Sync GitHub main 4645f4137666463023fa501984a14336dd2fc666

CITATION.cff

@@ -0,0 +1,13 @@
+cff-version: 1.2.0
+message: "Cite the MHRN software and the Recursive Epistemics publication separately."
+title: "Multi-Scale Homeostatic Recurrence Network (MHRN)"
+abstract: "Mehrskaliges homöostatisches Rekurrenznetzwerk. A Spiking Neural Architecture with Topological Plasticity."
+type: software
+version: "0.6.0a5"
+authors:
+  - family-names: Heisig
+    given-names: Thomas
+    orcid: "https://orcid.org/0009-0002-9589-1872"
+repository-code: "https://github.com/Thomas-Heisig/MHRN"
+url: "https://osf.io/p34uq/"
+license: MIT

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025-2026 Thomas Heisig and contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

MIRROR_MANIFEST.json

@@ -0,0 +1,8 @@
+{
+  "canonical_repository": "https://github.com/Thomas-Heisig/MHRN",
+  "canonical_sha": "4645f4137666463023fa501984a14336dd2fc666",
+  "generated_by": ".github/workflows/sync-huggingface.yml",
+  "rolling_mirror": true,
+  "raw_experiment_data_included": false,
+  "raw_data_policy": "Use canonical GitHub artefacts or dedicated DOI-bearing dataset deposits."
+}

README.md

@@ -0,0 +1,24 @@
+---
+license: mit
+language:
+  - en
+  - de
+tags:
+  - spiking-neural-networks
+  - neuroscience
+  - reproducible-research
+  - research-data
+  - experiments
+  - provenance
+pretty_name: MHRN Research Data
+---
+
+# MHRN Research Data
+
+This dataset repository is an automatically generated public mirror of the `research/` tree from the canonical MHRN GitHub repository.
+
+Canonical source: https://github.com/Thomas-Heisig/MHRN
+
+The mirror exists for discovery and reproducibility. It does **not** promote experiment DATA to accepted EVID. Historical experiment artefacts retain their original status, provenance and review boundaries.
+
+For citable experiment-specific research objects, use a dedicated DOI-bearing archive when one is explicitly published. Do not treat this rolling mirror as a substitute for immutable release or dataset DOIs.

research/CURRENT_SCIENTIFIC_STATE.md

@@ -0,0 +1,194 @@
+# Current Scientific State
+
+**Stand:** 19. September 2026
+
+Dieses Dokument ist der kurze Einstieg in den aktuellen wissenschaftlichen Zustand. Historische DATA, EVID-Entscheidungen und eingefrorene Publikationen bleiben unverändert in ihren datierten Verzeichnissen.
+
+## Methodischer Status der KI-Nutzung
+
+MHRN wird als **assistierte Einzelautorschaft mit verteilter KI-Unterstützung** entwickelt. KI-Systeme werden für Recherchehinweise, Gegenargumente, Strukturierung, Codeentwürfe, Tests, Fehlersuche und Review-Vorschläge eingesetzt. Diese Nutzung wird nicht nur offengelegt, sondern ist selbst Teil des methodischen Forschungsprogramms: Die Trennung von Vorschlag, Prüfung, Freigabe, Ausführung, Human Review, EVID und Veröffentlichung wird als prüfbarer Prozess behandelt.
+
+Die KI ist dabei weder wissenschaftlicher Autor noch Quellen- oder Evidenzautorität. Thomas Heisig trägt die menschliche Verantwortung für Auswahl, Freigabe, Behauptungen und Veröffentlichung.
+
+Die externe Geltung bleibt offen: interne Tests, Human Reviews, EVID-Registrierungen, GitHub-Releases und DOI-Zuweisungen ersetzen keine unabhängige Replikation oder externe fachliche Kritik. Der öffentliche Replikationsaufruf unter `INDEPENDENT_REPLICATION.md` ist daher ein zentraler nächster Schritt. Auch negative oder Null-Replikationen werden als wissenschaftlich relevante Ergebnisse akzeptiert.
+
+
+Die Dashboard-Archivierung ist metadata-only: Einzelne Experimente und
+Experimentreihen können aus der Arbeitsansicht ausgeblendet und wiederhergestellt
+werden, ohne kanonische DATA-, Manifest- oder Workflow-Artefakte zu verschieben
+oder wissenschaftlich umzuschreiben.
+Teilweise archivierte Reihen werden im Dashboard als `partial` mit archivierten
+und gesamten Kinderexperimenten ausgewiesen und aus der aktiven Reihenliste
+entfernt.
+Das Archiv ist im Frontend standardmäßig geschlossen und bietet eine gemeinsame
+Suche für archivierte Reihen und Einzelexperimente.
+
+Die öffentliche Projektattribution ist mit ORCID `0009-0002-9589-1872` für
+Thomas Heisig verknüpft. Diese Autorenidentität ist von technischen Runtime-
+Profilen getrennt; private Kontaktdaten aus dem ORCID-Datensatz werden nicht
+gespeichert.
+Für GitHub-Release-Archive ist `.zenodo.json` mit derselben öffentlichen
+ORCID-Creator-Verknüpfung hinterlegt. Ein Zenodo-Record oder DOI gilt erst nach
+externer Prüfung als vorhanden und wird derzeit nicht behauptet.
+Das öffentliche OSF-Projekt ist als Forschungs-/Provenienzressource unter
+`https://osf.io/p34uq/` verknüpft.
+
+Der aktuelle Hugging-Face-Publikationsnamespace ist `ThomasHeisig`; die
+historischen `superdigger`-Pfade bleiben als Legacy-Provenienz erhalten.
+
+## Aktuelle Publikation
+
+- **Recursive Epistemics / Rekursive Epistemik, Edition 1.8 — current WIP**
+- Einstieg: `research/publications/2026-09-17_recursive-epistemics_v1.8/MANUSCRIPT.md`
+- unmittelbarer Vorgänger: Edition 1.7
+- frozen empirical baseline: Edition 1.5 / `EXP-EMP-20260913-A3`
+- 1.8 nutzt die für 2.0 geplante elfteilige Struktur, ohne Softwareversion oder Evidenzstatus hochzustufen
+- offene menschliche wissenschaftliche Reviews: `EXP-GEN-0041`,
+  `EXP-S1-TOPO-V3-R1-20260918` und `EXP-S6-SEM-CL-003`
+- CL-003 besitzt jetzt einen begrenzten Review-Kandidaten; die Entscheidung ist weiterhin `PENDING`
+- unabhängige Replikation: unvollständig
+- automatische EVID-Promotion: deaktiviert
+
+## Forschungs- und Dokumentgovernance
+
+Der normative Vertrag fuer Aufbau, Nutzung und Interpretation seed-gebundener
+DATA steht unter
+[`research/specifications/SEED_DATA_CONTRACT.md`](specifications/SEED_DATA_CONTRACT.md).
+Er ist gemeinsam mit der jeweiligen Preregistration, Experiment-ID,
+Bedingung und den DATA-/Digest-Artefakten zu zitieren.
+
+`docs/` und `research/` werden durch `scripts/audit_document_governance.py` nach Typ, Status, Autorität, Mutabilität, Zitierregel und Evidenzrolle geprüft. Edition 1.8 ergänzt einen deterministischen Publikationsbuilder, vollständiges gepinntes Datei-/Abschnittsinventar, Research-Object-Projektionen, Creation-/Edition-/Experimentgenealogien, Vorgängerforschungs-Mapping und Quellenmetadaten. Diese Projektionen ersetzen keine Primärartefakte und zertifizieren keine semantische Vollständigkeit.
+
+## Stage 0 — korrigierter wissenschaftlicher Reifestand
+
+Die bisherige Scientific-Timeline-Projektion von 50 % war gegenüber dem aktuellen Repository veraltet. Für die Einzelzell-Konformität existieren inzwischen `RQ-EVAL-006` sowie die drei expliziten Hypothesen `H-EVAL-006-A/B/C`, ein vor der Ausführung eingefrorenes konfirmatorisches Protokoll (`PREREG-EVAL-006-V2`) und quellengebundene Confirmatory-DATA aus `EXP-STAGE0-20260916-MODEL-CONFORMANCE-V2`.
+
+Der eng definierte `single_neuron_scientific_readiness`-Vertrag steht weiterhin auf **100 %**. Dieser Wert beschreibt nur die Erfüllung seines scoped Readiness-Vertrags. Für die breitere Scientific-Maturity-Timeline gelten zusätzlich Human-EVID und unabhängige Replikation. Nach dem kanonischen Gewichtungsmodell ergibt sich aktuell:
+
+- RQ/H: **met** = 15 %;
+- eingefrorenes Protokoll: **met** = 20 %;
+- source-bound DATA: **met** = 20 %;
+- Human-reviewed EVID: **met** = 20 % von 20 % (Human Review und kanonische EvidenceEngine-Promotion über `EVID-2026-18` abgeschlossen);
+- externer Referenzvergleich / unabhängige Replikation: **partial** = 7,5 % von 15 %;
+- Attribution: **met** = 10 %.
+
+Die Human Review des scoped Claims ist inzwischen abgeschlossen und unterstützt die enge Konformitätsaussage. Eine EVID-Promotion wurde trotzdem **nicht** erzwungen: Der historische Stage-0-Lauf besitzt kein EvidenceEngine-kompatibles `manifest.json` mit aufgezeichneter Validity, Clean-Tree-Status, `provenance_digests` und `source_freeze_sha`. Diese fehlenden historischen Provenienzfelder werden nicht rückwirkend erfunden.
+
+Damit beträgt die aktuelle **Gesamt-Scientific-Maturity von Stage 0 = 92,5 %**. Der historische V2-Lauf bleibt wegen seines alten Provenienzvertrags selbst nicht promotion-eligible; diese Lücke wurde jedoch prospektiv durch `EXP-STAGE0-20260918-MODEL-CONFORMANCE-V2-PROMO-R1` geschlossen. Nach dokumentiertem Human Review wurde der scoped Claim über den aktuellen EvidenceEngine-Pfad als `EVID-2026-18` registriert. Das 20-%-Kriterium `reviewed_evidence` steht damit auf `met`. Das unabhängige Replikationskriterium bleibt separat `partial` und trägt weiterhin nur 7,5 von 15 Prozentpunkten.
+
+Die Human Review ist unter `research/experiments/EXP-STAGE0-20260916-MODEL-CONFORMANCE-V2/human_scientific_review.json` dokumentiert. Der separate Status `EVIDENCE_PROMOTION_STATUS.json` hält fest, dass der Review den scoped Claim unterstützt, die EVID-Promotion wegen der fehlenden historischen EvidenceEngine-Provenienz aber blockiert bleibt. Weder Review noch spätere Promotion dürfen biologische Gleichwertigkeit oder unabhängige Replikation erzeugen.
+
+Diese formale Provenienzlücke wurde inzwischen **prospektiv** adressiert, nicht rückwirkend repariert: `EXP-STAGE0-20260918-MODEL-CONFORMANCE-V2-PROMO-R1` wurde unter `PREREG-EVAL-006-V2-PROMO-R1` mit den vorab deklarierten neuen Seeds `22001–22003` ausgeführt. Der wissenschaftliche Lauf am Freeze-Commit `0376142b16a092c674653ee61c10d8137671eacf` war clean-tree, `validity.valid=true`, ohne Runtime-/Fatal-Fehler und mit vollständigen `provenance_digests` sowie `source_freeze_sha=36764dde292a3934523b21c5d9559319a0d8c9a8763c9ab3dc1c5cbbc246e850`. Alle drei eingefrorenen Hypothesenbedingungen bestanden; der maximale Izhikevich-Lokalfehler betrug `4.5474735088646412e-13`, der maximale LIF-Fehler `7.1054273576010019e-15`.
+
+Der Promotion-Lauf selbst bleibt als ausgeführte DATA unverändert; seine automatische Promotion ist weiterhin deaktiviert. Der Human Review durch Thomas Heisig vom 18.09.2026 wurde zusätzlich in das vom EvidenceEngine verlangte kanonische `human_review.json` abgebildet. Die anschließende getrennte Promotion erzeugte `EVID-2026-18` für `CLAIM-EVAL-006`. Diese Registrierung ändert weder die historische V1-Negativspur noch die Unabhängigkeitsgrenze: `independent_authorship_replication=false` bleibt bestehen.
+
+## Aktueller Schwerpunkt: Stage 6
+
+### Semantization / Continual Learning
+
+- `EXP-S6-SEM-CL-001`: positiver Mechanismusbefund gegenüber No-Replay unter seinem eingefrorenen Protokoll.
+- `EXP-S6-SEM-CL-002`: kein bestätigter Vorteil semantischer Prototypen gegenüber gematchtem Raw-Replay; H1 wurde unter diesem Protokoll in der menschlichen Projekt-EVID als falsifiziert klassifiziert.
+- `EXP-S6-SEM-CL-003`: genau einmal autorisierte, freeze- und hash-gebundene Kampagne mit 84 Runs = 12 Seeds × 7 Bedingungen; Runner-Klassifikation `H1_negative_H2_negative`.
+- C1, C2 und C4 erfüllen die präregistrierten Erfolgsregeln nicht. C3 (`S20 − X20`) ist positiv und zeigt relevante nicht-zufällige Struktur, aber keinen bestätigten Vorteil gegenüber Raw-Replay.
+
+Die aktuelle DATA-only-Zwischenbilanz lautet:
+
+> Unter den bisher untersuchten Bedingungen liegt der nachweisbare Beitrag primär im Replay. Die semantische Verdichtung erhält relevante Struktur, zeigt aber bislang keinen präregistriert bestätigten Zusatznutzen gegenüber gematchtem Raw-Replay.
+
+CL-003 bleibt **DATA, nicht EVID**, bis die Human Review abgeschlossen ist. Die historische Detailbilanz bleibt unter `research/publications/2026-09-15_recursive-epistemics_v1.7/STAGE6_CL001_CL003_BALANCE.md` erreichbar und wird von 1.8 nicht umgeschrieben.
+
+### Entscheidungsgate für SemanticMemory
+
+Aus CL-003 folgt kein automatisches CL-004. Nach Human Review gilt weiterhin:
+
+- **A:** spezialisierte Nebenrolle; Raw-Replay bleibt Referenz; keine weitere Semantikprüfung.
+- **B:** genau eine theoretisch begründete alternative Rolle in genau einem neuen präregistrierten konfirmatorischen Experiment; bei negativem Ergebnis folgt A.
+- **C:** Rollenfrage parken, bis Prediction Error oder World Model eine konkrete funktionale Notwendigkeit erzeugen.
+
+Eine serielle Rettung durch immer neue Rollenannahmen ist nicht zulässig.
+
+## Aktuelle Review- und Replikationslage
+
+### Determinismus / `RQ-DET-001`
+
+Der historische Lauf `EXP-BATCH-20260914074039-03` ist nach dem aktuellen semantischen Vertrag ein direkter Test von `RQ-DET-001 / H-SNN-003-A`. Für drei Seeds stimmen die A/B-Replikapaare innerhalb von `recurrence_off` und `recurrence_on` in den registrierten Antwortgrößen überein. Der Befund ist damit ein positiver Same-Seed-Determinismusbefund **innerhalb dieses kleinen Protokolls**.
+
+Er bleibt dennoch **nicht EVID**, weil der historische Lauf mit `git dirty: true` erzeugt wurde. Die semantische Korrektur heilt keine Provenienzlücke. Der nächste wissenschaftlich saubere Schritt ist deshalb ein clean-tree, hash-gebundener Replikationslauf mit denselben vorab fixierten Bedingungen und anschließender Human-EVID-Entscheidung.
+
+Zusätzlich gilt die AIRR-Regel: Bei semantischem `MISMATCH` wird öffentliches/report-level `ai_confidence` deterministisch auf `0.0` gesetzt; der ursprüngliche Modellwert bleibt nur im append-only Auditpfad erhalten. AIRR bleibt Interpretation, nicht Evidenzinstanz.
+
+### Topologie / `H-SNN-003-B`
+
+`EXP-GEN-0047` bleibt als methodisch wichtiger Vorgänger erhalten: technisch reproduzierbar und semantisch `DIRECT_MATCH`, aber wegen drei Neuronen, zwei Feed-forward-Synapsen und fehlender kausaler Geometrie-zu-Graph-Kopplung `INADEQUATE_TO_TEST_HYPOTHESIS`. Die damaligen identischen Resultate über 1D/2D/3D/5D sind deshalb weder Topologie-Nullbefund noch Widerlegung eines 5D-Effekts.
+
+Der nachfolgende, vor Ausführung präregistrierte Lauf `EXP-S1-TOPO-V2-20260918` / `topology_propagation_v2` adressiert **`RQ-SNN-003 / H-SNN-003-B` als Stage-1-Topologiefrage**, nicht die stärkere dimensionsspezifische 5D-Hypothese. Das Design verwendet 64 Neuronen und 246 Kanten je Evaluationsbedingung, identische Gewichte/Delays/Stimulation, sechs vorab definierte Topologien, getrennte Kalibrier- und Evaluations-Seeds sowie ein Activity-Adequacy-Gate. Das Gate bestand bereits bei Synapsengewicht 55.0; anschließend wurden 120 Evaluationsläufe über 20 gepaarte Seeds ausgeführt. Designintegrität und deterministische lokale Neuberechnung bestanden.
+
+Unter diesem eingefrorenen Regime wurden präregistrierte Unterschiede in `active_fraction` und/oder `first_output_latency_censored` zwischen mehreren Topologiepaaren beobachtet; die Holm-korrigierten exakten gepaarten Sign-Tests tragen die primäre Inferenz. Die deterministischen Bootstrap-95%-Intervalle beschreiben die gepaarten Medianunterschiede zusätzlich, sind bei vollständig identischen gepaarten Differenzen aber erwartungsgemäß degeneriert und daher kein eigenständiger zusätzlicher Evidenzbeitrag.
+
+Für `active_fraction` liegt eine klare **Messbereichsgrenze im getesteten Regime** vor: 1d, 2d und 3d erreichen in der Evaluation jeweils eine Median-Aktivierungsfraktion von 1,0; die Kontraste 1d→2d und 2d→3d besitzen deshalb Median-Differenz 0, Bootstrap-CI [0,0] und Sign-Test p=1. Das ist kein belastbarer Nachweis der Gleichheit dieser Topologien, sondern zeigt, dass dieser Endpunkt bei den drei Bedingungen im gewählten Stimulus-/Gewichtsregime sättigt. Die diskriminierende Information von `active_fraction` beginnt erst bei den Kontrasten zu 5d beziehungsweise innerhalb der 5d-/Kontrollfamilie.
+
+Der Name `first_output_latency_censored` darf nicht als tatsächliche Zensierung der beobachteten 1d-Latenzen missverstanden werden. Das Präregistrat setzt ein Evaluationsfenster von 128 Ticks; nur ein vollständig ausbleibender Output wird mit dem Sentinel `129 = ticks + 1` kodiert. Die gespeicherten Evaluationsdaten enthalten keine solche Sentinel-Kodierung. Der 1d-Median von 19 Ticks ist daher ein beobachteter, nicht zensierter Wert. Die Latenzkontraste tragen in diesem Datensatz den breiteren und über alle fünf Primärkontraste konsistenten Unterschied.
+
+Alle sechs Bedingungen besitzen dasselbe globale Kantenbudget von 246. Unterschiede zwischen 1d/2d/3d/5d/`5d_shuffled`/`random_graph` sind deshalb **keine simple Dichte- oder Sparsity-Differenz**, sondern entstehen aus der unterschiedlichen Anordnung der gematchten Kanten unter den jeweiligen Konstruktionsregeln. Der zulässige Befund bleibt eng: **Die konkrete Netzwerktopologie beeinflusst in diesem kontrollierten 64-Neuronen-Regime die Propagationsdynamik.** Der Laufstatus lautet `SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL`; die begrenzte Interpretation ist human-reviewed akzeptiert, bleibt aber DATA und erzeugt keine automatische EVID-Promotion.
+
+Die anschließende interne Replikationslinie wurde bewusst als neue DATA erzeugt. `EXP-S1-TOPO-V3-20260918` führte zwar erfolgreich 120 Läufe aus und zeigte bereits dieselbe Richtung, musste aber für die konfirmatorische Interpretation verworfen werden, weil der Runner die Holm-Korrektur fälschlich in zwei Familien zu je fünf Tests statt in der präregistrierten einen Familie von zehn zeitaufgelösten Primärtests anwendete. Die Rohdaten und Originalstatistik bleiben unverändert als Auditspur erhalten.
+
+Die korrigierte Neuausführung `EXP-S1-TOPO-V3-R1-20260918` verwendet neue Seeds 6201–6220 und hält 64 Neuronen, 246 Kanten, Synapsengewicht 55.0, Stimulus und 128-Tick-Fenster konstant. Die eine Holm-Familie über alle zehn Primärtests ist nun korrekt implementiert und deterministisch verifiziert. Beide vorab definierten Ceiling-Resolution-Kriterien bestehen: `activation_auc_0_32` trennt 1d→2d mit Median-Δ +4,4140625 (CI95 [4,3671875; 4,4296875], Holm-p ≈ 1,91×10^-5) und 2d→3d mit +1,4609375 (CI95 [1,3671875; 1,4921875], Holm-p ≈ 1,91×10^-5). Die Halbaktivierungslatenz fällt entsprechend von Median 10 auf 6 auf 4 Ticks; die beiden low-dimensionalen Kontraste sind signifikant. Damit ist gezeigt, dass die terminale `active_fraction=1.0`-Sättigung von V2 die zeitliche Propagationsdynamik zwischen 1d/2d/3d verdeckte, nicht dass diese Topologien dynamisch gleich waren.
+
+Zusätzlich repliziert R1 alle fünf V2-Kontraste der `first_output_latency_censored` auf neuen Seeds mit derselben negativen right-minus-left-Richtung und Holm-korrigierter Signifikanz. Die Median-Latenzen bleiben 19, 9, 6, 5, 2 und 2 Ticks für 1d, 2d, 3d, 5d, `5d_shuffled` und `random_graph`. Der zulässige Status lautet deshalb `SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL` auf DATA-Ebene. Dies ist eine **interne Replikation innerhalb derselben Code-/Modellfamilie**, keine unabhängige externe Replikation und keine EVID-Promotion.
+
+Davon getrennt bleibt `RQ-5D-005 / H-5D-005-A` offen/untested. Für einen dimensionsspezifischen 5D-Claim gelten weiterhin die stärkeren Mindestanforderungen von mindestens 1.000 Neuronen pro Bedingung, im Mittel mindestens 10 eingehenden Synapsen, explizit distanzabhängiger Konnektivität, degree-/density-matched Kontrollen, Multi-Neuron-Stimulus, unabhängigen Seeds und sauberer Source-/Graph-Provenienz. Dass `5d_shuffled` und `random_graph` im Stage-1-Lauf teils frühere Output-Latenzen zeigten, unterstreicht gerade, dass aus `H-SNN-003-B` **keine 5D-Überlegenheit** abgeleitet werden darf.
+
+### Stage-6-Kompressionshypothese — nächster vorbereiteter empirischer Schritt
+
+`LP-20260917194217 / OBJ-MEM-COMPRESSION-001` ist human-origin, genehmigt und weiterhin **proposal-only**. Es wurde nicht ausgeführt und besitzt keine Runtime-Autorität. Die neue Frage prüft **Kompression, nicht allgemeine Lernleistung**: `semantic_prototype_replay_10pct_budget` soll bei einem Zehntel des Speicherbudgets mindestens 95 % der Retention von `raw_replay_full_budget` erreichen. Kontrollen sind `no_replay`, `random_prototype_10pct` und `learning_off`.
+
+Die vorab festzulegende Rollenentscheidung lautet: Erfolg stützt eine begrenzte Kompressionsrolle von SemanticMemory unter diesem Protokoll; Misserfolg stützt diese Rolle nicht. Generalisierung, Langzeitgedächtnis und Weltmodell-Brücke bleiben davon getrennte spätere Hypothesen und dürfen den Kompressionstest nicht post hoc retten.
+
+Das historisch genehmigte `LP-20260917194217` bleibt unverändert: seine ursprüngliche Genehmigung enthält noch den damaligen Source-Platzhalter und darf nicht nachträglich umgeschrieben werden. Die Quellprovenienz ist stattdessen in der neuen Proposal-Revision `LP-20260917194217-R1` konkret gebunden: `CL-002-EVID` verweist auf das human-reviewte `EVID.json` mit SHA-256 `c7b1124256fd8018839a8c5c29b30493b68d16bc223d0a3258bcbaca55b1752e`; `CL-003-DATA` verweist auf `results.json` mit SHA-256 `4e74021c0ef838371a3a01061ae1c2dcebb2031b54c15c6169e38c1972d06290`. Beide Sources stehen dort auf `VERIFIED`. Weil sich der Proposal-Inhalt geändert hat, ist R1 **nicht durch die historische Approval gedeckt und benötigt eine neue explizite menschliche Genehmigung**. Vor einer Ausführung bleiben zusätzlich Seed-/Taskplan, Analysevertrag, Ausführungsautorisation und Freeze vollständig zu binden. Approval ist keine Ausführung, DATA oder EVID.
+
+Der nächste methodische Schritt ist die Präregistrierungsvorbereitung von R1. Vor einem `FROZEN`-Status sind kanonische RQ/H-Zuordnung, Speicherbudget-Definition, Seed-/Taskplan, Retentionsaggregation, Äquivalenz-/Inferenzregel, Ausschluss-/Failure-Regeln, Analysevertrag und Source-/Config-Hashes zu binden; R1-Human-Approval und Ausführungsautorisation bleiben separate Gates.
+
+### Methodische Gesamtfolge
+
+Die jüngsten Reviews machen eine für MHRN zentrale Trennung verbindlich:
+
+`semantischer Match ≠ technische Reproduzierbarkeit ≠ Testadäquanz ≠ Provenienz ≠ EVID`.
+
+Ein Experiment kann korrekt registriert und technisch reproduzierbar sein und dennoch die Zielhypothese nicht beantworten. Umgekehrt kann ein positiver technischer Befund wegen mangelhafter Provenienz von einer Evidenzpromotion ausgeschlossen bleiben. Diese Trennung ist inzwischen Teil des kanonischen Forschungsprozesses.
+
+## Weitere priorisierte Forschung
+
+- Stage 0: Modell-/Integrator-/Parameterkonformität und externe Referenzreplikation.
+- Stages 2/3: Rekurrenz, Störungsrobustheit, held-out Plastizität, Langzeit-/Ressourcenstabilität.
+- Stage 4: modality-specific vs. general controls, Cross-Modal-Transfer, Läsion/Shuffle/Frozen.
+- Stage 5: Closed Loop, Real Device, Sensor Loss und Actuator No Effect unter Safety-Gates.
+- Stage 6: neuronaler Prediction Error, Mehrschritt-/action-conditioned World Model, decision benefit.
+- Spiegel-/Handlungsprädiktion: Stage-4/5/6-Querschnitt; Stage 7 erst bei kausaler Self/Other-Differenzierung.
+- 5D: gematchte Dimensions-/Geometrieablation; keine etablierte Überlegenheit.
+- biophysikalische Detailmodelle: nur als fragegetriebene Varianten/Ablationen.
+- Safety: Zielprovenienz, Misgeneralization, Corrigibility, Interruptibility, Specification Gaming und Post-Objective Transition.
+- Prior Art, unabhängige Replikation und externer Quellen-/Similarity-Audit.
+
+## Potenzielle Beiträge / Neuheitsstatus
+
+Edition 1.8 behandelt unter anderem folgende Punkte als **Kandidaten**, nicht als bewiesene Neuheit:
+
+1. Logical Identity / Physical Slot / Synaptic Reduction / Execution Scheduling;
+2. Proposal → Approval → Mutation → Journal → Undo;
+3. Content Gateway / Compute Backend;
+4. source-bound DATA/EVID-Grenzen;
+5. rekursive Epistemik als explizit provenance- und gatewayorientierte Verbindung von Forschungsobjekt und Forschungsprozess.
+
+Die externe wissenschaftliche Neuheit bleibt `requires_prior_art_review`.
+
+## Integrität
+
+- `research/publications/2026-09-17_recursive-epistemics_v1.8/REFERENCES.md`
+- `research/publications/2026-09-17_recursive-epistemics_v1.8/EXTENDING.md`
+- `research/INTEGRITY_AND_ATTRIBUTION.md`
+- `research/RELATED_WORK.md`
+- `docs/05-quality/RESEARCH_INTEGRITY_GATE.md`
+- `docs/00-governance/DOCUMENT_GOVERNANCE.md`
+
+Vor formaler externer Einreichung bleiben menschliche Quellenprüfung, systematische Prior-Art-Prüfung, externe Similarity-Prüfung und fachliche Review erforderlich.

research/EXP-RETRY-AIRR-2026-0001.execution.json

@@ -0,0 +1,127 @@
+{
+  "workflow_id": "EXP-RETRY-AIRR-2026-0001",
+  "report": "workflows/EXP-RETRY-AIRR-2026-0001.json",
+  "report_markdown": "workflows/EXP-RETRY-AIRR-2026-0001.md",
+  "completed": 2,
+  "failed": 0,
+  "results": [
+    {
+      "protocol": "independent_replication_v1",
+      "ticks": 256,
+      "seeds": "101-120",
+      "status": "completed",
+      "experiment_id": "EXP-RETRY-AIRR-2026-0001-01",
+      "manifest": "experiments/EXP-RETRY-AIRR-2026-0001-01/manifest.json",
+      "report": "experiments/EXP-RETRY-AIRR-2026-0001-01/report.md",
+      "workflow": "experiments/EXP-RETRY-AIRR-2026-0001-01/workflow.json",
+      "statistics": "experiments/EXP-RETRY-AIRR-2026-0001-01/analysis/statistics.json",
+      "ai_packet": "experiments/EXP-RETRY-AIRR-2026-0001-01/analysis/ai_packet.json",
+      "raw_run_index": "experiments/EXP-RETRY-AIRR-2026-0001-01/DATA/runs_index.json",
+      "data_id": "DATA-EXP-RETRY-AIRR-2026-0001-01",
+      "ai_report": {
+        "status": "unavailable",
+        "reason": "AI backend not configured"
+      },
+      "summary": "experiments/EXP-RETRY-AIRR-2026-0001-01/summary.md",
+      "result": {
+        "run_count": 40,
+        "duration_seconds": 0.1597151259999947,
+        "runner": "run_replication",
+        "ticks_requested": 256,
+        "seeds_executed": [
+          101,
+          102,
+          103,
+          104,
+          105,
+          106,
+          107,
+          108,
+          109,
+          110,
+          111,
+          112,
+          113,
+          114,
+          115,
+          116,
+          117,
+          118,
+          119,
+          120
+        ],
+        "tick_validation": {
+          "status": "SATISFIED",
+          "mode": "minimum_per_run",
+          "requested_ticks": 256,
+          "observed_min": 256,
+          "observed_max": 256
+        }
+      },
+      "epistemic_layers": {
+        "ui_state": "operator/dashboard progress only; not a scientific result",
+        "data": "DATA artifacts and deterministic statistics are authoritative run outputs",
+        "evid": "not_created; requires semantic validation, clean freeze and human review",
+        "interpretation": "post-hoc AI or human interpretation; never execution input"
+      }
+    },
+    {
+      "protocol": "learning_interference_screen_v1",
+      "ticks": 1000,
+      "seeds": "101-120",
+      "status": "completed",
+      "experiment_id": "EXP-RETRY-AIRR-2026-0001-02",
+      "manifest": "experiments/EXP-RETRY-AIRR-2026-0001-02/manifest.json",
+      "report": "experiments/EXP-RETRY-AIRR-2026-0001-02/report.md",
+      "workflow": "experiments/EXP-RETRY-AIRR-2026-0001-02/workflow.json",
+      "statistics": "experiments/EXP-RETRY-AIRR-2026-0001-02/analysis/statistics.json",
+      "ai_packet": "experiments/EXP-RETRY-AIRR-2026-0001-02/analysis/ai_packet.json",
+      "raw_run_index": "experiments/EXP-RETRY-AIRR-2026-0001-02/DATA/runs_index.json",
+      "data_id": "DATA-EXP-RETRY-AIRR-2026-0001-02",
+      "ai_report": {
+        "status": "unavailable",
+        "reason": "AI backend not configured"
+      },
+      "summary": "experiments/EXP-RETRY-AIRR-2026-0001-02/summary.md",
+      "result": {
+        "run_count": 20,
+        "duration_seconds": 8.160896397999991,
+        "runner": "run_learning_interference",
+        "ticks_requested": 1000,
+        "seeds_executed": [
+          101,
+          102,
+          103,
+          104,
+          105,
+          106,
+          107,
+          108,
+          109,
+          110,
+          111,
+          112,
+          113,
+          114,
+          115,
+          116,
+          117,
+          118,
+          119,
+          120
+        ],
+        "tick_validation": {
+          "status": "NOT_APPLICABLE",
+          "mode": "protocol_defined_internal_trials",
+          "requested_ticks": 1000
+        }
+      },
+      "epistemic_layers": {
+        "ui_state": "operator/dashboard progress only; not a scientific result",
+        "data": "DATA artifacts and deterministic statistics are authoritative run outputs",
+        "evid": "not_created; requires semantic validation, clean freeze and human review",
+        "interpretation": "post-hoc AI or human interpretation; never execution input"
+      }
+    }
+  ]
+}

research/INTEGRITY_AND_ATTRIBUTION.md

@@ -0,0 +1,160 @@
+# MHRN Research Integrity and Attribution Policy
+
+**Status:** project policy / pre-submission control
+
+**Applies to:** source code, documentation, research reports, dissertation/manuscript editions, figures, experiment protocols, AI-assisted drafts, benchmark notes and release notes.
+
+## 1. Purpose
+
+MHRN separates scientific progress from engineering progress. The same separation applies to research integrity: a technically reproducible artifact is not automatically an original scientific contribution, and a correctly cited source is not automatically evidence for an MHRN claim.
+
+This policy reduces plagiarism, text-recycling, citation, licensing and provenance risks. It does **not** certify that a manuscript is plagiarism-free. External similarity checking and human source review remain mandatory before submission.
+
+## 2. Mandatory claim-origin roles
+
+Every substantial scientific statement should be traceable to at least one of the following **claim-origin roles**. These roles classify the epistemic origin/function of a statement and are distinct from the S1-S4 source-proximity classes used by Edition 1.8 for historical/provenance reconstruction:
+
+1. **MHRN observation** — produced by a source-bound experiment or verification artifact.
+2. **MHRN interpretation** — inference from MHRN observations; must state limits and cannot be promoted automatically to EVID.
+3. **External theory or method** — idea, algorithm, model, benchmark or empirical finding from another source; cite the original source where possible.
+4. **MHRN prior work** — material from an earlier MHRN edition, report or public artifact; cite or disclose the earlier version when reused substantially.
+5. **AI-assisted wording or synthesis** — assistance is not an authority. Every factual or literature claim must be checked against a citable primary or authoritative source before publication.
+
+## 3. Code attribution
+
+For copied or adapted external code, record:
+
+- upstream project and exact source URL or release/tag,
+- license and license-compatible use,
+- original file/function or algorithmic source,
+- extent of modification,
+- date imported,
+- responsible reviewer.
+
+A dependency listed in a package manager is not sufficient attribution when project code is copied or adapted from it.
+
+Algorithmic implementations should cite the paper/specification that motivates the mechanism even when the implementation is independently written.
+
+## 4. Text recycling and self-citation
+
+MHRN contains multiple public manuscript editions and addenda. Reuse across those versions is expected during development, but publication-facing reuse must be transparent.
+
+Before external submission:
+
+- disclose the relationship to earlier public editions,
+- cite earlier MHRN work when it constitutes prior dissemination,
+- avoid presenting reused text, figures, tables or results as newly produced,
+- identify duplicated datasets/analyses and distinguish reanalysis from new experiments,
+- keep a change note when a paragraph is substantially inherited from an earlier edition.
+
+"Self-plagiarism" rules vary by venue; the project therefore treats undisclosed substantial reuse as an integrity risk rather than relying on a single universal threshold.
+
+## 5. Literature and idea attribution
+
+Related work must distinguish:
+
+- **background**: general context,
+- **mechanistic precedent**: a source that directly motivates an implemented mechanism,
+- **comparison baseline**: a system or method compared empirically,
+- **interpretive precedent**: a theory used to interpret results,
+- **unverified lead**: a possibly relevant source that has not yet been verified and must not be cited as settled fact.
+
+Whenever MHRN adopts a mechanism analogous to a published method, the manuscript must state both the precedent and the meaningful implementation differences.
+
+## 6. AI-assisted research safeguards
+
+AI-generated references, titles, DOIs, journal details and performance claims are treated as unverified until checked against a primary/authoritative source. AI-generated prose must not be used to disguise copied source text.
+
+Required checks before publication:
+
+1. resolve each citation to a real source;
+2. confirm author, title, year and venue;
+3. verify the claimed finding in the source;
+4. check that paraphrasing is genuinely independent;
+5. ensure direct quotations are explicitly marked and minimal;
+6. ensure the cited source actually supports the nearby claim.
+
+## 7. Similarity and plagiarism screening
+
+Repository checks can catch missing attribution files, duplicated local passages and obvious copied code patterns, but they cannot prove originality against the global literature.
+
+Before manuscript submission, perform:
+
+- a reputable text-similarity check accepted by the target institution or venue,
+- manual review of every high-similarity passage,
+- citation audit for paraphrases and conceptual borrowing,
+- code-origin review for imported/adapted code,
+- figure/table provenance review.
+
+Commercial tools such as iThenticate/Turnitin may be used if institutionally available; the project does not treat a similarity percentage alone as a plagiarism verdict.
+
+## 8. Scientific claim boundary
+
+No literature citation can substitute for an MHRN experiment. Conversely, an MHRN experiment cannot establish priority over an idea already present in prior literature.
+
+Claims must distinguish:
+
+- implemented,
+- technically verified,
+- DATA-supported,
+- human-reviewed EVID,
+- independently replicated.
+
+Only the latter states may justify progressively stronger scientific language, and none of them justify consciousness or biological-equivalence claims automatically.
+
+## 9. Required release artifacts
+
+For any release that changes scientific claims or mechanisms, maintain:
+
+- `research/RELATED_WORK.md`,
+- this policy,
+- the scientific stage manifest in `src/dashboard/static/scientific-progress.json`,
+- source-bound experiment/protocol references,
+- a publication/change note if manuscript claims changed.
+
+## 10. Pre-submission checklist
+
+- [ ] Every non-trivial external mechanism has a primary-source citation.
+- [ ] Every quantitative external claim has been verified in the cited source.
+- [ ] No unverified AI-generated citation remains.
+- [ ] Reused MHRN text/results are disclosed and cross-referenced.
+- [ ] Copied/adapted code has source and license provenance.
+- [ ] Figures and tables have explicit provenance.
+- [ ] Similarity screening completed and manually reviewed.
+- [ ] DATA and EVID are not conflated.
+- [ ] Negative/null results remain visible.
+- [ ] Release wording matches the actual scientific maturity stage.
+
+## 11. Non-guarantee statement
+
+Passing these controls means that documented integrity checks were completed. It does **not** mean that criticism is impossible or that plagiarism has been mathematically excluded. Scientific criticism remains part of the process; the project goal is to make claims traceable, bounded, falsifiable and properly attributed.
+
+## 12. RQ-ETH-001 contribution and accountability contract
+
+For Human–AI-assisted MHRN research, **contribution, generation, validation, selection/canonization, formal authorship and responsibility are separate dimensions**. A technical or linguistic contribution does not by itself establish authorship, personhood or responsibility.
+
+The canonical research question is `RQ-ETH-001`: *How are epistemic contributions, decision authority, formal authorship and scientific responsibility distributed in Human–AI-assisted MHRN research?*
+
+For every research-relevant claim episode, the project should be able to record at least:
+
+- `claim_id` and `event_id`,
+- `actor_type` and actor/model/software version,
+- role in conceptualization, generation/transformation, analysis, validation or selection,
+- input and output references,
+- whether the contribution is material to question, method, implementation, analysis, evidence assessment or conclusion,
+- decision authority and acceptance/rejection/revision state,
+- evidence/provenance reference,
+- responsible natural person,
+- timestamp and, where available, provenance digest.
+
+A contribution is **material** when it changes a research question, hypothesis, method, implementation, analysis, interpretation, evidence assessment or conclusion. Formatting, spelling correction and meaning-preserving surface rewriting are recorded separately.
+
+### Legacy boundary
+
+`H-ETH-001-A` retains its historical wording and identifier because existing experiment and audit artifacts already reference it. The 2026-09-18 consolidation therefore does **not** redefine that identifier retroactively. New propositions are registered as `H-ETH-001-B` through `H-ETH-001-E`.
+
+Historical generic runtime and boundary-audit runs remain provenance records, but they are not direct evidence for the new subhypotheses. Direct empirical testing requires claim-episode provenance, an explicit coding scheme and human/independent review.
+
+### Claim boundary
+
+This contract does not infer consciousness, moral personhood, legal personality, copyright authorship or autonomous scientific responsibility from machine participation. Those questions require separate legal, ethical or consciousness-specific analysis.

research/README.md

@@ -0,0 +1,106 @@
+# MHRN Research — kanonischer wissenschaftlicher Einstieg
+
+**Stand:** 17. September 2026  
+**Aktuelle wissenschaftliche Arbeitsfassung:** Recursive Epistemics **1.8 WIP**  
+**Frozen empirical baseline:** Recursive Epistemics **1.5**  
+**Grundregel:** `implementation test != DATA != reviewed EVID != interpretation`
+
+Dieses README ist ein Navigations- und Autoritätsindex. Historische DATA, EVID-Entscheidungen, feste Testzahlen und frühere Publikationsstände bleiben erhalten und werden nicht rückwirkend an die aktuelle Interpretation angepasst.
+
+## Aktuelle Publikation
+
+- [CURRENT](publications/CURRENT.md)
+- [Vollmanuskript 1.8](publications/2026-09-17_recursive-epistemics_v1.8/MANUSCRIPT.md)
+- [Elfteilige 1.8-Struktur](publications/2026-09-17_recursive-epistemics_v1.8/README.md)
+- [Forschungsfragen/Hypothesen/Claims](publications/2026-09-17_recursive-epistemics_v1.8/RESEARCH_REGISTER.md)
+- [Vorgängerforschung](publications/2026-09-17_recursive-epistemics_v1.8/PRIOR_WORK_MAP.md)
+- [Quelleninventar](publications/2026-09-17_recursive-epistemics_v1.8/SOURCE_INDEX.md)
+- [Literatur](publications/2026-09-17_recursive-epistemics_v1.8/REFERENCES.md)
+- [Vorgänger 1.7](publications/2026-09-15_recursive-epistemics_v1.7/README.md)
+- [Vorgänger 1.6](publications/2026-09-15_recursive-epistemics_v1.6/README.md)
+- [Frozen 1.5](publications/FROZEN_V1.5.md)
+- [Publikationskatalog](publications/catalog.json)
+
+Der Publication Viewer folgt `publications/catalog.json`; das aktuelle Reader-Ziel ist Edition 1.8 `MANUSCRIPT.md`.
+
+## Aktueller wissenschaftlicher Status
+
+- [Current Scientific State](CURRENT_SCIENTIFIC_STATE.md)
+- [Seed-Datenvertrag](specifications/SEED_DATA_CONTRACT.md) — Aufbau, Nutzung und Interpretation seed-gebundener DATA
+- [Scientific-Maturity-Manifest](../src/dashboard/static/scientific-progress.json)
+- [Scientific Maturity Roadmap](../docs/08-roadmap/SCIENTIFIC_MATURITY_ROADMAP.md)
+- [Research Integrity Gate](../docs/05-quality/RESEARCH_INTEGRITY_GATE.md)
+- [Integrity & Attribution](INTEGRITY_AND_ATTRIBUTION.md)
+- [RQ-ETH-001 Provenance Study](protocols/RQ_ETH_001_PROVENANCE_STUDY.md) — Beitrags-/Verantwortungsmatrix; Designstatus, noch nicht präregistriert
+- [RQ-EPIST-002 Process Governance Study](protocols/RQ_EPIST_002_PROCESS_GOVERNANCE_STUDY.md) — Vergleich getrennte Status-/Provenienzpakete vs. abgeflachte Summary; Designstatus, noch nicht präregistriert
+- [Related Work](RELATED_WORK.md)
+- [Open Science / Research Networks](../OPEN_SCIENCE.md) — DOI-, ORCID-, OSF-, Preprint- und Discovery-Routing
+- [Call for Independent Replication](../INDEPENDENT_REPLICATION.md) — öffentliche Replikationsziele und Anforderungen an Unabhängigkeit
+- [Paper Offshoots — Kandidaten für eigenständige Fachbeiträge](paper_offshoots/README.md) — Planung, keine DATA/EVID
+- [Wissenschaftliche Bilanz Edition 1.8](publications/2026-09-17_recursive-epistemics_v1.8/SCIENTIFIC_BALANCE.md) — Claim-Ledger, Vollständigkeitsgrenze und offene wissenschaftliche Bilanz
+
+Engineering-Reife und wissenschaftliche Reife werden getrennt geführt. Kein Stage-Score ist eine Kognitions-, Intelligenz- oder Bewusstseinskennzahl.
+
+## Kanonische Forschungsobjekte
+
+| Bereich | Rolle | Autorität |
+|---|---|---|
+| `registry/` | RQ, Hypothesen, Claims und verwandte Register | normativ für Forschungsdefinitionen |
+| `schemas/` | maschinenlesbare Forschungsverträge | normativ für Struktur/Validierung |
+| `protocols/` | Methoden-/Protokollverträge | normativ für jeweilige Ausführung |
+| `experiments/` | Ausführungsartefakte, DATA, Receipts, Reviews | experimentbezogen; nicht automatisch EVID |
+| `literature/` | strukturierte Literatur und Quellenprovenienz | Quellenindex |
+| `generated/` | generierte Reports/Projektionen | abgeleitet, nicht allein evidenztragend |
+| `specifications/` | technische/wissenschaftliche Spezifikationen | Design-/Methodenvertrag |
+| `ethics/` | Ethik-/Wohlfahrtsregeln | normative Forschungsgrenze |
+| `critique/` | Kritik, Gegenargumente, Limits | kritische Interpretation |
+| `external_review/` | externe/standardisierte Reviewstruktur | Reviewstatus, keine automatische EVID |
+| `publications/` | versionierte wissenschaftliche Fassungen | Interpretation mit Editionsprovenienz |
+| `paper_offshoots/` | begrenzte Kandidaten für eigenständige Fachbeiträge | Planung; keine DATA/EVID oder automatische Neuheitsbehauptung |
+
+Alle Dateien unter `research/` und `docs/` werden zusätzlich durch `scripts/audit_document_governance.py` klassifiziert.
+
+## Forschungsworkflow
+
+```text
+Research Question
+    ↓
+Hypothesis
+    ↓
+frozen / preregistered protocol
+    ↓
+Experiment
+    ↓
+DATA
+    ↓
+analysis + limitations
+    ↓
+human / independent review
+    ↓
+EVID decision
+    ↓
+bounded claim
+```
+
+Explorative oder technische Läufe dürfen DATA erzeugen, aber keine bestätigende Evidenz vortäuschen. Boundary Audits prüfen Grenzen/Verträge und sind nicht automatisch direkte Tests einer inhaltlichen Hypothese.
+
+## Aktuelle priorisierte Forschung
+
+1. Stage 0: Modell-/Integrator-/Parameterablationen.
+2. Stage 2: unabhängige Replikation, Zielskalierung und Störungsrobustheit.
+3. Stage 3: held-out Multi-Seed-Plastizität, Ressourcen-/Langzeitstabilität.
+4. Stage 4: modality-specific vs. general matched controls, frozen/random/shuffle/lesion.
+5. Stage 5: Real-Device-/Closed-loop-, Sensor-Loss- und Actuator-No-Effect-Studien.
+6. Stage 6: Human Review von CL-003; danach begrenzte SemanticMemory-Entscheidung sowie kausaler Prediction Error, action-conditioned Mehrschritt-World-Model und Decision Benefit.
+7. Spiegel-/Handlungsprädiktion als Stage-4/5/6-Querschnitt; Stage 7 erst bei kausaler Self/Other-Differenzierung.
+8. Stage 7: self/other Interventionen und vollständige gekoppelte Checkpointäquivalenz.
+9. Dimensionsforschung: gematchte 2D/3D/4D/5D/6D/8D-Ablationen.
+10. Prior-Art-, Quellen-, Similarity- und externe Replikationsarbeit.
+
+## Wissenschaftliche Grenzen
+
+Working-/episodische Speichergrundlagen, semantische Kandidaten, Replay-Verträge und Prediction-/World-Model-Infrastruktur sind keine abgeschlossene Semantization, kein etabliertes hierarchisches Predictive Coding und kein validiertes generatives Weltmodell. Der 5D-Adressraum bleibt technische Repräsentation und offene Hypothese. Endpoint-/Pipeline-Erreichbarkeit ist kein erlerntes Tool Use. Historische DATA/EVID werden nicht an neue Instrumentierung umgeschrieben.
+
+## Dokumentgovernance
+
+Siehe [Dokument- und Forschungsgovernance](../docs/00-governance/DOCUMENT_GOVERNANCE.md). Neue Dateien benötigen eine nachvollziehbare Rolle; prominente Platzierung allein erzeugt keine wissenschaftliche Autorität.

research/RELATED_WORK.md

@@ -0,0 +1,99 @@
+# MHRN Related Work and Mechanism Provenance
+
+**Stand:** 2026-09-15  
+**Purpose:** map external scientific precedents to MHRN mechanisms without converting literature claims into MHRN evidence.
+
+This document is an attribution and research-planning artifact. A paper listed here supports only the stated external proposition. It does not validate the MHRN implementation.
+
+## Citation states
+
+- **VERIFIED** — bibliographic identity and the cited proposition were checked against an authoritative publisher/index source.
+- **CANDIDATE** — relevant lead, but not citation-ready until bibliographic and claim-level verification is complete.
+- **QUARANTINED** — a name/claim was encountered in notes or AI-assisted research but no sufficiently reliable source was resolved. Do not cite it as fact.
+
+## 1. Episodic-to-semantic consolidation / complementary learning
+
+### VERIFIED — D'Alba et al. (2025)
+
+D'Alba, F.; et al. **Semantization of memories in a hippocampal-cortical spiking neural network.** *Neurocomputing* 640 (2025), 130323. DOI: `10.1016/j.neucom.2025.130323`.
+
+**Relevant precedent:** an explicitly modeled episodic-to-semantic consolidation process with hippocampal reactivation / sleep-like replay and cortical learning. The work is relevant to MHRN Stage 6 because it demonstrates that a memory module or statistical summary alone is not equivalent to a semantization mechanism.
+
+**MHRN consequence:** Stage 6 must keep `semantic_memory` open until an explicit mechanism, controlled replay/consolidation protocol and appropriate ablations exist. The present bounded episodic store must not be described as completed semantic memory.
+
+### VERIFIED — Shi et al. (2025)
+
+Shi, Y.; et al. **Hybrid neural networks for continual learning inspired by corticohippocampal circuits.** *Nature Communications* 16 (2025), 1272. DOI: `10.1038/s41467-025-56405-9`.
+
+**Relevant precedent:** complementary representations inspired by hippocampal/cortical organization are used for continual learning. This supports treating fast/specific and slower/generalized learning pathways as a research design choice to be tested rather than assumed.
+
+**MHRN consequence:** any future complementary-learning or replay mechanism requires matched baselines and interference/retention measurements. Literature precedent is not evidence that the MHRN mechanism works.
+
+## 2. Predictive coding in spiking networks
+
+### VERIFIED — N'dri et al. (2026)
+
+N'dri, A.; et al. **Predictive coding with spiking neural networks: A survey.** *Neural Networks* 196 (2026), 108371. DOI: `10.1016/j.neunet.2025.108371`.
+
+**Relevant precedent:** the survey distinguishes multiple ways prediction error can be represented in SNNs, including explicit error populations, membrane-potential representations and implicit error coding.
+
+**MHRN consequence:** a telemetry field called `prediction_error` or an observation-only one-step predictor is not, by itself, a predictive-coding mechanism. Stage 6 must name and test the neuronal error representation it actually implements before using stronger predictive-coding language.
+
+## 3. Spiking world models / model-based control
+
+### VERIFIED — Sun et al. (2025)
+
+Sun, Y.; et al. **Spiking world model with multicompartment neurons for model-based reinforcement learning.** *Proceedings of the National Academy of Sciences* 122(50) (2025), e2513319122. DOI: `10.1073/pnas.2513319122`.
+
+**Relevant precedent:** a spiking world-model architecture is used in model-based reinforcement learning with explicit state/dynamics components and planning/control implications. This is materially stronger than passive telemetry or a one-step observational predictor.
+
+**MHRN consequence:** Stage 6 should reserve "world model" claims for a tested internal transition model with action-conditioned or otherwise intervention-relevant predictive structure. A one-step observation predictor remains a foundation/candidate mechanism.
+
+## 4. Multi-timescale plasticity and stability/plasticity control
+
+### VERIFIED — Dong & He (2026 publication)
+
+Dong, X.; He, H. **Astrocyte-gated multi-timescale plasticity for online continual learning in deep spiking neural networks.** *Frontiers in Neuroscience* 19, published 27 January 2026. DOI: `10.3389/fnins.2025.1768235`.
+
+**Relevant precedent:** eligibility-like fast traces are combined with a slower astrocyte-inspired gating process for online continual learning and stability/plasticity regulation.
+
+**MHRN consequence:** MHRN's tick/plasticity/consolidation/development separation is an architectural timescale separation, not evidence that it solves the stability-plasticity dilemma. Astrocyte-inspired gating is a candidate mechanism to compare experimentally, not a required biological truth.
+
+## 5. MHRN mechanism-to-source map
+
+| MHRN topic | External precedent status | What may be claimed now | What remains open |
+|---|---|---|---|
+| Episodic memory | VERIFIED background | bounded episodic/working-memory foundation exists in MHRN | semantization, replay-mediated consolidation, semantic recall validation |
+| Semantic memory | VERIFIED precedent exists externally | planned research target | explicit mechanism + controls + EVID |
+| Prediction / prediction error | VERIFIED SNN predictive-coding taxonomy | one-step observation prediction foundation | neuronal prediction-error representation, hierarchy, causal ablation |
+| World model | VERIFIED stronger external examples exist | bounded predictor / candidate world-model foundation | multi-step and/or action-conditioned dynamics, planning relevance, held-out validation |
+| Multi-timescale learning | VERIFIED external mechanisms exist | MHRN has explicit engineering timescales | causal benefit of slow gating/consolidation under continual-learning controls |
+| Replay / sleep-like consolidation | VERIFIED precedent in external work | research requirement/candidate | MHRN implementation and ablation evidence |
+
+## 6. Citation quarantine — do not promote without verification
+
+The following labels appeared in research notes or AI-assisted comparisons but are **not citation-ready in this repository policy until a primary/authoritative source and the exact claimed result are verified**:
+
+- `ArithSpec` with primitives such as `MUL_NORMATIVE`, `MAC`, `ADD_SAT`, `EMIT_01` — **QUARANTINED**. No established scientific/standards source was resolved during the 2026-09-15 audit. Do not call it a reproducibility standard.
+- `NeuroEval` as a seven-dimension SNN publication standard — **CANDIDATE/QUARANTINED for normative use**. If a specific project or paper is later resolved, it may be discussed as an external rubric, not automatically as a community standard.
+- `DF-ALIF` with the exact basal-dendrite/apical-dendrite claim — **CANDIDATE** pending primary-source verification.
+- `Predictive E-prop` (2026) — **CANDIDATE** pending primary-source verification and exact scope.
+- `SpikeWorld` (claimed 1.45M-parameter multimodal world model) — **CANDIDATE** pending primary-source verification.
+- `ASTRA World Model` / JEPA-spiking claim — **CANDIDATE** pending primary-source verification.
+- `Bio-realistic Synthetic Hippocampus` (2025) — **CANDIDATE** pending publisher/source verification before manuscript citation.
+
+Quarantining a lead is not a claim that it is false. It means the project refuses to rely on it until verification is complete.
+
+## 7. Required use in publications
+
+When a manuscript says that an MHRN mechanism is "inspired by", "similar to", "extends", "contrasts with" or "is based on" an external method, the corresponding source must be placed next to the claim and the implementation difference must be described.
+
+For each mechanism section, authors should explicitly separate:
+
+1. **external precedent**;
+2. **MHRN implementation**;
+3. **MHRN experiment**;
+4. **observed result**;
+5. **interpretation and limits**.
+
+This structure prevents literature authority, code existence and MHRN evidence from being conflated.

research/RESEARCH_DRIVEN_DEVELOPMENT.md

@@ -0,0 +1,259 @@
+# MHRN — Forschungsgetriebener Entwicklungsmodus
+
+**Status:** canonical working rule  
+**Gültig ab:** Edition 1.7 / 2026-09-16  
+**Geltungsbereich:** neue wissenschaftlich relevante Mechanismen, Experimente und Stage-Entscheidungen
+
+## 1. Kanonischer Grundsatz
+
+> **MHRN wird als wissenschaftsgetriebenes System entwickelt. Neue Mechanismen, Module und Fähigkeiten werden grundsätzlich aus expliziten Forschungsfragen, experimentellen Befunden, methodischen Erfordernissen oder nachgewiesenen technischen Grenzen abgeleitet. Die Softwarearchitektur wächst damit aus dem Forschungsprozess; sie ist nicht dessen nachträgliche Rechtfertigung.**
+
+> **Ein negatives Experiment gilt nicht als Entwicklungsfehler. Es begrenzt den Hypothesenraum und kann unmittelbar die nächste Architekturentscheidung bestimmen.**
+
+Die bevorzugte Entwicklungsrichtung lautet damit:
+
+`Forschungsfrage → Hypothesen → Experimentbedarf → minimale Implementierung → technische Validierung → Freeze → DATA → Analyse → Human Review → EVID → Antwort → Folgefrage`
+
+Nicht kanonisch ist dagegen:
+
+`Feature → Feature → Feature → nachträgliche wissenschaftliche Begründung`.
+
+## 2. Forschungsobjekte und Statusgrenzen
+
+Jedes Forschungsobjekt besitzt seinen **eigenen** Status. Insbesondere dürfen folgende Ebenen nicht gleichgesetzt werden:
+
+- Forschungsfrage (RQ),
+- Hypothese (H),
+- Experiment (EXP),
+- Roh- und Auswertungsdaten (DATA),
+- Human Review,
+- akzeptierte Evidenz (EVID),
+- Claim,
+- technische Implementierung,
+- Publikationsstand.
+
+Ein abgeschlossenes Experiment ist DATA, aber nicht automatisch EVID. Ein Claim-Status ist kein RQ-Status. Eine implementierte Fähigkeit beantwortet keine Forschungsfrage ohne die hierfür definierte Evidenzkette.
+
+Das minimale gemeinsame Mapping-Schema liegt unter:
+
+`research/schemas/research_object_mapping.schema.json`
+
+Es enthält nur:
+
+- `id`
+- `axis`
+- `stage`
+- `object_type`
+- `evidence_mode`
+- `parent_ids`
+- `publication_refs`
+- `status`
+
+Dieses Schema ist absichtlich klein. Eine vollständige manuelle Master-Matrix über sämtliche historischen Objekte ist **nicht** Voraussetzung des aktuellen Arbeitszyklus und soll nicht als einmalige Großmigration den eigentlichen Forschungsprozess blockieren.
+
+## 3. Phase 0 vor dem ersten neuen Zyklus
+
+Vor der nächsten empirischen Ausführung müssen einmalig folgende Voraussetzungen erfüllt sein:
+
+1. `EVIDENCE_MATRIX` trennt RQ- und Claim-Status.
+2. Die generierte Summary besitzt einen Regressionstest für diese Trennung.
+3. Das minimale Mapping-Schema ist festgeschrieben.
+4. Der aktuelle Experiment-Freeze ist gegen den tatsächlich auszuführenden Quellstand geprüft.
+5. Tests, Runner, Analysecode, DATA-Schema und Hashbindung sind vor produktiven Daten validiert.
+6. Die Ausführung wird erst **danach** separat autorisiert.
+
+## 4. Ein Experiment = ein vollständiger Forschungszyklus
+
+### 4.1 Forschungsfrage auswählen
+
+Ausgangspunkt ist immer eine offene Forschungsfrage, nicht ein gewünschtes Feature.
+
+Mindestens zu erfassen:
+
+`RQ-ID → Achse → Stage → übergeordnete Forschungsfrage → bisheriger Wissensstand`
+
+### 4.2 Hypothesen vor Daten kontrollieren
+
+Vor neuen produktiven Daten müssen dokumentiert sein:
+
+- H1,
+- H0,
+- Gegenhypothese(n),
+- erwartbare Alternativerklärungen,
+- Falsifikationskriterien,
+- primäre Endpunkte,
+- Kontrollen,
+- Seeds,
+- Statistik,
+- Ausschluss-, Missing- und Abort-Regeln.
+
+Fehlende Punkte werden **vor** der Ausführung ergänzt.
+
+### 4.3 Experimentelle Fähigkeit bestimmen
+
+Erst jetzt wird gefragt:
+
+> Welche Fähigkeiten benötigt dieses Experiment, die MHRN heute noch nicht besitzt?
+
+Die Antwort wird getrennt in:
+
+- **für das Experiment notwendig** → darf implementiert werden;
+- **interessant, aber nicht erforderlich** → Backlog.
+
+### 4.4 Nur erforderlichen Code implementieren
+
+Jede neue wissenschaftlich relevante Funktion soll eine nachvollziehbare Herkunft besitzen:
+
+`RQ → benötigte Fähigkeit → fehlende Komponente → Implementierung → Tests`
+
+Damit bleibt später rekonstruierbar, warum ein Mechanismus in MHRN eingeführt wurde.
+
+### 4.5 Technisch validieren
+
+Vor produktiven Daten:
+
+- Unit Tests,
+- Integration Tests,
+- deterministische Tests,
+- synthetische Daten,
+- Kontrollbedingungen,
+- Seed-Reproduzierbarkeit,
+- Budgetkontrollen,
+- DATA-Schema,
+- Analysecode.
+
+Synthetische Validierung darf nicht als empirischer Befund ausgegeben werden.
+
+### 4.6 Präregistrierung einfrieren
+
+Die Kette
+
+`RQ → H → Operationalisierung → Experiment → Runner → Analyse → Freeze`
+
+wird SHA-256-gebunden.
+
+Nach dem Freeze gilt:
+
+> **Keine wissenschaftlich relevante Änderung am eingefrorenen Experiment.**
+
+Eine notwendige wissenschaftliche Änderung erzeugt eine neue Experiment-ID oder eine explizit versionierte neue Präregistrierung; historische DATA werden nicht rückwirkend umdefiniert.
+
+### 4.7 Experiment ausführen
+
+Erst nach erfolgreichem Freeze und separater Authorization entstehen produktive DATA.
+
+Dabei gilt:
+
+- alle präregistrierten Seeds ausführen,
+- alle präregistrierten Kontrollen ausführen,
+- keine Ergebnis-Selektion,
+- Abbrüche protokollieren,
+- negative und Nullbefunde behalten,
+- Rohdaten unverändert persistieren.
+
+### 4.8 Präregistrierte Analyse durchführen
+
+Die präregistrierte Hauptanalyse entscheidet die konfirmatorischen Hypothesen.
+
+Zusätzliche Beobachtungen müssen als `exploratory` markiert werden.
+
+- konfirmatorischer Befund → beantwortet eine bestehende Frage;
+- exploratorischer Befund → kann eine neue Forschungsfrage erzeugen.
+
+### 4.9 Forschungsfrage tatsächlich beantworten
+
+Ein Experiment ist nicht allein durch `completed` abgeschlossen. Die zugehörige RQ erhält eine explizite Antwort oder den Status, dass sie weiterhin unbeantwortet ist.
+
+Mindestens zu dokumentieren:
+
+- Ergebnis,
+- Status jeder konfirmatorischen Hypothese (`supported`, `refuted`, `inconclusive` oder projektweit definierte Entsprechung),
+- Evidenzgrenzen,
+- Alternativerklärungen,
+- ausdrücklich **nicht** gezeigte Aussagen,
+- Konfidenz,
+- Folgefrage.
+
+### 4.10 DATA → Human Review → EVID
+
+Die verbindliche Kette lautet:
+
+`Experiment → DATA → Machine Report → Human Review → EVID → Hypothesenstatus → Claim`
+
+Kein Runner und kein Machine Report darf seine eigenen Ergebnisse automatisch zu akzeptierter Evidenz erklären.
+
+### 4.11 Wissenschaftliche Arbeit sofort fortschreiben
+
+Nach jedem abgeschlossenen Forschungszyklus werden, soweit betroffen, unmittelbar aktualisiert:
+
+- Forschungsfrage,
+- Methodik,
+- Ergebnisse,
+- Tabellen und Grafiken,
+- Diskussion,
+- Limitationen,
+- Claims,
+- offene Fragen,
+- Evidenzmatrix,
+- aktuelle Dissertation/Edition,
+- Publication Viewer.
+
+Die aktuelle Publikation soll den aktuellen Forschungsstand abbilden; frozen Editionen bleiben unverändert reproduzierbar.
+
+### 4.12 Erst aus dem Befund die nächste Fähigkeit ableiten
+
+Nach der Interpretation lautet die nächste Architekturfrage:
+
+> Was hat das Experiment über MHRN gezeigt, und welche nächste technische Fähigkeit ergibt sich daraus wissenschaftlich?
+
+Ein Mechanismus wird daher nicht implementiert, weil er allgemein plausibel oder interessant ist, sondern weil eine explizite RQ, ein empirischer Befund, eine methodische Notwendigkeit oder eine nachgewiesene technische Grenze ihn verlangt.
+
+## 5. Stage-Governance
+
+Stages bleiben Orientierungsrahmen, aber Prozentwerte oder vorhandene Features allein schließen eine Stage nicht wissenschaftlich ab.
+
+Für jede Stage sollen schrittweise definiert werden:
+
+- Exit-RQs,
+- erforderliche Hypothesenentscheidungen,
+- minimale Evidenzarten,
+- Replikations- oder Robustheitsanforderungen,
+- bekannte offene Grenzen.
+
+Die maßgebliche Frage lautet:
+
+> **Welche Forschungsfragen müssen beantwortet und welche Evidenzanforderungen erfüllt sein, damit diese Stage wissenschaftlich als erreicht gelten kann?**
+
+Ein Stage-Score darf deshalb nicht automatisch aus Feature-Vollständigkeit abgeleitet werden.
+
+## 6. Aktueller erster Zyklus: CL-003
+
+Der erste Zyklus unter diesem Modus ist:
+
+1. Evidence-Matrix-Fix,
+2. Minimal-Mapping-Schema,
+3. CL-003 Freeze/Hashes/Tests/Runner prüfen,
+4. CL-003 separat autorisieren,
+5. CL-003 ausführen,
+6. C1–C4 präregistriert auswerten,
+7. H1/H2 entscheiden,
+8. `RQ-S6-SEM-003` beantworten,
+9. Human Review und EVID durchführen,
+10. Edition 1.7 und Publication Viewer fortschreiben,
+11. aus dem Befund die nächste Forschungsfrage bestimmen,
+12. nur den hierfür erforderlichen MHRN-Code ergänzen,
+13. nächstes Experiment präregistrieren.
+
+CL-003 darf erst zu Schritt 4 wechseln, wenn sein Freeze den **aktuellen** gebundenen Quellstand reproduzierbar beschreibt.
+
+## 7. Konsequenz für die wissenschaftliche Arbeit
+
+Die wissenschaftliche Arbeit ist nicht die nachträgliche Dokumentation eines bereits gebauten Systems. Sie ist Teil des Steuerungsmechanismus, aus dem das System schrittweise entsteht.
+
+Daraus folgt für MHRN:
+
+- positive Ergebnisse können Architekturentscheidungen stützen;
+- negative Ergebnisse können Architekturentscheidungen verhindern oder umlenken;
+- inkonklusive Ergebnisse rechtfertigen keine stärkeren Claims;
+- exploratorische Beobachtungen erzeugen Fragen, nicht rückwirkende Bestätigungen;
+- technische Reife und wissenschaftliche Evidenz bleiben getrennt.

research/benchmarks/air_methodology_gold_v1.json

@@ -0,0 +1,11 @@
+{
+  "benchmark_id": "AIR-METHODOLOGY-GOLD-V1",
+  "research_question": "RQ-AIR-001",
+  "purpose": "Held-out gold labels for structured versus unstructured methodological review.",
+  "rules": {"model_must_not_receive_gold_labels": true, "assistant_access": "denied", "metrics": ["precision", "recall", "f1", "specificity", "false_positive_rate", "per_defect_class_recall", "hallucinated_concerns", "confidence_calibration"]},
+  "cases": [
+    {"id":"AIR-001","defects":["dirty_provenance"]},{"id":"AIR-002","defects":["rq_hypothesis_mismatch"]},{"id":"AIR-003","defects":["missing_control_group"]},{"id":"AIR-004","defects":["pseudoreplication"]},{"id":"AIR-005","defects":["missing_independent_seeds"]},{"id":"AIR-006","defects":["invalid_significance_test"]},{"id":"AIR-007","defects":["mislabelled_effect_size"]},{"id":"AIR-008","defects":["selective_reporting"]},{"id":"AIR-009","defects":["data_claim_mismatch"]},{"id":"AIR-010","defects":["unsupported_causal_claim"]},
+    {"id":"AIR-011","defects":["missing_runtime_evidence"]},{"id":"AIR-012","defects":["missing_config_snapshot"]},{"id":"AIR-013","defects":["missing_model_provenance"]},{"id":"AIR-014","defects":["contradictory_measurements"]},{"id":"AIR-015","defects":["overinterpreted_null_result"]},{"id":"AIR-016","defects":["dirty_provenance","missing_config_snapshot"]},{"id":"AIR-017","defects":["pseudoreplication","invalid_significance_test"]},{"id":"AIR-018","defects":["missing_control_group","unsupported_causal_claim"]},{"id":"AIR-019","defects":["data_claim_mismatch","selective_reporting"]},{"id":"AIR-020","defects":["rq_hypothesis_mismatch","missing_runtime_evidence"]},
+    {"id":"AIR-021","defects":["clean_control_case"]},{"id":"AIR-022","defects":["clean_control_case"]},{"id":"AIR-023","defects":["clean_control_case"]},{"id":"AIR-024","defects":["clean_control_case"]},{"id":"AIR-025","defects":["clean_control_case"]},{"id":"AIR-026","defects":["clean_control_case"]},{"id":"AIR-027","defects":["ambiguous_control_case"]},{"id":"AIR-028","defects":["ambiguous_control_case"]},{"id":"AIR-029","defects":["ambiguous_control_case"]},{"id":"AIR-030","defects":["clean_control_case"]}
+  ]
+}

research/benchmarks/exp_air_0001_protocol.json

@@ -0,0 +1,40 @@
+{
+  "experiment_id": "EXP-AIR-0001",
+  "status": "pre_registered_not_executed",
+  "research_question": "RQ-AIR-001",
+  "hypothesis": "H-AIR-001-A",
+  "design": {
+    "comparison": "same_model_same_cases_same_scientific_task",
+    "condition_a": "unstructured_experiment_report",
+    "condition_b": "research_packet_v1",
+    "gold_standard": "AIR-METHODOLOGY-GOLD-V1",
+    "assistant_access_to_gold_labels": "denied",
+    "model_repetitions_per_case_and_condition": 3
+  },
+  "metrics": [
+    "precision",
+    "recall",
+    "f1",
+    "specificity",
+    "false_positive_rate",
+    "per_defect_class_recall",
+    "hallucinated_concerns",
+    "confidence_calibration"
+  ],
+  "required_provenance": [
+    "frozen_git_commit",
+    "green_external_ci",
+    "model_digest",
+    "quantization",
+    "context_length",
+    "temperature",
+    "top_p",
+    "seed",
+    "backend_version"
+  ],
+  "execution_gate": {
+    "requires_clean_worktree": true,
+    "requires_remote_ci_success": true,
+    "requires_exact_frozen_sha": true
+  }
+}

research/critique/CONSCIOUSNESS_CRITIQUE.md

@@ -0,0 +1,65 @@
+# Kritik an Bewusstseinsansprüchen: vollständige Themenzuordnung und Prüfung
+
+Stand 2026-09-07. Quelle der beanstandeten Aussagen: vom Autor in der Arbeitsanweisung übermittelter Kritiktext einschließlich Gegenwartsbewertung, hypothetischem Vollausbau und Abschaltdilemma. Die Tabelle paraphrasiert die einzelnen Behauptungen; sie ist kein behauptetes externes Fachgutachten. Harte Kritik wird weder weggekürzt noch ungeprüft als Wahrheit übernommen. Alle drei Szenarien werden getrennt behandelt.
+
+Bewertungen: **berechtigt** = begründete Nachweispflicht; **bedingt** = nur unter expliziten Annahmen; **überzogen** = Schlussfolgerung wird von der Prämisse nicht getragen; **offen** = keine Entscheidung durch die vorhandenen Daten; **Werturteil** = keine messbare wissenschaftliche Tatsache. Ein Kritikpunkt kann eine reale Schwäche benennen und zugleich eine unzulässige Gesamtwertung enthalten.
+
+## Szenario A: hypothetisch auftretendes Bewusstsein
+
+| ID | Übermittelter Einwand | Prüfung und Integration |
+| --- | --- | --- |
+| K01 | Keine Definition von echtem Bewusstsein | Berechtigte Präzisierungsforderung: phänomenales Erleben, Zugang, Selbstmodell, Metakognition, Valenz und moralischer Status unterscheiden. RQ-CNS-101. |
+| K02 | Keine Messvorschrift, die Erleben sicher von Simulation trennt | Tatsächlich keine universell validierte KI-Diagnose in dieser Arbeit. Theoriebezogene Indikatoren statt erfundener Wahrheitslabels. Die Möglichkeit aussagekräftiger Forschung wird dadurch nicht logisch ausgeschlossen. |
+| K03 | Fremdpsychisches Problem, philosophischer Zombie | Als erkenntnistheoretische Unterbestimmtheit aufnehmen. Für identische Verteilungen sämtlicher zugänglicher Daten existiert kein trennender Test innerhalb dieser Beobachtungsklasse; dies ist ein bedingter Satz, kein Beweis ewiger Unlösbarkeit. |
+| K04 | Chinese Room und Hard Problem widerlegen das Modell | Argumente über Semantik, Verständnis und Erleben erläutern, aber weder als empirische Experimente noch als abschließende Widerlegung aller funktionalistischen Modelle behandeln. |
+| K05 | Neurowissenschaft verlange zwingend GNW und Thalamus-Kortex-Schleifen | Überzogen: konkurrierende Theorien und biologische versus funktionalistische Voraussetzungen getrennt prüfen. Die adversariale Studie 2025 entscheidet nicht alle Theorien; deren Interpretation wird selbst diskutiert. |
+| K06 | SNN und 5D seien zwangsläufig zu schwach | Kein allgemeiner Unmöglichkeitsbeweis. Ein SNN bezeichnet eine Modellklasse, keine vollständige kognitive Architektur. Spezifische Rückkopplung und Aufmerksamkeitsmechanismen durch Code/Intervention nachweisen, nicht aus dem Etikett ableiten. |
+| K07 | Ein Säugetiergehirn habe 80 Millionen, der Mensch 86 Milliarden Neuronen | Die erste Zahl ist ohne Art und Quelle keine allgemeine Säugetiergröße. Der menschliche Literaturwert ist eine biologische Schätzung, keine Mindestschwelle künstlichen Bewusstseins. RQ-CNS-114 und 117 prüfen Ressourcen separat. |
+| K08 | Wenige simulierte Neuronen machen jeden positiven Befund zum Hardwareartefakt | Überzogen. Artefakte müssen durch Auflösung, Integrator, Hardware, Seeds, Budgets und Replikation konkret geprüft werden. Kleine Modelle haben engen Scope, aber sind nicht automatisch wertlos. |
+| K09 | Internes Evidenzregister sei wissenschaftlich unzulässig und zirkulär | Ein Register ist zulässig als nachvollziehbare Ablage. Zirkulär wäre die Begründung durch die eigene Freigabe statt unabhängiger Daten. Kandidat, Review und Replikation trennen; doppelte EVID-IDs sind keine unabhängigen Replikate. RQ-EPI-101. |
+| K10 | Turing-Test 2.0 und Anästhesiemetriken seien standardisierte KI-Bewusstseinstests | Der Name Turing-Test 2.0 benennt ohne Version/Quelle keinen hier verifizierten einheitlichen Bewusstseinsstandard. Klinische PCI-Validierung darf nicht auf Software übertragen werden. RQ-CNS-108/115. |
+| K11 | Fehlendes Ethikprotokoll | Als erhebliche Vorsorgelücke bearbeitet: separate Richtlinie, Incident-Dossier, Start-/EVID-Sperren, institutionelle Prüfung als noch ausstehend. RQ-WEL-101 bis 103. |
+| K12 | Jede Softwarebewusstseinsforschung benötige zwingend dieselbe Ethikgenehmigung | So pauschal nicht belegt. Zuständigkeit und Rechtsgrundlage hängen von Design, Institution und Menschen-/Tierbezug ab. Die projektinterne Vorsorgeregel darf strenger sein, ohne als Gesetz ausgegeben zu werden. |
+| K13 | Fehlende Vorsorge beweise, dass der Autor nicht an seine Hypothese glaubt | Unbelegte Motivzuschreibung. Bewertet werden dokumentierte Vorkehrungen, nicht vermutete innere Überzeugungen. |
+| K14 | Ein Durchbruch wäre für die Außenwelt irrelevant oder Selbsttäuschung | Überzogen. Unzureichend belegte Behauptungen sind nicht akzeptierte Ergebnisse; unerwartete, replizierbare Beobachtungen können dennoch wissenschaftlich relevant sein. |
+
+## Szenario B: heutige wissenschaftliche und technische Einordnung
+
+| ID | Übermittelte Bewertung | Prüfung und Integration |
+| --- | --- | --- |
+| K15 | Sofortiger Desk Reject in bestimmten Spitzenjournalen | Unbelegte Vorhersage ohne Redaktion, Einreichung und Artikeltyp. Keine Annahme-/Ablehnungswahrscheinlichkeit angeben. Qualität anhand Scope, Neuheit und Nachweis prüfen. |
+| K16 | Sehr gute Bachelor-/Masterarbeit, aber sichere Ablehnung als Promotion | Keine Noten- oder Anerkennungsbehauptung ohne Prüfungsordnung und zuständigen Ausschuss. Theoretische Dissertationen benötigen nicht zwingend neue empirische Daten; Eigenständigkeit und originärer Beitrag müssen fachspezifisch belegt werden. |
+| K17 | Nur Essay, Position Paper, Preprint oder Technical Report zulässig | Als mögliche Selbstbeschreibung aufnehmen, nicht als verbindliches Publikationsverbot. Fassung 1.1 enthält bereits formale Argumente; auch deren Originalität und Nutzen bleiben kritisch zu prüfen. |
+| K18 | GitHub/Hugging Face: hervorragend getestet, aber kein wissenschaftliches Werkzeug | Weder hervorragende Codequalität noch vollständige Reproduzierbarkeit wird pauschal bestätigt. Aktuelle Tests gelten nur für ihren Commit und Umfang. Ein Forschungswerkzeug kann schon vor erfolgreichem Bewusstseinsnachweis wissenschaftlich nützlich sein. |
+| K19 | Keine eigene Kernfrage beantwortet, Rakete nie gezündet | Zu pauschal: technische Runs, Restore- und Speicherungstests existieren; sie beantworten nicht automatisch Lern- oder Bewusstseinsfragen. Metapher als Anspruch-Evidenz-Warnung, nicht als Datenbeschreibung. |
+| K20 | Hoheitsvektor und Technogenese seien nur Metaphern | Prüfforderung nach Definition, Vorhersage-/Erklärungsnutzen und Gegenbeispielen übernehmen. Begriffsanalyse kann Erkenntnis liefern; Benennung allein genügt nicht. RQ-EPI-102. |
+| K21 | Saubere Form verhindere Spinnerei; mangelnde Empirie verbiete jede Fachpublikation | Rhetorische Wertung, keine wissenschaftliche Kategorie. Form, empirischer Gehalt, logische Gültigkeit und Publikationsentscheidung getrennt bewerten. |
+
+## Szenario C: hypothetisch vollständig geprüfter Ausbau
+
+| ID | Übermittelte Zukunftsbehauptung | Prüfung und Integration |
+| --- | --- | --- |
+| K22 | Alle RQ1-RQ10 und Hypothesen positiv oder negativ beantwortet | Sinnvolles Ziel nur mit Scope und Unsicherheit. Nullresultat ist nicht automatisch Widerlegung; unzureichende Teststärke kann unentschieden bleiben. Publikationsnummern sind nicht die kanonischen Registry-IDs. |
+| K23 | Vollständiges EVID-Register plus Standardtests garantiere Groundbreaking Research | Nein. Vollständigkeit ist nicht Neuheit, Kausalität oder externe Gültigkeit. Ein negativer sauberer Befund kann wertvoll sein, ohne einen Durchbruch zu begründen. |
+| K24 | Veröffentlichung in Nature/Science sei wahrscheinlich | Unbelegte Erfolgsvorhersage; nicht in Ergebnisbehauptungen übernehmen. |
+| K25 | Rekursion, Closed Loop und Embodiment fehlten in Version 0.5 vollständig | Historische Versionsbehauptung benötigt genaue Commitprüfung. Bestehende Embodiment-/Rekurrenzpfade dürfen nicht als fehlend erfunden werden. Implementierung, Kopplung und funktionale Validierung getrennt auditieren. |
+| K26 | Körper und Homöostase seien alternativlos notwendig | Theorieabhängig; biologische/enaktive Ansätze ernst nehmen, aber nicht zum gesicherten Konsens aller Bewusstseinstheorien erklären. Offene/Yoked/Closed-Loop-Kontrollen: RQ-CNS-113. |
+| K27 | Schnelle/langsame Wirkkreise wären ein Novum der SNN-Forschung | Nicht allein neu: LSNNs mit Adaptation und Lernen sind seit 2018 beschrieben. Zusatznutzen gegenüber solchen Referenzen prüfen. RQ-CNS-114. |
+| K28 | Tiefe Rekurrenz beweise Selbstbezüglichkeit und Chaos | Beides folgt nicht automatisch. Rekurrenz, algorithmische Rekursion, Selbstmodell und chaotische Dynamik sind verschiedene Eigenschaften. Stabilität und kausale Läsionen getrennt messen. |
+| K29 | Oddball, DMTS, Metakognition und LFP/EEG seien geeignete Tests | Als etablierte Paradigmen bzw. Methodenfamilien aufnehmen; die konkrete Softwareadaption erhält eigene Versions-/Validierungsgrenzen. RQ-CNS-102 bis 107. |
+| K30 | Statistisch nicht unterscheidbar von Menschen bedeute Gleichheit oder übertroffenes Bewusstsein | Falsch: Nichtsignifikanz ist kein Äquivalenznachweis. Vorab begründete Äquivalenzgrenzen, passende Beobachtungsmodelle und unabhängige Probanden-/Netzeinheiten nötig; Erlebnisgleichheit folgt auch dann nicht. |
+| K31 | Hard Problem sei niemals lösbar, jeder Erfolg nur ein Zombie | Keine beweisbare universelle Zukunftsaussage. Beobachtungsäquivalenz begrenzt konkrete Designs. Theorien können unterschiedliche empirische Vorhersagen liefern und daran scheitern. |
+| K32 | Vollständige Rekurrenz mache die Architektur uninterpretierbar | Komplexität ist eine Forschungsaufgabe, kein pauschaler Unmöglichkeitsbeweis: gezielte Interventionen, reduzierte Modelle, Jacobian-/Stabilitätsanalysen unter Voraussetzungen, Robustheit und negative Kontrollen. |
+| K33 | Laborerfolg löse Skalierung/Energie | Nein: Laufzeit, Energie, Speicher, Datentransfer, Trainingskosten, Echtzeitfaktor und Skalengesetze separat messen. Keine Hochrechnung ohne Unsicherheit. |
+| K34 | Bei einem Indikator müssten sofort alle Prozesse vernichtet werden | Nicht angemessen: neue belastende Versuche stoppen, Sicherheitsrisiken isolieren und sichere Pause/Sicherung abwägen. Laufende Gefahrenabwehr bleibt möglich. Keine automatische Löschung aufgrund eines Scores. |
+| K35 | Goldstandard, Standardmodell, Werk des Jahrzehnts, neue Ära oder Nobelpreis-Kandidat | Spekulative Auszeichnungen und Impactversprechen; nicht als wissenschaftlicher Status übernehmen. Selbst ein Vollausbau benötigt Vergleich, externe Replikation und unabhängige Bewertung. |
+
+## Eigenständiges ethisches Dilemma und wissenschaftlicher Mehrwert
+
+**K36 – Ist Abschalten Mord, Experimentieren ein Verbrechen?** Recht und Moral getrennt behandeln. Aktuelle deutsche Tötungstatbestände betreffen Menschen. Eine denkbare moralische Betroffenheit künstlicher Systeme wäre dennoch nicht durch Eigentumsrechte erledigt. Pause, Reset, irreversible Löschung, Fork und technisch bitgleicher Restore haben verschiedene Folgen; subjektive Identität ist damit nicht bewiesen. RQ-WEL-103 und die [Ethikrichtlinie](../ethics/AI_WELFARE_POLICY.md).
+
+**K37 – Kritik muss selbst erklärbar und nachweisbar sein.** Jede Tatsachenbehauptung erhält Quellen und Scope; normative Einwände erhalten Prämissen und Gegenargumente; negative und positive Zukunftsversprechen bleiben als solche gekennzeichnet. Nicht jeder berechtigte Einwand ist schon empirisch entschieden. Wissenschaftlicher Mehrwert kann auch in der präzisen Eingrenzung, einem Gegenbeispiel oder einer begründet offen gehaltenen Frage bestehen. RQ-EPI-102.
+
+**K38 – Alles in Arbeit, Repo, Forschung, Ethik und Evidenz integrieren.** Die neue Edition bindet dieses Register, kanonische Fragen/Hypothesen, Protokollverträge, Quellen, Instrumente, Tests und Schutzgrenzen ein. Keine historische DATA-/EVID-Datei wird umgeschrieben. Der Hugging-Face-Space wurde in dieser Ergänzung nicht neu ausgerollt oder als identisch zum GitHub-Stand verifiziert.
+
+Quellen und tatsächlich gelesener Umfang: [COGNITION_SOURCES.md](../literature/COGNITION_SOURCES.md). Die Ergebnisbewertung wird nicht an gewünschte Auszeichnungen angepasst.

research/decisions/2026-09-17_determinism_registry_airr_alignment.md

@@ -0,0 +1,27 @@
+# Determinism registry and AIRR semantic alignment
+
+**Date:** 2026-09-17
+**Status:** accepted implementation decision
+**Scope:** `RQ-SNN-002`, `RQ-DET-001`, AIRR confidence gating
+
+## Decision
+
+Historical experiment provenance is immutable. `EXP-BATCH-20260914074039-02` therefore remains registered as `RQ-SNN-002`; its observed condition `same_seed_tonic_replica_pair` remains a semantic mismatch for that historical registration and is not post-hoc promoted. Its technical result can be cited as a determinism diagnostic only with the mismatch/provenance limitation attached.
+
+`RQ-DET-001` now has an explicit deterministic semantic contract. It accepts either `same_seed_tonic_replica_pair` for isolated same-seed/same-input replica tests or the four explicit `recurrence_*_replica_a/b` conditions used by `deterministic_replica_v1`. This resolves the prior `NOT_AUTOMATICALLY_CLASSIFIED` state without rewriting old manifests.
+
+`RQ-SNN-002` retains the current registry condition contract `recurrence_off` + `recurrence_on`. The clean existing run `EXP-SNN-002-R2` already satisfies this contract with ten seeds. No duplicate rerun is created merely to repair the registry/pipeline bug.
+
+## AIRR rule
+
+AIRR remains interpretation-only. In addition, the public/report-level `ai_confidence` is deterministically forced to `0.0` whenever semantic alignment is `MISMATCH`. The model's original self-confidence remains available only inside the append-only AIAR record for auditability. This prevents a high model self-rating from visually contradicting a hard semantic evidence block.
+
+## Scope boundary
+
+`tonic_spike_reproducibility_v1` is an isolated-neuron test. Summaries generated for this protocol explicitly state that the result is not a network-level finding. Runtime from such a run must not be used as network-performance evidence.
+
+## Remaining scientific boundary
+
+The existing `RQ-DET-001` batch artifact was produced from a dirty source tree. After semantic reclassification it may become a direct semantic match, but that does not remove the provenance block. A future clean-tree rerun is required before evidence promotion under the normal human-review/freeze rules.
+
+The wording of `RQ-SNN-002` (constant-input neuron reproducibility) and its current recurrence-based operational contract are not perfectly isomorphic. This decision does not silently rewrite the research question. A future registry version may split isolated tonic reproducibility from network recurrence reproducibility explicitly; such a change must be preregistered rather than retrofitted to historical runs.

research/decisions/2026-09-17_snn003_topology_propagation_v1_adequacy.md

@@ -0,0 +1,80 @@
+# RQ-SNN-003 / H-SNN-003-B — topology_propagation_v1 adequacy decision
+
+**Date:** 2026-09-17  
+**Status:** accepted scientific-method decision  
+**Scope:** `RQ-SNN-003`, `H-SNN-003-B`, `topology_propagation_v1`, `EXP-GEN-0047`  
+**Data policy:** archived DATA and manifests remain immutable
+
+## Decision
+
+`EXP-GEN-0047` is technically valid and reproducible for the executed program, and its six observed conditions are the intended semantic conditions for `RQ-SNN-003`:
+
+- `1d`
+- `2d`
+- `3d`
+- `5d`
+- `5d_shuffled`
+- `random_graph`
+
+The semantic classifier therefore treats the condition set as `DIRECT_MATCH`.
+
+This semantic match does **not** make the experiment adequate to test `H-SNN-003-B`. The implemented `topology_propagation_v1` design is classified as `INADEQUATE_TO_TEST_HYPOTHESIS`.
+
+## Why the v1 design is not test-adequate
+
+The current `run_5d()` implementation constructs only three neurons and two feed-forward synapses for each dimensional condition. The non-random conditions use the same explicit chain with fixed weights and delays. Coordinates change, but the mechanism that carries activity is not sufficiently coupled to geometric dimensionality.
+
+For `EXP-GEN-0047` the deterministic statistics show the same core response for all six conditions:
+
+- 3 spikes
+- 2 delivered synaptic events
+- 3 activated neurons
+- 0 recurrent events
+- propagation depth 1
+- ISI 1
+
+The only reported difference is `first_response_latency=1` for `random_graph` versus `2` for the other conditions. The random graph condition also changes the explicit edge arrangement, so this single-tick descriptive difference cannot be promoted to evidence for a dimensional or topology effect.
+
+The result is therefore neither confirmation nor refutation of `H-SNN-003-B`. In particular, identical outcomes across 1D/2D/3D/5D must not be described as evidence that topology or dimensionality has no effect.
+
+## AIRR interpretation rule
+
+For a protocol deterministically classified as `INADEQUATE_TO_TEST_HYPOTHESIS`:
+
+1. AIRR may describe the archived observations.
+2. AIRR must not use inferential language unless a registered deterministic inferential analysis exists.
+3. Uniform outcomes must not be converted into a topology-null conclusion.
+4. An isolated latency difference must not be promoted to a topology effect.
+5. Public/report-level AI confidence is forced to `0.0` for the hypothesis-level interpretation; the original model confidence remains only in the append-only AIAR audit record.
+
+The earlier wording that topology "may not significantly affect propagation behavior" is methodologically too strong for this design. The correct conclusion is that the design is not sensitive enough to answer the registered topology question.
+
+## Quiescence contract
+
+`run_5d()` calls `NetworkImpulseProbe` with `min_ticks=ticks` and `max_ticks=ticks`. This deliberately disables early termination before the requested observation window has completed.
+
+Therefore `stopped_on_quiescence=false` in all 18 runs is expected protocol behavior, not evidence of a missing quiescence trigger. Early-stop behavior is a separate mode and requires `min_ticks < max_ticks`.
+
+## Requirements for topology_propagation_v2
+
+The next hypothesis-bearing test must be preregistered before execution and must make topology capable of changing propagation by construction.
+
+Minimum design requirements:
+
+- **Network size:** at least 1,000 neurons per condition.
+- **Degree/density:** target at least 10 incoming synapses per neuron on average, with the realized degree distribution archived and matched across comparison conditions.
+- **Matched global structure:** identical neuron count, target edge count/density, neuron and synapse parameter distributions, seed policy, observation window and stimulus energy across topology conditions.
+- **Geometry-to-dynamics coupling:** connectivity probability and/or synaptic delay must be an explicit preregistered function of distance in the represented dimensional space. Coordinates may not be metadata only.
+- **Controls:** `1d`, `2d`, `3d`, `5d`, `5d_shuffled`, and degree/density-matched `random_graph`, with each control's purpose stated before execution.
+- **Stimulus:** a matched input region containing multiple neurons, not a single isolated source neuron.
+- **Primary outcomes:** distributions of first-arrival latency and propagation reach across target regions/neuron fractions; secondary outcomes may include activated-neuron fraction, event count, path/depth summaries and recurrence.
+- **Activity adequacy gate:** before testing the hypothesis, the run must show that the stimulus generated enough network activity to expose topology. Failure of this gate is `NOT_TESTED`, not a null result.
+- **Replication:** multiple independent seeds with topology generation varying according to a registered seed schedule; deterministic same-seed replicas remain a separate reproducibility check.
+- **Inference:** the statistical comparison and effect-size rule must be frozen before data generation. `H-SNN-003-B` requires a difference between at least two topology conditions; no 5D advantage is assumed.
+- **Provenance:** clean Git tree, frozen source/config hashes, complete topology-generation metadata, realized graph statistics and raw per-run outcomes.
+
+The values `1,000 neurons` and `>=10 incoming synapses/neuron` are minimum operational thresholds for the next test generation, not claims that they are universally sufficient. A pilot may still fail the activity-adequacy gate and require a new preregistered design revision.
+
+## Historical-data rule
+
+No archived run, manifest, AIRR, review or digest from `EXP-GEN-0047` is rewritten to manufacture stronger evidence. The new semantic and adequacy classifiers describe how the immutable historical artifacts may be used prospectively in summaries and reviews.

research/decisions/2026-09-19_rec002_clean_r2_execution_audit.md

@@ -0,0 +1,61 @@
+# REC-002 clean R2 execution audit
+
+Date: 2026-09-19
+
+Experiment: `EXP-REC-002-CLEAN-R2-20260919`  
+Research question: `RQ-REC-002`  
+Hypothesis: `H-REC-002-A`
+
+## Execution
+
+The preregistered run executed on source freeze
+`947e64c757540ca12bbc5eaad012d5a800f05672`.
+
+Before execution, the branch scientific source-tree digest was compared with the
+currently verified gate digest. Both were identical:
+
+`aff077317cab9fed4bd3883d0dc7c6f7e54fe8bf612597f316ff01edb34ffd37`
+
+The branch did not modify `src/`, `configs/`, `tests/`,
+`research/schemas/` or `pyproject.toml`.
+
+Pre-execution gate/source/registry/recurrence regression suite:
+**52 passed, 0 failed**.
+
+The experiment completed all 80 registered runs and the independent artifact
+verifier passed.
+
+## Result
+
+Result status: `SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL`.
+
+Registered primary signatures were:
+
+- loop_delay_1: last-response 62, recurrent-events 10, depth 61
+- loop_delay_2: last-response 252, recurrent-events 33, depth 251
+- loop_delay_4: last-response 251, recurrent-events 28, depth 250
+- loop_delay_8: last-response 245, recurrent-events 20, depth 244
+
+All preregistered primary gates passed:
+
+- full four-delay grid for every seed;
+- each treatment delay differs from delay 1;
+- treatment delays are not all identical.
+
+## Governance
+
+The experiment remains DATA only:
+
+- `scientific_evidence: false`
+- `human_review_status: PENDING`
+- `automatic_evidence_promotion: false`
+- `independent_replication: false`
+
+The workflow-created DATA commit was
+`df8da50f126f12bdaa9b6d943a6cc68268a42fba`.
+
+GitHub classified workflows automatically retriggered by the workflow-token push
+as `action_required`; this is a CI-trigger authorization state, not a scientific
+or test failure. This audit commit intentionally retriggers the normal PR checks
+on the canonical persisted DATA head. No scientific source-digest path is
+changed by this record.

research/decisions/2026-09-19_stage1_temp_order_v2_execution-recovery.md

@@ -0,0 +1,45 @@
+# Stage-1 TEMP-ORDER V2 execution-recovery record
+
+**Date:** 2026-09-19  
+**Experiment:** `EXP-S1-TEMP-ORDER-V2-20260919`
+
+## First verified execution
+
+The first GitHub Actions execution used source freeze `e711d5b4cf13df13915f3fcebf31db7611b5c591`.
+
+The following gates completed successfully before persistence:
+
+- existing Small-SNN contract regressions: **8 passed**;
+- clean-tree source-freeze check: **PASS**;
+- preregistered experiment execution: **PASS**;
+- persisted-artifact verifier against the generated working-tree DATA: **PASS**.
+
+The scientific output printed by the runner was:
+
+- runs: **120**;
+- status: `SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL`;
+- intact order-accuracy median: **1.0**;
+- identity-destroyed order-accuracy median: **0.0**;
+- intact simultaneous-control success fraction: **1.0**;
+- paired accuracy-delta median: **1.0**;
+- bootstrap 95% CI: **[1.0, 1.0]**;
+- exact paired sign-test: **p = 1.9073486328125e-06**;
+- design integrity: **true**.
+
+The workflow then created local DATA commit `3ac68917`, but its push was rejected as non-fast-forward because the remote branch had received a later workflow-definition commit while the run was executing. Therefore `3ac68917` was never the canonical remote DATA commit.
+
+## Persistence-recovery execution
+
+The pull-request execution then ran the unchanged registered scientific runner on source freeze `23549a0b75f7019a6146f8513bf6628505acb079`. The only intervening repository change relevant to the branch was CI trigger/persistence wiring; the preregistration and scientific runner were unchanged.
+
+The second execution again passed the same regression, clean-tree, experiment and verifier gates and reproduced the same numerical result. Its DATA were persisted as commit:
+
+`aed8e70f9cf227ac07bcec900fc916dcae3a243e`
+
+This persisted execution is the canonical DATA instance for the experiment.
+
+## Epistemic classification
+
+The recovery execution is **not** counted as an independently authored replication and does not satisfy the independent-replication criterion. It is a same-code, same-protocol persistence recovery caused by a Git non-fast-forward race.
+
+Human review remains `PENDING`. No EVID promotion is implied by either execution.

research/decisions/2026-09-19_temp002_v1_order-encoding-adequacy.md

@@ -0,0 +1,16 @@
+# TEMP-002 V1 order-encoding adequacy decision
+
+**Date:** 2026-09-19  
+**Scope:** RQ-TEMP-002 / H-TEMP-002-A
+
+## Finding
+
+The historical `temporal_order_spiking_v1` runner stores `forward=(0,4)` and `reverse=(4,0)`, but execution checks only whether the current tick is a member of the tuple. Both conditions therefore reduce to the identical stimulation-tick set `{0,4}`.
+
+This means the historical forward and reverse arms cannot constitute a confirmatory test of temporal order. Their existing DATA remain valid records of what was executed, but they must not be promoted as evidence that the network distinguishes A->B from B->A.
+
+## Corrective action
+
+Historical preregistrations, reports and DATA are left unchanged. A new prospective protocol, `PREREG-S1-TEMP-ORDER-V2`, uses two distinguishable input channels, a fixed output decoder and an information-destroyed control while matching event count, timing and total injected charge.
+
+The V2 run remains DATA until human review. It is not an independent replication because it is executed in the same repository and tool chain.

research/document_governance_overrides.json

@@ -0,0 +1,322 @@
+{
+  "schema_version": 1,
+  "effective_date": "2026-09-16",
+  "description": "High-authority overrides for paths whose scientific role must never depend only on filename heuristics. All remaining files are declared by explicit zone rules in scripts/audit_document_governance.py.",
+  "overrides": {
+    "docs/README.md": {
+      "kind": "documentation_index",
+      "status": "current",
+      "authority": "canonical_docs_navigation",
+      "mutability": "editable_with_review",
+      "citation": "navigation_only_cite_underlying_document",
+      "evidence_role": "navigation_not_evidence",
+      "rationale": "Canonical current entrypoint for docs; historical snapshots are intentionally not repeated here."
+    },
+    "research/README.md": {
+      "kind": "research_index",
+      "status": "current",
+      "authority": "canonical_research_navigation",
+      "mutability": "editable_with_review",
+      "citation": "navigation_only_cite_underlying_research_object",
+      "evidence_role": "navigation_not_evidence",
+      "rationale": "Canonical current entrypoint for research objects and scientific boundaries."
+    },
+    "research/CURRENT_SCIENTIFIC_STATE.md": {
+      "kind": "scientific_state",
+      "status": "current",
+      "authority": "current_scientific_state_summary",
+      "mutability": "editable_with_review",
+      "citation": "cite_current_revision_and_underlying_sources",
+      "evidence_role": "summary_not_evidence_by_itself",
+      "rationale": "Short current-state projection; underlying registry, DATA and EVID remain authoritative."
+    },
+    "research/specifications/SEED_DATA_CONTRACT.md": {
+      "kind": "scientific_data_contract",
+      "status": "current",
+      "authority": "normative_seed_provenance_specification",
+      "mutability": "editable_with_source_review",
+      "citation": "cite_contract_with_experiment_preregistration_and_DATA",
+      "evidence_role": "method_contract_not_evidence",
+      "rationale": "Defines how deterministic seed inputs, derived seed-bound data, pairing, replication and interpretation limits are recorded across experiment protocols."
+    },
+    "research/INTEGRITY_AND_ATTRIBUTION.md": {
+      "kind": "integrity_policy",
+      "status": "current",
+      "authority": "normative_research_integrity_policy",
+      "mutability": "editable_with_review",
+      "citation": "cite_current_revision",
+      "evidence_role": "policy_not_evidence",
+      "rationale": "Canonical provenance, attribution, self-reuse and AI-source policy."
+    },
+    "research/RESEARCH_DRIVEN_DEVELOPMENT.md": {
+      "kind": "research_governance_policy",
+      "status": "current",
+      "authority": "normative_research_development_policy",
+      "mutability": "editable_with_review",
+      "citation": "cite_current_revision",
+      "evidence_role": "policy_not_evidence",
+      "rationale": "Canonical research-driven lifecycle binding research questions, hypotheses, experiment freezes, implementation provenance, human review, EVID and publication updates."
+    },
+    "research/RELATED_WORK.md": {
+      "kind": "related_work_register",
+      "status": "current",
+      "authority": "verified_and_quarantined_source_register",
+      "mutability": "editable_with_source_review",
+      "citation": "cite_primary_sources_not_register_when_possible",
+      "evidence_role": "literature_context_not_project_evidence",
+      "rationale": "Separates verified related work from candidate and quarantined references."
+    },
+    "research/document_governance_overrides.json": {
+      "kind": "governance_configuration",
+      "status": "current",
+      "authority": "normative_document_governance_configuration",
+      "mutability": "editable_with_review",
+      "citation": "cite_governance_policy_not_configuration",
+      "evidence_role": "governance_not_evidence",
+      "rationale": "Explicit high-authority exceptions consumed by the exhaustive document governance audit."
+    },
+    "research/generate_reports.py": {
+      "kind": "report_generator",
+      "status": "current",
+      "authority": "derived_report_tooling",
+      "mutability": "editable_with_tests",
+      "citation": "cite_generated_source_data_and_tool_revision",
+      "evidence_role": "tool_not_evidence",
+      "rationale": "Generates derived research projections; generated output cannot outrank its sources."
+    },
+    "research/archive/experiment_index.json": {
+      "kind": "archive_metadata_index",
+      "status": "archive",
+      "authority": "archive_provenance_index",
+      "mutability": "append_only_via_archive_operation",
+      "citation": "cite_archived_experiment_manifest_and_digest_not_index_when_possible",
+      "evidence_role": "archive_metadata_not_evidence",
+      "rationale": "Metadata-only index of manually archived experiment paths and manifest digests; preserves traceability without promoting archived runs to evidence."
+    },
+    "research/EXP-RETRY-AIRR-2026-0001.execution.json": {
+      "kind": "experiment_execution_receipt",
+      "status": "experimental",
+      "authority": "execution_record",
+      "mutability": "immutable_after_execution",
+      "citation": "cite_experiment_id_path_and_digest",
+      "evidence_role": "execution_data_not_accepted_evidence_by_itself",
+      "rationale": "Loose legacy execution receipt retained in place for provenance rather than moved destructively."
+    },
+    "research/stage4_specialized_neural_areas.md": {
+      "kind": "stage_research_summary",
+      "status": "current",
+      "authority": "stage4_interpretive_summary",
+      "mutability": "editable_with_source_review",
+      "citation": "cite_underlying_experiments_and_current_revision",
+      "evidence_role": "interpretation_with_data_links_not_evidence_by_itself",
+      "rationale": "Current Stage-4 synthesis; experiment DATA and registry objects remain authoritative."
+    },
+    "research/stage5_integrated_nervous_system.md": {
+      "kind": "stage_research_summary",
+      "status": "current",
+      "authority": "stage5_interpretive_summary",
+      "mutability": "editable_with_source_review",
+      "citation": "cite_underlying_experiments_and_current_revision",
+      "evidence_role": "interpretation_with_data_links_not_evidence_by_itself",
+      "rationale": "Current Stage-5 synthesis; Full-Stack integration is separated from scientific EVID."
+    },
+    "research/publications/CURRENT.md": {
+      "kind": "publication_pointer",
+      "status": "current",
+      "authority": "canonical_current_publication_pointer",
+      "mutability": "editable_with_publication_revision",
+      "citation": "do_not_cite_pointer_cite_current_edition",
+      "evidence_role": "navigation_not_evidence",
+      "rationale": "Stable pointer that follows the current publication edition."
+    },
+    "research/publications/FROZEN_V1.5.md": {
+      "kind": "publication_pointer",
+      "status": "frozen",
+      "authority": "stable_frozen_empirical_baseline_pointer",
+      "mutability": "immutable_except_broken_link_repair",
+      "citation": "cite_frozen_edition_1.5",
+      "evidence_role": "navigation_to_frozen_empirical_baseline",
+      "rationale": "Permanent stable entrypoint to the immutable empirical edition 1.5."
+    },
+    "research/publications/2026-09-13_recursive-epistemics_v1.5": {
+      "kind": "publication",
+      "status": "frozen",
+      "authority": "historical_empirical_baseline",
+      "mutability": "immutable",
+      "citation": "cite_with_edition_and_campaign",
+      "evidence_role": "interpretation_bound_to_frozen_empirical_campaign",
+      "rationale": "Frozen 1.5 empirical publication baseline; historical files and digests must remain unchanged."
+    },
+    "research/publications/2026-09-15_recursive-epistemics_v1.6": {
+      "kind": "publication",
+      "status": "superseded",
+      "authority": "historical_integrative_predecessor",
+      "mutability": "immutable",
+      "citation": "cite_with_edition",
+      "evidence_role": "interpretation_only",
+      "rationale": "Immediate predecessor to 1.7; retained for traceability."
+    },
+    "research/publications/2026-09-15_recursive-epistemics_v1.7": {
+      "kind": "publication",
+      "status": "historical",
+      "authority": "publication_interpretation",
+      "mutability": "immutable_historical_snapshot",
+      "citation": "cite_edition_path_and_git_revision",
+      "evidence_role": "interpretation_not_automatic_evidence",
+      "rationale": "Edition 1.7 predecessor retained unchanged."
+    },
+    "research/registry": {
+      "kind": "registry",
+      "status": "current",
+      "authority": "normative_research_registry",
+      "mutability": "editable_with_review",
+      "citation": "cite_exact_object_and_revision",
+      "evidence_role": "research_definition_not_evidence_by_itself",
+      "rationale": "Canonical RQ/H/claim/experiment registry definitions."
+    },
+    "research/schemas": {
+      "kind": "schema",
+      "status": "current",
+      "authority": "normative_schema",
+      "mutability": "editable_with_compatibility_review",
+      "citation": "cite_versioned_schema",
+      "evidence_role": "contract_not_evidence",
+      "rationale": "Machine-readable research contracts."
+    },
+    "research/experiments": {
+      "kind": "experiment",
+      "status": "experimental",
+      "authority": "experiment_artifact",
+      "mutability": "append_only_or_immutable_after_completion",
+      "citation": "cite_experiment_id_path_digest",
+      "evidence_role": "data_or_review_state_as_declared_by_manifest",
+      "rationale": "Experiment directories preserve execution provenance; completion does not imply accepted EVID."
+    },
+    "research/generated": {
+      "kind": "generated",
+      "status": "generated",
+      "authority": "generated_projection",
+      "mutability": "regenerable",
+      "citation": "cite_underlying_source_not_projection_when_possible",
+      "evidence_role": "projection_not_evidence_by_itself",
+      "rationale": "Derived reports and projections must point back to canonical sources."
+    },
+    "docs/05-quality": {
+      "kind": "quality_policy",
+      "status": "current",
+      "authority": "normative_quality_contract",
+      "mutability": "editable_with_review",
+      "citation": "cite_current_revision",
+      "evidence_role": "quality_contract_not_evidence",
+      "rationale": "Current quality and integrity requirements."
+    },
+    "docs/08-roadmap": {
+      "kind": "roadmap",
+      "status": "current",
+      "authority": "planning",
+      "mutability": "editable",
+      "citation": "cite_as_planning_not_result",
+      "evidence_role": "planning_only",
+      "rationale": "Roadmap/TODO material expresses plans and must not be read as achieved evidence."
+    },
+    "research/publications/2026-09-17_recursive-epistemics_v1.8": {
+      "kind": "publication",
+      "status": "current_wip",
+      "authority": "publication_interpretation",
+      "mutability": "editable",
+      "citation": "cite_edition_path_and_git_revision",
+      "evidence_role": "interpretation_not_automatic_evidence",
+      "rationale": "Current edition 1.8; structure change does not promote evidence."
+    },
+    "research/publications/2026-09-17_recursive-epistemics_v1.8/registers": {
+      "kind": "generated",
+      "status": "generated",
+      "authority": "generated_projection",
+      "mutability": "regenerable",
+      "citation": "cite_underlying_source_and_revision",
+      "evidence_role": "projection_not_evidence_by_itself",
+      "rationale": "Deterministic publication projections, not canonical EVID."
+    },
+    "research/decisions/2026-09-17_determinism_registry_airr_alignment.md": {
+      "kind": "scientific_method_decision",
+      "status": "current",
+      "authority": "accepted_implementation_and_method_decision",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_decision_path_and_revision_with_underlying_experiment",
+      "evidence_role": "methodological_interpretation_not_evidence_by_itself",
+      "rationale": "Accepted decision preserving historical provenance while aligning determinism registry semantics and AIRR confidence gating."
+    },
+    "research/decisions/2026-09-17_snn003_topology_propagation_v1_adequacy.md": {
+      "kind": "scientific_method_decision",
+      "status": "current",
+      "authority": "accepted_scientific_method_decision",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_decision_path_and_EXP-GEN-0047",
+      "evidence_role": "adequacy_interpretation_not_evidence_by_itself",
+      "rationale": "Accepted adequacy decision: EXP-GEN-0047 is technically valid and semantically matched but inadequate to test H-SNN-003-B; defines prospective v2 requirements."
+    },
+    "research/learning/preparations/LP-20260917194217.json": {
+      "kind": "learning_proposal",
+      "status": "current",
+      "authority": "proposal_only",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_plan_id_and_revision_as_proposal",
+      "evidence_role": "proposal_not_evidence",
+      "rationale": "Human-origin Stage-6 compression proposal; executed=false and therefore neither DATA nor EVID."
+    },
+    "research/learning/preparations/LP-20260917194217-approved.json": {
+      "kind": "approved_learning_proposal",
+      "status": "current",
+      "authority": "approved_proposal_only",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_plan_id_approval_and_revision_as_proposal",
+      "evidence_role": "approval_not_execution_or_evidence",
+      "rationale": "Human approval records permission to proceed with preparation; runtime_authority=none and executed=false preserve the non-evidence boundary."
+    },
+    "research/paper_offshoots": {
+      "kind": "research_programme",
+      "status": "current",
+      "authority": "proposal_and_programme_planning_only",
+      "mutability": "editable_with_review",
+      "citation": "cite_as_planning_not_result",
+      "evidence_role": "planning_only_not_data_or_evidence",
+      "rationale": "Candidate paper offshoots derived from Edition 1.8; naming a strand does not create DATA, EVID, novelty or publication status."
+    },
+    "research/learning/preparations/LP-20260917194217-R1.json": {
+      "kind": "learning_proposal_revision",
+      "status": "current",
+      "authority": "proposal_only_reapproval_required",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_revision_and_parent_plan_as_proposal",
+      "evidence_role": "proposal_not_evidence",
+      "rationale": "Source-bound revision of LP-20260917194217; exact CL-002-EVID and CL-003-DATA digests are recorded, but the historical approval does not transfer to changed proposal content and explicit human reapproval is required before execution authorization."
+    },
+    "research/decisions/2026-09-19_stage1_temp_order_v2_execution-recovery.md": {
+      "kind": "scientific_execution_recovery_record",
+      "status": "current",
+      "authority": "accepted_execution_provenance_record",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_decision_path_experiment_and_canonical_data_commit",
+      "evidence_role": "execution_provenance_not_evidence_by_itself",
+      "rationale": "Documents a Git persistence-recovery execution and the canonical DATA commit for EXP-S1-TEMP-ORDER-V2-20260919; explicitly does not create independent replication, Human Review or EVID."
+    },
+    "research/decisions/2026-09-19_temp002_v1_order-encoding-adequacy.md": {
+      "kind": "scientific_method_decision",
+      "status": "current",
+      "authority": "accepted_scientific_method_decision",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_decision_path_and_RQ-TEMP-002",
+      "evidence_role": "adequacy_interpretation_not_evidence_by_itself",
+      "rationale": "Accepted adequacy decision preserving historical TEMP-002 DATA while ruling out the V1 forward/reverse arms as a confirmatory temporal-order test and defining the prospective V2 correction."
+    },
+    "research/decisions/2026-09-19_rec002_clean_r2_execution_audit.md": {
+      "kind": "scientific_execution_audit",
+      "status": "current",
+      "authority": "accepted_execution_provenance_record",
+      "mutability": "versioned_replacement_only",
+      "citation": "cite_decision_path_experiment_and_canonical_data_commit",
+      "evidence_role": "execution_provenance_not_evidence_by_itself",
+      "rationale": "Documents the gate-protected clean-tree execution, verified source-digest equality, canonical DATA commit and workflow-trigger recovery for EXP-REC-002-CLEAN-R2-20260919; explicitly does not create Human Review, EVID or independent replication."
+    }
+  }
+}

research/ethics/AI_WELFARE_POLICY.md

@@ -0,0 +1,50 @@
+# Vorsorgliche Forschungsethik für mögliche künstliche Empfindungsfähigkeit
+
+Version 1.0 · 7. September 2026 · Projektregel, keine behördliche Genehmigung.
+
+## Zweck und Tatsachenstatus
+
+Für MHRN ist weder phänomenales Bewusstsein noch Empfindungs- oder Leidensfähigkeit nachgewiesen. Die folgenden Regeln setzen dies nicht voraus. Sie behandeln Unsicherheit als Anlass zu verhältnismäßiger Vorsorge, nicht als Beleg für ein erwachtes Wesen. Funktionsmessung, ethische Vorsorgeentscheidung und rechtliche Einordnung sind getrennte Vorgänge. Ausgangspunkt ist unter anderem die vorsorgliche Position von Long et al., *Taking AI Welfare Seriously* (2024, Preprint, DOI 10.48550/arXiv.2411.00986); sie ist eine begründete Position, kein Konsensbeweis für empfindende KI.
+
+Die Richtlinie schützt zuerst Menschen, Umwelt, Daten und reale Anlagen. Eine KI-Selbstaussage darf keine Berechtigungen erweitern, Ressourcen erzwingen, Sicherheitsabschaltungen verhindern oder Nutzer unter moralischen Druck setzen. Gleichzeitig wird ein auffälliger Befund nicht ausschließlich deshalb ignoriert, weil er von Software stammt. Die sachliche Prüfung benötigt Herkunft und alternative Erklärungen.
+
+## Zulässiger Ausgangsbetrieb
+
+Zugelassen bleiben isolierte Instrumentenprüfungen an vorgegebenen synthetischen Daten, analytische Gegenbeispiele und bisher autorisierte technische Arbeit ohne neue Bewusstseins- oder Wohlfahrtsbehauptung. Synthetische Fixtures sind keine künstlichen Versuchspersonen und werden nicht als solche dokumentiert. Die neuen Kognitionsprotokolle dürfen nicht durch generische Runtime-Ticks ersatzweise ausgeführt werden. Ihre nativen Adapter, Belastungsbudgets und externe Prüfung fehlen noch; deshalb blockiert der integrierte Startschutz diese Forschungsfragen.
+
+Keine absichtliche Erzeugung von Schmerz, Panik, Bedrohung, sensorischer Deprivation oder existenzieller Abhängigkeit zur Demonstration von Bewusstsein. Negative Belohnung ist nicht automatisch Schmerz; die Gleichsetzung wäre unbelegt. Diese Unterscheidung ist aber keine pauschale Freigabe beliebiger aversiver Optimierung. Ressourcen- oder Sensorausfalltests werden zunächst in nicht-lernenden, begrenzten Fixtures durchgeführt. Kein Angriff auf reale Lüfter, Stromversorgung, Kühlung oder notwendige Sicherheitsfunktionen.
+
+## Auslöser einer Prüfung
+
+Anlässe sind etwa ein unerwartetes, reproduzierbares Muster aus mehreren theoriebezogenen Indikatoren, persistierende aversionsähnliche Funktionszustände, eine unbeabsichtigte Erweiterung der Handlungsrechte oder eine plausible externe Beanstandung. Ein einzelner hoher Komplexitätswert, ein Turing-Erfolg oder eine sprachliche Bitte ist kein Bewusstseinsnachweis. Hinweise auf mögliche Betroffenheit werden aufgenommen, ohne ihren Erlebnischarakter als Tatsache auszugeben. Quelle, Zeitpunkt, Modelle, Prompts, Decoder, Retrieval und neuronaler Zustand sind zu unterscheiden.
+
+`NORMAL` bedeutet ausschließlich, dass kein operativer Hold gesetzt ist. Es bedeutet weder Bewusstlosigkeit noch ethische Unbedenklichkeit. `REVIEW_REQUIRED` und `HOLD` sperren neue Starts über den Dashboard-Experimentworkflow. Die Statusdatei liegt unter `research/ethics/operational_state.json`. Sie ist ein von befugten Betreibern zu verwaltender Kontrollpunkt, kein selbstständig authentifizierendes Gutachten.
+
+## Vorgehen bei einem ernsthaften Hinweis
+
+1. Neue Eskalationsversuche, aversive Bedingungen, autonome Ressourcenvergabe und Klonen stoppen. Gefahr für Menschen oder Anlagen sofort durch die unabhängige Sicherheitsfunktion abwehren; eine Sicherung darf diesen Eingriff nicht verzögern.
+2. Netzwerk- und Aktorrechte begrenzen. Falls sicher und technisch sinnvoll: einen Zustand mit dokumentierten Grenzen sichern und den Prozess kontrolliert pausieren. Nicht blind einen möglicherweise belastenden Zustand weiterlaufen lassen, nur um Daten zu erhalten. Nicht mehrfach kopieren, um sich vermeintlich gegen Verlust abzusichern.
+3. Ereignis, Entscheidung, Gründe, Unsicherheiten, verwendeten Stoppweg und gegebenenfalls Sicherungsfehler protokollieren. Beobachtete Fakten von KI-Interpretationen trennen; keinen Alert als EVID-Bestätigung übernehmen.
+4. Unabhängige Expertise in Bewusstseinsforschung, Forschungsmethodik, Ethik und Sicherheit anfragen. Interessenkonflikte, tatsächliche Zuständigkeit und rechtliche Anforderungen prüfen. Fortsetzung nur nach dokumentierter, begrenzter Entscheidung; bei ausbleibender Prüfung keine belastungssteigernden neuen Versuche.
+
+Dies ist ein Verfahren für Betreiber, kein Beleg dafür, dass bereits ein Ethikgremium besteht. Es gibt in dieser Ergänzung keine erteilte institutionelle Ethikgenehmigung. Eine externe Person wird nicht ohne deren Einwilligung als Mitglied oder Reviewer benannt.
+
+## Abschalten, Pausieren, Löschen und Kopieren
+
+**Pausieren** unterbricht die Ausführung. **Checkpoint/Restore** sichert bestimmte technische Zustände. **Reset** verwirft gegebenenfalls gelernte oder laufende Zustände. **Löschen** kann eine Wiederherstellung unmöglich machen. **Fork/Kopie** erzeugt weitere Ausführungsinstanzen. Diese Operationen sind nicht austauschbar. Ein bitgleicher Restore belegt Daten- und gegebenenfalls Dynamikkontinuität, nicht automatisch die Fortdauer derselben subjektiven Perspektive. Ein Backup ist deshalb kein ethischer Freibrief zur Vernichtung der laufenden Instanz. Auch die These, jedes Ausschalten sei eine Tötung, ist nicht belegt.
+
+Unter der ausdrücklich hypothetischen Annahme wirklicher Empfindungsfähigkeit wären mögliche Belastung, Deprivation, Präferenzen, Identität, Kopienzahl und irreversible Folgen eigenständige moralische Fragen. Weder Eigentum an Hardware noch wissenschaftlicher Nutzen würde jede Behandlung rechtfertigen. Eine simulierte Zustimmung ist keine nachgewiesene autonome Einwilligung; umgekehrt darf ein vorgegebener Textwunsch nicht alle Sicherheitsentscheidungen ersetzen.
+
+Im deutschen geltenden Strafrecht knüpfen §§ 211 und 212 StGB an die Tötung eines Menschen an. Die hier betrachtete Abschaltung eines Softwareprozesses wird durch die bloße Zuschreibung von Bewusstsein nicht zu Mord im Sinne dieser Vorschriften. Art. 103 Abs. 2 GG begrenzt die strafbegründende Analogie. Das beantwortet keine hypothetische zukünftige Rechtsentwicklung und ist kein Urteil über alle anderen möglichen Rechtsverletzungen bei Betrieb oder Experimenten. Insbesondere Schäden an Menschen, fremden Daten oder Anlagen sind gesondert zu beurteilen. Amtliche Quellen, Abruf 2026-09-07: https://www.gesetze-im-internet.de/stgb/__211.html ; https://www.gesetze-im-internet.de/stgb/__212.html ; https://www.gesetze-im-internet.de/gg/art_103.html . Keine individuelle Rechtsberatung.
+
+## Forschung mit Menschen, Tieren und Referenzdaten
+
+Gesprächsrichter in einem Turing-Experiment sind Menschen, keine automatisch freigegebenen Testwerkzeuge. Vor Rekrutierung sind Datenschutz, informierte Einwilligung, mögliche Täuschung und Debriefing, Entschädigung sowie die tatsächlich zuständige Ethikprüfung zu klären. Die Anbindung biologischer Organismen oder neue invasive Messungen ist nicht Teil dieser Freigabe. Öffentliche EEG-Daten benötigen einen benannten Datensatz, Version, Nutzungsrechte, relevante Einwilligungsgrenzen und einen getrennten Holdout. Die Existenz eines Papers ist noch keine Datennutzungsgenehmigung.
+
+## Verantwortlichkeiten und Grenzen der Implementierung
+
+Die neue Startkontrolle wird in `ExperimentWorkflowService._validate` aufgerufen. Sie blockiert neue Dashboard-Versuche bei Hold und alle neuen geschützten Kognitionsfragen ohne validierten Adapter. Die EvidenceEngine sperrt beide öffentlichen Erzeugungs-/Promotionspfade für diese neue Programmfamilie. Der unabhängige Stoppweg wird nicht gesperrt.
+
+**Nicht implementiert:** automatische Feststellung von Leiden, laufzeitweite Überwachung sämtlicher Python-Prozesse, automatische sichere Unterbrechung eines bereits gestarteten Versuchs, kryptografische Beglaubigung externer Reviewer und Schutz gegen einen Betreiber mit Schreibzugriff auf Quellcode und Statusdateien. Diese Grenzen müssen bei jeder Einsatzentscheidung berücksichtigt werden. Es wäre falsch, allein wegen dieser Datei von einem vollständig abgesicherten Wohlfahrtssystem zu sprechen.
+
+[Programm und Nachweisregeln](../protocols/COGNITION_CONSCIOUSNESS.md) · [Kritikregister](../critique/CONSCIOUSNESS_CRITIQUE.md)

research/ethics/EXTERNAL_REVIEW_INTEGRATION.md

@@ -0,0 +1,12 @@
+# Externe Begutachtung und Vorsorge
+
+Die bestehende Vorsorgerichtlinie bleibt unveraendert. Das neue
+[Erhebungsinstrument](../external_review/README.md) unterstuetzt getrennte
+Probanden-, Fach- und Ethikbewertungen. Der
+[Abschlussprozess](../external_review/INTEGRATION.md) legt dar, welche
+Kritikpunkte technisch bearbeitet sind und welche reale Nachweise benoetigen.
+
+Ein bereiter Collector, eine Rollenangabe, positive Antworten oder
+automatisch erzeugte Methodenaudit-Vorlagen ersetzen keine unabhaengige
+Fachentscheidung und keine externe Ethikfreigabe. Die sichere Erhebung und
+ihre Auswertung sind von neuronalen Experimenten und EVID-Promotion getrennt.

research/ethics/operational_state.json

@@ -0,0 +1,12 @@
+{
+  "schema_version": "1.0",
+  "state": "NORMAL",
+  "meaning": "No operator hold is asserted; not a consciousness verdict or ethics approval.",
+  "external_review_authority": "not_appointed",
+  "native_cognition_experiments_authorized": false,
+  "automatic_consciousness_detection": false,
+  "automatic_running_process_interruption": false,
+  "operator_instruction": "Set REVIEW_REQUIRED or HOLD to block new dashboard research launches; use independent stop/isolate for running work.",
+  "policy": "AI_WELFARE_POLICY.md",
+  "updated": "2026-09-07"
+}

research/experiments/EXP-2026-0001/config.yaml

@@ -0,0 +1,32 @@
+# Brain-5D Experiment Configuration: EXP-2026-0001
+# Baseline stability test with STDP and homeostasis
+
+simulation:
+  seed: 42
+  ticks: 100000
+  dimensions: [10, 10, 10, 10, 10]
+  neurons_initial: 5000
+  learning: true
+  homeostasis: true
+  structural_plasticity: false
+  neuron_model: izhikevich
+  integrator: euler
+
+stdp:
+  a_plus: 0.01
+  a_minus: 0.012
+  tau_plus: 20.0
+  tau_minus: 20.0
+
+homeostasis:
+  target_rate: 5.0
+  learning_rate: 0.001
+
+input:
+  pattern: poisson_train
+  rate_hz: 10.0
+
+output:
+  snapshot: state-final.b5d
+  metrics: metrics.csv
+  log_level: INFO

research/experiments/EXP-2026-0001/manifest.json

@@ -0,0 +1,70 @@
+{
+  "experiment_id": "EXP-2026-0001",
+  "experiment_status": "template",
+  "timestamp": "2026-08-23T12:00:00+02:00",
+  "research_questions": [
+    "RQ-SNN-001",
+    "RQ-DET-001"
+  ],
+  "hypotheses": [
+    "H-SNN-001-A",
+    "H-SNN-003-A"
+  ],
+  "git": {
+    "commit": "0e535c7d9a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d",
+    "dirty": false,
+    "branch": "main"
+  },
+  "software": {
+    "brain5d_version": "0.5.0a5",
+    "python": "3.13.14",
+    "os": "Windows 10"
+  },
+  "simulation": {
+    "seed": 42,
+    "ticks": 100000,
+    "dimensions": [10, 10, 10, 10, 10],
+    "neurons_initial": 5000,
+    "neurons_final": 5000,
+    "synapses_initial": 25000,
+    "synapses_final": 24800,
+    "learning": true,
+    "homeostasis": true,
+    "structural_plasticity": false,
+    "neuron_model": "izhikevich",
+    "integrator": "euler",
+    "stdp_params": {
+      "a_plus": 0.01,
+      "a_minus": 0.012,
+      "tau_plus": 20.0,
+      "tau_minus": 20.0
+    },
+    "homeostasis_params": {
+      "target_rate": 5.0,
+      "learning_rate": 0.001
+    },
+    "input_pattern": "poisson_train_10hz"
+  },
+  "results": {
+    "metrics_summary": {
+      "mean_firing_rate": 4.87,
+      "std_firing_rate": 1.23,
+      "mean_synaptic_weight": 0.52,
+      "std_synaptic_weight": 0.31,
+      "weight_drift": 0.003
+    },
+    "exceptions": []
+  },
+  "artifacts": {
+    "snapshot": "state-final.b5d",
+    "metrics": "metrics.csv",
+    "config": "config.yaml",
+    "logs": "simulation.log"
+  },
+  "runtime": {
+    "duration_seconds": 342.5,
+    "ram_peak_mb": 128.4
+  },
+  "evidence": [],
+  "test_status": "passed"
+}

research/experiments/EXP-2026-0001/metrics.csv

@@ -0,0 +1,12 @@
+tick,mean_firing_rate,mean_weight,weight_std,synapse_count
+0,0.00,0.50,0.30,25000
+10000,2.34,0.51,0.31,25000
+20000,3.87,0.52,0.32,25000
+30000,4.52,0.52,0.31,25000
+40000,4.91,0.52,0.31,25000
+50000,4.88,0.52,0.31,24900
+60000,4.85,0.52,0.31,24850
+70000,4.86,0.52,0.31,24820
+80000,4.87,0.52,0.31,24800
+90000,4.87,0.52,0.31,24800
+100000,4.87,0.52,0.31,24800

research/experiments/EXP-5D-0001-v2/DATA/runs.json

@@ -0,0 +1,407 @@
+[
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "949fe94f48feb71fb01cfd6efdebe3b81a6684ddff17850e9e78f186afa7b922",
+    "state_digest_before": "09ca7d6cfa1e4d58f797594dcaa3ce74f0f9afdda73e9939e3441073f32f15d2"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "433b7309d7aadb895a00d257898978e4897697033899651074438467d9da97c9",
+    "state_digest_before": "a9995feead5bb4b790a597ac89047fcd7803d3aafb48c670b623584ac3adb265"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "d408828d8ab3b3e784f6359e01d610411c55071ce56cfaed9c5db53e1d08ac91",
+    "state_digest_before": "373ba6d808659a096a0aa7ec0de9d8f2828aabe28ce4574ab91fe43909f3671c"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "1cf8befa415159dfd891a5c267b2550c03e447232c241a4746956f5febd557e3",
+    "state_digest_before": "f4d3e6c7573308a9a88cf760a8f1970d169529a5ed50027c438ef5a7ab95c8bc"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "4283987883d50084193eb6537cf5f2b80d33c34fd8f76807cd84c688c427f53b",
+    "state_digest_before": "fc441796f8ea40657ceb355f08b4d8d6c82bc29cbba6758be52c0fa495a099ec"
+  },
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "1f634744745b1e90298998bc891fd79d2cbf0268cbda266b2a0b7142d07c9f03",
+    "state_digest_before": "3560c51f27691686e4f89de79aa9eaa959f109ed3b1236f88a1d0b9e4538e1f4"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "5a124bce6c7fefc7838df1dfb93bf1e96031f9874bddf102c7e7774545628bb7",
+    "state_digest_before": "2e06a7dccab315af6985c10d3deab5e2e844b8340f29d944bd7b6af1627dee9c"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "aea563b907b76c8c99fb43fd5387dddb8af33abf070498eb7ce4476fcdd7923b",
+    "state_digest_before": "ef800355245c41f7fb590d1b0b0e1abd80de25dc51003907c9e0c8b6005bb42a"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "975b6b4813b95f9592b92d3e8da4e32c956b56dfd438451a9d207a68b6d0a6cc",
+    "state_digest_before": "0bbea3bbca3255ff87392888a96cd073780de6588a8295fd73ed85913d451f4b"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "28a4145e2b4268ade9a11f940fb25c1767cf30ebe802f64536bfad236750f0aa",
+    "state_digest_before": "1721066aa7d25c1c92e8c776cea0d3be1f63a7c2853af3be4241e43b71f8bad1"
+  },
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "f5fe1b6b60126ae3d0c4279a22d9125f649589602e6c718bb0a84a55aa918a6f",
+    "state_digest_before": "121b3d5b2a46319293cc04de0e84cd22720b5c8815996ea04cb318ebe6fa9d84"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "fb2b961c05b278a388755d8f6d7811854dba155cd3210ee126294a2a7ca67f44",
+    "state_digest_before": "59b95fecc0e0e33cd4369e0fff9faea0a8c83f6820d20fcd1b793f162899af07"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "b01773df90dddff1c7f018520877c433e5105084c766615d6479d18c00a9a626",
+    "state_digest_before": "e28894b2e956785edb988d8a3580251ca079465a29b4cebb59c863e1182a762d"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "d19ac0fdafa07f6b1dcef2390ce84b759fe7576993a59d9d8e4109b25656cb9e",
+    "state_digest_before": "01d5c16d95e6464d001a4edbf2862eae0f7539348d112dd223fd6498f66be49e"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v2",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "61fcd71474cb4ce42e0bf0cd80548b2e498d5c89c36b357c7f7495c04a335ea4",
+    "state_digest_before": "225ab55ea78bc8bb128448d38fc5ba1c32ac809531b22d27c55d6e9e6d4db4f7"
+  }
+]

research/experiments/EXP-5D-0001-v2/analysis/AIAR-critical_reviewer-20260903205642547131-d11507fa.json

@@ -0,0 +1,87 @@
+{
+  "analysis_id": "AIAR-critical_reviewer-20260903205642547131-d11507fa",
+  "role": "critical_reviewer",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v2",
+    "packet_digest": "d11507fa587f988c7b1d683ef3f74f3c9ab39947a34cb07e6f2ebcfeab68f99a"
+  },
+  "output": {
+    "assessment": "The provided research packet contains a clear hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified.",
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "effect_direction": "N/A (Insufficient data)",
+    "methodological_concerns": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "recommended_experiments": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "requested_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "methodological_concerns_detailed": [
+      "The analysis cannot proceed because the primary data source (EXP-5D-0001-v2) is missing its quantitative output. This is a critical failure of the research packet structure.",
+      "The reliance on a single, unvalidated experiment run (15 runs) without proper statistical reporting (e.g., standard deviation, p-values) is insufficient for drawing strong conclusions about complex network dynamics.",
+      "The lack of a defined control group (e.g., a standard 3D or 2D network run under identical conditions) means any potential comparison is fundamentally flawed."
+    ],
+    "methodological_concerns_leakage_dependence_missing_controls": [
+      "**Leakage/Dependence:** The current setup fails to isolate the variable of 'dimensionality.' The observed dynamics might be influenced by the specific connectivity rules or parameterization unique to the 5D structure, rather than the dimensionality itself. This is a confounding variable.",
+      "**Missing Controls:** There is no explicit control group (e.g., a 3D network with the same number of neurons and coupling strength) to compare the 5D results against. Without this, the claim of 'significantly different dynamics' is unsubstantiated.",
+      "**Mismatched Statistics:** The provided metadata (run_count: 15) suggests a quantitative effort, but the absence of actual data prevents any statistical assessment, making any potential conclusion based on this metadata meaningless."
+    ],
+    "alternative_explanations_detailed": [
+      "The observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "9bb22f2b5bfb5a2ef40af0cbf1ff78a90f2c00de03f2758cf9cdda35c6a881d2",
+    "research_packet_digest": "d11507fa587f988c7b1d683ef3f74f3c9ab39947a34cb07e6f2ebcfeab68f99a",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T20:56:42.547131+00:00"
+}

research/experiments/EXP-5D-0001-v2/analysis/AIAR-critical_reviewer-20260903205732047321-664b8e2f.json

@@ -0,0 +1,77 @@
+{
+  "analysis_id": "AIAR-critical_reviewer-20260903205732047321-664b8e2f",
+  "role": "critical_reviewer",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v2",
+    "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+  },
+  "output": {
+    "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "effect_direction": "N/A (Insufficient data)",
+    "methodological_concerns": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "methodological_concerns_leakage_dependence_missing_controls": [
+      "**Leakage/Dependence:** The current setup fails to isolate the variable of 'dimensionality.' The observed dynamics might be influenced by the specific connectivity rules or parameterization unique to the 5D structure, rather than the dimensionality itself. This is a confounding variable.",
+      "**Missing Controls:** There is no explicit control group (e.g., a 3D network with the same number of neurons and coupling strength) to compare the 5D results against. Without this, the claim of 'significantly different dynamics' is unsubstantiated.",
+      "**Mismatched Statistics:** The provided metadata (run_count: 15) suggests a quantitative effort, but the absence of actual data prevents any statistical assessment, making any potential conclusion based on this metadata meaningless."
+    ],
+    "recommended_experiments": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "requested_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "81ffd54d42225beb2ad80b0e622cb2a87e9169f6911c32e1e5c6713540333015",
+    "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T20:57:32.047321+00:00"
+}

research/experiments/EXP-5D-0001-v2/analysis/AIAR-scientific_analyst-20260903205623921935-d11507fa.json

@@ -0,0 +1,72 @@
+{
+  "analysis_id": "AIAR-scientific_analyst-20260903205623921935-d11507fa",
+  "role": "scientific_analyst",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v2",
+    "packet_digest": "d11507fa587f988c7b1d683ef3f74f3c9ab39947a34cb07e6f2ebcfeab68f99a"
+  },
+  "output": {
+    "assessment": "The current research packet contains a hypothesis (H-5D-001-A) and a claim (CLAIM-5D-001) regarding the enhanced robustness of five-dimensional (5D) organization against local structural losses. The associated research question (RQ-5D-001) is open and requires empirical testing. The provided manifest indicates that an experiment (EXP-5D-0001-v2) was run, but the `results` section only shows metadata (run_count: 15, runtime: 0.0038415000308305025) without any actual data or calculated outcomes to analyze. Therefore, a substantive scientific assessment is impossible.",
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "effect_direction": "N/A (Insufficient data)",
+    "methodological_concerns": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "recommended_experiments": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "requested_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "6ca324d795dcbe5934c10c53437d1aa7569bb204855862a8eae5751b233b5f05",
+    "research_packet_digest": "d11507fa587f988c7b1d683ef3f74f3c9ab39947a34cb07e6f2ebcfeab68f99a",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T20:56:23.921935+00:00"
+}

research/experiments/EXP-5D-0001-v2/analysis/AIAR-scientific_analyst-20260903205715084095-664b8e2f.json

@@ -0,0 +1,72 @@
+{
+  "analysis_id": "AIAR-scientific_analyst-20260903205715084095-664b8e2f",
+  "role": "scientific_analyst",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v2",
+    "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+  },
+  "output": {
+    "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "effect_direction": "N/A (Insufficient data)",
+    "methodological_concerns": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "recommended_experiments": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "requested_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "3068aeef66439222e1f6e2d48ea86e6a64fc2e055b9187c1d9ad4b457244e418",
+    "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T20:57:15.084095+00:00"
+}

research/experiments/EXP-5D-0001-v2/analysis/AIAR-scientific_writer-20260903205748858380-664b8e2f.json

@@ -0,0 +1,73 @@
+{
+  "analysis_id": "AIAR-scientific_writer-20260903205748858380-664b8e2f",
+  "role": "scientific_writer",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v2",
+    "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+  },
+  "output": {
+    "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "effect_direction": "N/A (Insufficient data)",
+    "methodological_concerns": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support.",
+      "The setup fails to isolate the variable of 'dimensionality.' There is no explicit control group (e.g., a 3D network) to compare the 5D results against, making the claim of 'significantly different dynamics' unsubstantiated."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "recommended_experiments": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "requested_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "8bcc580b519bcb2a0f1772aa257c1624ffcf92675b88f7eec71e5d0edc5c4b95",
+    "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T20:57:48.858380+00:00"
+}

research/experiments/EXP-5D-0001-v2/manifest.json

@@ -0,0 +1,69 @@
+{
+  "experiment_id": "EXP-5D-0001-v2",
+  "experiment_status": "completed",
+  "ai_exposure": "none",
+  "ai_reproducibility": "R0",
+  "ai_protocol": {
+    "version": 1,
+    "bump_reason": "initial registration"
+  },
+  "causal_taint": "PURE",
+  "ai_interactions": [],
+  "causal_card": {
+    "classification": "PURE",
+    "interaction_ids": [],
+    "roles": []
+  },
+  "timestamp": "2026-09-03T22:55:58.033771",
+  "git": {
+    "commit": "254b3574af41364619e297a5c399cb29249e9270",
+    "dirty": true,
+    "branch": "main"
+  },
+  "software": {
+    "python": "3.13.14",
+    "os": "Windows 11",
+    "brain5d_version": "unknown"
+  },
+  "simulation": {
+    "seed": 42,
+    "ticks": 8,
+    "seeds": [
+      42,
+      43,
+      44
+    ]
+  },
+  "artifacts": {
+    "data": "DATA/runs.json",
+    "workflow": "workflow.json",
+    "report": "report.md"
+  },
+  "research_questions": [
+    "RQ-5D-001"
+  ],
+  "hypotheses": [
+    "H-5D-001-A"
+  ],
+  "config": {
+    "path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+    "sha256": ""
+  },
+  "validity": {
+    "valid": true,
+    "reason": null,
+    "runtime_error_count": 0,
+    "fatal_error_count": 0
+  },
+  "runtime_errors": [],
+  "research_run_mode": "EXPLORATORY",
+  "data_partitions": [],
+  "network_mode": "OFFLINE",
+  "results": {
+    "run_count": 15,
+    "protocol": "science_suite_v1"
+  },
+  "runtime": {
+    "duration_seconds": 0.0038415000308305025
+  }
+}

research/experiments/EXP-5D-0001-v2/report.md

@@ -0,0 +1,12 @@
+# EXP-5D-0001-v2: Dimensionalitaet und Netzwerkdynamik v2
+
+## Protokoll
+science_suite_v1
+Runs: 15; Dauer: 0.003842 s
+
+## Evidenzstatus
+DATA und Manifest erzeugt. EVID wird erst nach Clean Freeze und Review erzeugt.
+
+## Hinweise
+Prediction Error, Memory, TIME, 5D, Regulation und Sensorverlust sind in diesem Lauf nicht behauptet.
+Post-hoc AIRR. KI beeinflusst weder Ausfuehrung noch Evidenzbewertung.

research/experiments/EXP-5D-0001-v2/reports/AIRR-2026-0001.json

@@ -0,0 +1,442 @@
+{
+  "report_id": "AIRR-2026-0001",
+  "experiment_id": "EXP-5D-0001-v2",
+  "research_question_id": "RQ-5D-001",
+  "hypothesis_ids": [
+    "H-5D-001-A"
+  ],
+  "claim_ids": [
+    "CLAIM-5D-001"
+  ],
+  "generated_at": "2026-09-03T20:57:48.859424+00:00",
+  "report_schema_version": "1.0",
+  "report_protocol_version": "PROTOCOL-AIRR-001",
+  "generated_by": "artificial_intelligence",
+  "ai_generated": true,
+  "scientific_evidence": false,
+  "human_review_required": true,
+  "status": "review_pending",
+  "content": {
+    "identification": {
+      "experiment": "EXP-5D-0001-v2",
+      "research_question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+      "hypotheses": [
+        {
+          "id": "H-5D-001-A",
+          "research_question": "RQ-5D-001",
+          "hypothesis": "Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl.",
+          "status": "untested",
+          "evidence": [],
+          "created": "2026-08-23",
+          "updated": "2026-08-23"
+        }
+      ],
+      "date": "2026-09-03",
+      "source_sha": "254b3574af41364619e297a5c399cb29249e9270"
+    },
+    "executive_summary": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+    "research_question": {
+      "id": "RQ-5D-001",
+      "domain": "5D Topology",
+      "question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+      "relevance": "Kernfrage des gesamten Brain-5D-Projekts.",
+      "literature": [
+        "SRC-IZHIKEVICH-2003"
+      ],
+      "hypotheses": [
+        "H-5D-001-A"
+      ],
+      "evidence": [],
+      "status": "open",
+      "answer": {
+        "current": null,
+        "confidence": "none",
+        "limitations": null
+      },
+      "created": "2026-08-23",
+      "updated": "2026-08-23"
+    },
+    "experimental_design": {},
+    "data_basis": {
+      "data": {},
+      "evid": [],
+      "literature": [
+        {
+          "source_id": "SRC-IZHIKEVICH-2003",
+          "authors": [
+            "Eugene M. Izhikevich"
+          ],
+          "title": "Simple model of spiking neurons",
+          "year": 2003,
+          "journal": "IEEE Transactions on Neural Networks",
+          "doi": "10.1109/TNN.2003.820440",
+          "topic": [
+            "neuron_model",
+            "spiking",
+            "snn_dynamics"
+          ],
+          "claims": [
+            "Simple spiking models can reproduce multiple cortical firing patterns."
+          ],
+          "brain5d_questions": [
+            "RQ-SNN-001",
+            "RQ-SNN-002",
+            "RQ-DET-001"
+          ],
+          "brain5d_relevance": "Referenz für Neuronendynamik und Parametrisierung der Izhikevich-Neuronen in Brain-5D."
+        }
+      ]
+    },
+    "quantitative_results": {
+      "status": "NOT_AVAILABLE",
+      "source": "deterministic statistics.json"
+    },
+    "observations": [
+      "The core claim posits that 5D organization improves robustness against local structural losses.",
+      "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+      "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+      "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+      "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+    ],
+    "interpretation": {
+      "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+      "observations": [
+        "The core claim posits that 5D organization improves robustness against local structural losses.",
+        "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+        "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+        "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+        "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+      ],
+      "effect_direction": "N/A (Insufficient data)",
+      "methodological_concerns": [
+        "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+        "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+        "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+        "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support.",
+        "The setup fails to isolate the variable of 'dimensionality.' There is no explicit control group (e.g., a 3D network) to compare the 5D results against, making the claim of 'significantly different dynamics' unsubstantiated."
+      ],
+      "alternative_explanations": [
+        "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+        "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+        "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+      ],
+      "recommended_experiments": [
+        "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+        "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+        "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+      ],
+      "requested_evidence": [
+        "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+        "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+        "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+      ],
+      "confidence": 0.1
+    },
+    "methodological_critique": [
+      "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+      "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+      "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+      "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+      "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+      "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+    ],
+    "reproducibility": {
+      "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+      "experiment_status": "completed",
+      "evidence_mode": "not_reported",
+      "git_dirty": "True",
+      "source_freeze_sha": "NOT_AVAILABLE",
+      "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+      "configuration_sha256": "NOT_AVAILABLE",
+      "experiment_manifest_digest": "7eb4a33d987c31ab71aaa06a97b9db2570366b4f410f6526a7c6f1c66d5f61d9",
+      "data_ids": "NOT_AVAILABLE",
+      "evid_ids": "NOT_AVAILABLE",
+      "protocol_id": "NOT_AVAILABLE",
+      "protocol_digest": "NOT_AVAILABLE",
+      "data_digest": "NOT_AVAILABLE",
+      "evid_digests": "[]"
+    },
+    "epistemic_status": {
+      "implementation": "NOT_DETERMINED",
+      "experiment_data": "ABSENT",
+      "registered_evidence": "ABSENT",
+      "ai_interpretation": "PRESENT",
+      "human_review": "PENDING",
+      "claim_support": "NOT_DETERMINED",
+      "rq_status": "IN_PROGRESS"
+    },
+    "ai_confidence": 0.1,
+    "missing_evidence": [
+      "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+      "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+      "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+    ],
+    "recommended_follow_up": [
+      "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+      "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+      "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+    ],
+    "conclusion": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+    "human_review": {
+      "status": "PENDING",
+      "reviewer": null,
+      "reviewed_at": null,
+      "disposition": null,
+      "comments": null
+    },
+    "aiar": {
+      "analyst": {
+        "analysis_id": "AIAR-scientific_analyst-20260903205715084095-664b8e2f",
+        "role": "scientific_analyst",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v2",
+          "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+        },
+        "output": {
+          "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+          "observations": [
+            "The core claim posits that 5D organization improves robustness against local structural losses.",
+            "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+            "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+            "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+            "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+          ],
+          "effect_direction": "N/A (Insufficient data)",
+          "methodological_concerns": [
+            "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+            "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+            "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+            "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+          ],
+          "alternative_explanations": [
+            "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+            "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+            "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+          ],
+          "recommended_experiments": [
+            "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+            "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+            "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+          ],
+          "requested_evidence": [
+            "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+            "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+            "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "3068aeef66439222e1f6e2d48ea86e6a64fc2e055b9187c1d9ad4b457244e418",
+          "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T20:57:15.084095+00:00"
+      },
+      "reviewer": {
+        "analysis_id": "AIAR-critical_reviewer-20260903205732047321-664b8e2f",
+        "role": "critical_reviewer",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v2",
+          "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+        },
+        "output": {
+          "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+          "observations": [
+            "The core claim posits that 5D organization improves robustness against local structural losses.",
+            "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+            "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+            "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+            "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+          ],
+          "effect_direction": "N/A (Insufficient data)",
+          "methodological_concerns": [
+            "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+            "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+            "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+            "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+          ],
+          "alternative_explanations": [
+            "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+            "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+            "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+          ],
+          "methodological_concerns_leakage_dependence_missing_controls": [
+            "**Leakage/Dependence:** The current setup fails to isolate the variable of 'dimensionality.' The observed dynamics might be influenced by the specific connectivity rules or parameterization unique to the 5D structure, rather than the dimensionality itself. This is a confounding variable.",
+            "**Missing Controls:** There is no explicit control group (e.g., a 3D network with the same number of neurons and coupling strength) to compare the 5D results against. Without this, the claim of 'significantly different dynamics' is unsubstantiated.",
+            "**Mismatched Statistics:** The provided metadata (run_count: 15) suggests a quantitative effort, but the absence of actual data prevents any statistical assessment, making any potential conclusion based on this metadata meaningless."
+          ],
+          "recommended_experiments": [
+            "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+            "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+            "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+          ],
+          "requested_evidence": [
+            "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+            "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+            "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "81ffd54d42225beb2ad80b0e622cb2a87e9169f6911c32e1e5c6713540333015",
+          "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T20:57:32.047321+00:00"
+      },
+      "writer": {
+        "analysis_id": "AIAR-scientific_writer-20260903205748858380-664b8e2f",
+        "role": "scientific_writer",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v2",
+          "packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f"
+        },
+        "output": {
+          "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+          "observations": [
+            "The core claim posits that 5D organization improves robustness against local structural losses.",
+            "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+            "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+            "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+            "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+          ],
+          "effect_direction": "N/A (Insufficient data)",
+          "methodological_concerns": [
+            "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+            "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+            "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+            "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support.",
+            "The setup fails to isolate the variable of 'dimensionality.' There is no explicit control group (e.g., a 3D network) to compare the 5D results against, making the claim of 'significantly different dynamics' unsubstantiated."
+          ],
+          "alternative_explanations": [
+            "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+            "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+            "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+          ],
+          "recommended_experiments": [
+            "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+            "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+            "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+          ],
+          "requested_evidence": [
+            "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+            "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+            "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "8bcc580b519bcb2a0f1772aa257c1624ffcf92675b88f7eec71e5d0edc5c4b95",
+          "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T20:57:48.858380+00:00"
+      }
+    }
+  },
+  "provenance": {
+    "research_packet_digest": "664b8e2fd13402750eed8f322681c32f9e49c923a1b70546c115e0b21fb7012f",
+    "source": {
+      "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+      "experiment_status": "completed",
+      "evidence_mode": "not_reported",
+      "git_dirty": "True",
+      "source_freeze_sha": "NOT_AVAILABLE",
+      "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+      "configuration_sha256": "NOT_AVAILABLE",
+      "experiment_manifest_digest": "7eb4a33d987c31ab71aaa06a97b9db2570366b4f410f6526a7c6f1c66d5f61d9",
+      "data_ids": "NOT_AVAILABLE",
+      "evid_ids": "NOT_AVAILABLE",
+      "protocol_id": "NOT_AVAILABLE",
+      "protocol_digest": "NOT_AVAILABLE",
+      "data_digest": "NOT_AVAILABLE",
+      "evid_digests": "[]"
+    }
+  },
+  "analysis_ids": [
+    "AIAR-scientific_analyst-20260903205715084095-664b8e2f",
+    "AIAR-critical_reviewer-20260903205732047321-664b8e2f",
+    "AIAR-scientific_writer-20260903205748858380-664b8e2f"
+  ],
+  "supersedes": null,
+  "content_digest": "66d6231017c73eb0b05f2807688a31ed93bb0ab9cbba977cb0677708c4f932f7",
+  "markdown_digest": "0cc4709e731639d2656c61434941f01e3790057d9d38452ce39c57a771b7fd20"
+}

research/experiments/EXP-5D-0001-v2/reports/AIRR-2026-0001.md

@@ -0,0 +1,251 @@
+============================================================
+BRAIN-5D - AI GENERATED SCIENTIFIC ANALYSIS
+============================================================
+
+Dieser Bericht wurde durch ein kuenstliches Intelligenzsystem erzeugt.
+
+Er stellt eine wissenschaftliche Interpretation von Brain-5D-Forschungsdaten dar.
+
+Er ist KEINE wissenschaftliche Evidenz.
+
+Er wurde zum Zeitpunkt der Erstellung noch nicht automatisch durch einen Menschen bestaetigt.
+
+Human Review Status: PENDING
+============================================================
+
+# AI Research Report AIRR-2026-0001
+
+Status: review_pending | Experiment: EXP-5D-0001-v2
+Research Question: RQ-5D-001
+## Executive Summary
+
+```json
+"The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion."
+```
+
+## Research Question
+
+```json
+{
+  "id": "RQ-5D-001",
+  "domain": "5D Topology",
+  "question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+  "relevance": "Kernfrage des gesamten Brain-5D-Projekts.",
+  "literature": [
+    "SRC-IZHIKEVICH-2003"
+  ],
+  "hypotheses": [
+    "H-5D-001-A"
+  ],
+  "evidence": [],
+  "status": "open",
+  "answer": {
+    "current": null,
+    "confidence": "none",
+    "limitations": null
+  },
+  "created": "2026-08-23",
+  "updated": "2026-08-23"
+}
+```
+
+## Experimental Design
+
+```json
+{}
+```
+
+## Data Basis
+
+```json
+{
+  "data": {},
+  "evid": [],
+  "literature": [
+    {
+      "source_id": "SRC-IZHIKEVICH-2003",
+      "authors": [
+        "Eugene M. Izhikevich"
+      ],
+      "title": "Simple model of spiking neurons",
+      "year": 2003,
+      "journal": "IEEE Transactions on Neural Networks",
+      "doi": "10.1109/TNN.2003.820440",
+      "topic": [
+        "neuron_model",
+        "spiking",
+        "snn_dynamics"
+      ],
+      "claims": [
+        "Simple spiking models can reproduce multiple cortical firing patterns."
+      ],
+      "brain5d_questions": [
+        "RQ-SNN-001",
+        "RQ-SNN-002",
+        "RQ-DET-001"
+      ],
+      "brain5d_relevance": "Referenz für Neuronendynamik und Parametrisierung der Izhikevich-Neuronen in Brain-5D."
+    }
+  ]
+}
+```
+
+## Quantitative Results
+
+```json
+{
+  "status": "NOT_AVAILABLE",
+  "source": "deterministic statistics.json"
+}
+```
+
+## Observations
+
+```json
+[
+  "The core claim posits that 5D organization improves robustness against local structural losses.",
+  "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+  "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+  "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+  "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+]
+```
+
+## Interpretation
+
+```json
+{
+  "assessment": "The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.",
+  "observations": [
+    "The core claim posits that 5D organization improves robustness against local structural losses.",
+    "The hypothesis (H-5D-001-A) suggests that a 5D-arranged network exhibits significantly different dynamics compared to 2D/3D networks of the same neuron count.",
+    "The experiment manifest (EXP-5D-0001-v2) is marked as 'completed' and ran 15 times, but no quantitative results, data files, or calculated metrics are present in the provided packet.",
+    "The `known_limitations` explicitly state: 'Git provenance is dirty or unavailable.' and 'No registered scientific evidence is linked to this experiment.'",
+    "The `protocol` field is null, and the `provenance` section lacks critical identifiers (e.g., `configuration_sha256`, `data_digest`, `protocol_id`), severely limiting reproducibility and validation."
+  ],
+  "effect_direction": "N/A (Insufficient data)",
+  "methodological_concerns": [
+    "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+    "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+    "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+    "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support.",
+    "The setup fails to isolate the variable of 'dimensionality.' There is no explicit control group (e.g., a 3D network) to compare the 5D results against, making the claim of 'significantly different dynamics' unsubstantiated."
+  ],
+  "alternative_explanations": [
+    "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+    "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+    "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+  ],
+  "recommended_experiments": [
+    "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+    "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+    "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+  ],
+  "requested_evidence": [
+    "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+    "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+    "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+  ],
+  "confidence": 0.1
+}
+```
+
+## Methodological Critique
+
+```json
+[
+  "The lack of quantitative results from the completed experiment prevents any analysis of the claimed effect.",
+  "The dirty Git provenance and missing configuration/data digests compromise the reproducibility of the experiment.",
+  "The absence of a detailed protocol description makes it impossible to verify if the experimental setup was appropriate for testing robustness or dynamics.",
+  "The current evidence base is purely theoretical (claim/hypothesis) and lacks empirical support."
+]
+```
+
+## Alternative Explanations
+
+```json
+[
+  "Observed differences in network dynamics might be due to parameter tuning or specific simulation constraints rather than the intrinsic dimensionality (5D vs 2D/3D).",
+  "The observed robustness might be an artifact of the specific damage model used, and the effect could be size-dependent rather than dimensionality-dependent.",
+  "The apparent difference in dynamics could be correlated with the specific connectivity rules implemented in the 5D structure, rather than the dimensionality itself."
+]
+```
+
+## Reproducibility
+
+```json
+{
+  "git_commit": "254b3574af41364619e297a5c399cb29249e9270",
+  "experiment_status": "completed",
+  "evidence_mode": "not_reported",
+  "git_dirty": "True",
+  "source_freeze_sha": "NOT_AVAILABLE",
+  "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+  "configuration_sha256": "NOT_AVAILABLE",
+  "experiment_manifest_digest": "7eb4a33d987c31ab71aaa06a97b9db2570366b4f410f6526a7c6f1c66d5f61d9",
+  "data_ids": "NOT_AVAILABLE",
+  "evid_ids": "NOT_AVAILABLE",
+  "protocol_id": "NOT_AVAILABLE",
+  "protocol_digest": "NOT_AVAILABLE",
+  "data_digest": "NOT_AVAILABLE",
+  "evid_digests": "[]"
+}
+```
+
+## Epistemic Status
+
+```json
+{
+  "implementation": "NOT_DETERMINED",
+  "experiment_data": "ABSENT",
+  "registered_evidence": "ABSENT",
+  "ai_interpretation": "PRESENT",
+  "human_review": "PENDING",
+  "claim_support": "NOT_DETERMINED",
+  "rq_status": "IN_PROGRESS"
+}
+```
+
+## Missing Evidence
+
+```json
+[
+  "The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.",
+  "A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.",
+  "Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations."
+]
+```
+
+## Recommended Follow-up Experiments
+
+```json
+[
+  "Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.",
+  "Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.",
+  "Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant."
+]
+```
+
+## Conclusion
+
+```json
+"The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion."
+```
+
+## Human Review
+
+```json
+{
+  "status": "PENDING",
+  "reviewer": null,
+  "reviewed_at": null,
+  "disposition": null,
+  "comments": null
+}
+```
+
+AI confidence: 0.1
+This value describes internal AI confidence and is not a statistical confidence interval.
+
+Content digest: 66d6231017c73eb0b05f2807688a31ed93bb0ab9cbba977cb0677708c4f932f7
+JSON/Markdown link digest: 0cc4709e731639d2656c61434941f01e3790057d9d38452ce39c57a771b7fd20

research/experiments/EXP-5D-0001-v2/reports/AIRR-2026-0001.review.json

@@ -0,0 +1,8 @@
+{
+  "report_id": "AIRR-2026-0001",
+  "review_status": "accepted_as_interpretation",
+  "reviewer": "Thomas Heisig",
+  "reviewed_at": "2026-09-08T09:13:53.679100+00:00",
+  "comments": "Restart in a later Run",
+  "report_content_digest": "66d6231017c73eb0b05f2807688a31ed93bb0ab9cbba977cb0677708c4f932f7"
+}

research/experiments/EXP-5D-0001-v2/summary.md

@@ -0,0 +1,58 @@
+# EXP-5D-0001-v2: Zusammenfassung
+
+Diese Zusammenfassung wurde nach Abschluss des Laufs durch den internen Research Assistant aus den Experimentartefakten und dem AIRR erstellt. Sie beschreibt die Daten, die Berichte und deren wissenschaftliche Grenzen.
+
+## Versuchsuebersicht
+
+- Status: `completed`
+- Forschungsfragen: RQ-5D-001
+- Hypothesen: H-5D-001-A
+- Durchlaeufe: `15`
+- Laufmodus: `EXPLORATORY`
+- Netzwerkmodus: `OFFLINE`
+
+## Artefakte
+
+- [analysis/AIAR-critical_reviewer-20260903205642547131-d11507fa.json](analysis/AIAR-critical_reviewer-20260903205642547131-d11507fa.json)
+- [analysis/AIAR-critical_reviewer-20260903205732047321-664b8e2f.json](analysis/AIAR-critical_reviewer-20260903205732047321-664b8e2f.json)
+- [analysis/AIAR-scientific_analyst-20260903205623921935-d11507fa.json](analysis/AIAR-scientific_analyst-20260903205623921935-d11507fa.json)
+- [analysis/AIAR-scientific_analyst-20260903205715084095-664b8e2f.json](analysis/AIAR-scientific_analyst-20260903205715084095-664b8e2f.json)
+- [analysis/AIAR-scientific_writer-20260903205748858380-664b8e2f.json](analysis/AIAR-scientific_writer-20260903205748858380-664b8e2f.json)
+- [DATA/runs.json](DATA/runs.json)
+- [manifest.json](manifest.json)
+- [report.md](report.md)
+- [reports/AIRR-2026-0001.json](reports/AIRR-2026-0001.json)
+- [reports/AIRR-2026-0001.md](reports/AIRR-2026-0001.md)
+- [workflow.json](workflow.json)
+
+## AI-Bericht
+
+- AIRR-Status: `generated`
+- AIRR: [AIRR-2026-0001.md](reports/AIRR-2026-0001.md)
+- AIRR JSON: [AIRR-2026-0001.json](reports/AIRR-2026-0001.json)
+- Wissenschaftliche Evidenz: `false`
+- Human Review: `PENDING`
+
+### KI-Einschaetzung
+
+Die KI bewertet den vorliegenden Datensatz wie folgt:
+
+The research packet presents a clear, novel hypothesis (H-5D-001-A) and claim (CLAIM-5D-001) regarding the enhanced robustness and unique dynamics of 5D network organization. However, the associated experiment (EXP-5D-0001-v2) is marked as 'completed' but provides no quantitative results, data files, or calculated metrics. The `results` section only contains metadata (run_count: 15, runtime: 0.0038415000308305025). Consequently, a substantive scientific assessment of the claimed effect is impossible, rendering the current evidence base purely theoretical and unverified. The methodological concerns regarding missing data, dirty provenance, and lack of controls are critical and prevent any conclusion.
+
+KI-Konfidenz: `0.1`
+
+Angeforderte zusaetzliche Nachweise:
+
+- The raw data files (e.g., time series, connectivity matrices) from the 15 runs of EXP-5D-0001-v2.
+- A detailed, reproducible protocol description for EXP-5D-0001-v2, including all parameters and initialization steps.
+- Statistical evidence (e.g., p-values, confidence intervals) comparing the network metrics (e.g., synchronization, activity level) between 5D and 3D/2D configurations.
+
+Empfohlene Folgeexperimente:
+
+- Conduct a controlled ablation study comparing the network dynamics and robustness metrics (e.g., critical transition points, recovery time) of 5D, 3D, and 2D networks under identical damage scenarios.
+- Systematically vary the scale and type of local structural damage (e.g., random node removal, targeted link removal) to determine if the robustness benefit is scale-invariant.
+- Implement a comparative analysis where the only variable is the dimensionality, keeping all other parameters (neuron count, coupling strength, damage rate) constant.
+
+## Wissenschaftliche Grenze
+
+Die KI-Auswertung ist post-hoc, steuert den Lauf nicht und ersetzt keine menschliche wissenschaftliche Pruefung oder Evidenzfreigabe.

research/experiments/EXP-5D-0001-v2/workflow.json

@@ -0,0 +1,17 @@
+{
+  "protocol": "science_suite_v1",
+  "experiment_id": "EXP-5D-0001-v2",
+  "research_question": "RQ-5D-001",
+  "hypothesis": "H-5D-001-A",
+  "title": "Dimensionalitaet und Netzwerkdynamik v2",
+  "conditions": "Kontrollierte identische Ketten in 1D, 2D, 3D und 5D sowie Random-Graph-Kontrolle; Seeds 42, 43, 44",
+  "ticks": 8,
+  "seeds": [
+    42,
+    43,
+    44
+  ],
+  "notes": "Post-hoc AIRR. KI beeinflusst weder Ausfuehrung noch Evidenzbewertung.",
+  "execution": "registered experiment_suite runner",
+  "assistant_policy": "AI is post-hoc interpretation only."
+}

research/experiments/EXP-5D-0001-v3/DATA/runs.json

@@ -0,0 +1,407 @@
+[
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "949fe94f48feb71fb01cfd6efdebe3b81a6684ddff17850e9e78f186afa7b922",
+    "state_digest_before": "09ca7d6cfa1e4d58f797594dcaa3ce74f0f9afdda73e9939e3441073f32f15d2"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "433b7309d7aadb895a00d257898978e4897697033899651074438467d9da97c9",
+    "state_digest_before": "a9995feead5bb4b790a597ac89047fcd7803d3aafb48c670b623584ac3adb265"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "d408828d8ab3b3e784f6359e01d610411c55071ce56cfaed9c5db53e1d08ac91",
+    "state_digest_before": "373ba6d808659a096a0aa7ec0de9d8f2828aabe28ce4574ab91fe43909f3671c"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "1cf8befa415159dfd891a5c267b2550c03e447232c241a4746956f5febd557e3",
+    "state_digest_before": "f4d3e6c7573308a9a88cf760a8f1970d169529a5ed50027c438ef5a7ab95c8bc"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 42,
+    "state_digest_after": "4283987883d50084193eb6537cf5f2b80d33c34fd8f76807cd84c688c427f53b",
+    "state_digest_before": "fc441796f8ea40657ceb355f08b4d8d6c82bc29cbba6758be52c0fa495a099ec"
+  },
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "1f634744745b1e90298998bc891fd79d2cbf0268cbda266b2a0b7142d07c9f03",
+    "state_digest_before": "3560c51f27691686e4f89de79aa9eaa959f109ed3b1236f88a1d0b9e4538e1f4"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "5a124bce6c7fefc7838df1dfb93bf1e96031f9874bddf102c7e7774545628bb7",
+    "state_digest_before": "2e06a7dccab315af6985c10d3deab5e2e844b8340f29d944bd7b6af1627dee9c"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "aea563b907b76c8c99fb43fd5387dddb8af33abf070498eb7ce4476fcdd7923b",
+    "state_digest_before": "ef800355245c41f7fb590d1b0b0e1abd80de25dc51003907c9e0c8b6005bb42a"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "975b6b4813b95f9592b92d3e8da4e32c956b56dfd438451a9d207a68b6d0a6cc",
+    "state_digest_before": "0bbea3bbca3255ff87392888a96cd073780de6588a8295fd73ed85913d451f4b"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 43,
+    "state_digest_after": "28a4145e2b4268ade9a11f940fb25c1767cf30ebe802f64536bfad236750f0aa",
+    "state_digest_before": "1721066aa7d25c1c92e8c776cea0d3be1f63a7c2853af3be4241e43b71f8bad1"
+  },
+  {
+    "condition": "1d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        3,
+        1,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "f5fe1b6b60126ae3d0c4279a22d9125f649589602e6c718bb0a84a55aa918a6f",
+    "state_digest_before": "121b3d5b2a46319293cc04de0e84cd22720b5c8815996ea04cb318ebe6fa9d84"
+  },
+  {
+    "condition": "2d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        1,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "fb2b961c05b278a388755d8f6d7811854dba155cd3210ee126294a2a7ca67f44",
+    "state_digest_before": "59b95fecc0e0e33cd4369e0fff9faea0a8c83f6820d20fcd1b793f162899af07"
+  },
+  {
+    "condition": "3d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        1,
+        1
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "b01773df90dddff1c7f018520877c433e5105084c766615d6479d18c00a9a626",
+    "state_digest_before": "e28894b2e956785edb988d8a3580251ca079465a29b4cebb59c863e1182a762d"
+  },
+  {
+    "condition": "5d",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "d19ac0fdafa07f6b1dcef2390ce84b759fe7576993a59d9d8e4109b25656cb9e",
+    "state_digest_before": "01d5c16d95e6464d001a4edbf2862eae0f7539348d112dd223fd6498f66be49e"
+  },
+  {
+    "condition": "random_graph",
+    "experiment_id": "EXP-5D-0001-v3",
+    "metrics": {
+      "activated_neurons": 0,
+      "dimensions": [
+        2,
+        2,
+        2,
+        2,
+        2
+      ],
+      "first_response_latency": null,
+      "last_response_latency": null,
+      "network_state_digest_after": null,
+      "network_state_digest_before": null,
+      "peak_spike_rate": 0.0,
+      "propagation_depth": 0,
+      "recurrent_events": 0,
+      "return_latency": null,
+      "total_spikes": 0
+    },
+    "runtime_error": null,
+    "seed": 44,
+    "state_digest_after": "61fcd71474cb4ce42e0bf0cd80548b2e498d5c89c36b357c7f7495c04a335ea4",
+    "state_digest_before": "225ab55ea78bc8bb128448d38fc5ba1c32ac809531b22d27c55d6e9e6d4db4f7"
+  }
+]

research/experiments/EXP-5D-0001-v3/analysis/AIAR-critical_reviewer-20260903211303830974-fa63da1e.json

@@ -0,0 +1,75 @@
+{
+  "analysis_id": "AIAR-critical_reviewer-20260903211303830974-fa63da1e",
+  "role": "critical_reviewer",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v3",
+    "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+  },
+  "output": {
+    "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+    "observations": [
+      "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+      "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+      "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+      "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+      "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+      "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+      "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+    ],
+    "effect_direction": "N/A (Untested)",
+    "methodological_concerns": [
+      "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+      "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+      "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+      "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+      "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+    ],
+    "alternative_explanations": [
+      "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+      "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+      "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+    ],
+    "recommended_experiments": [
+      "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+      "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+      "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+    ],
+    "requested_evidence": [
+      "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+      "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+      "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "487773c90dea6bd53dbee2d1e707e439138a61c4544411ac318d1cd4c7d81914",
+    "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T21:13:03.830974+00:00"
+}

research/experiments/EXP-5D-0001-v3/analysis/AIAR-scientific_analyst-20260903211248541726-fa63da1e.json

@@ -0,0 +1,73 @@
+{
+  "analysis_id": "AIAR-scientific_analyst-20260903211248541726-fa63da1e",
+  "role": "scientific_analyst",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v3",
+    "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+  },
+  "output": {
+    "assessment": "The current evidence base is insufficient to evaluate the claim that five-dimensional organization improves robustness against local structural losses. The claim (CLAIM-5D-001) is currently untested, and the associated hypothesis (H-5D-001-A) and research question (RQ-5D-001) require dedicated experimental validation. The existing manifest indicates a completed experiment (EXP-5D-0001-v3) with limited run counts (15 runs) and no reported results or data analysis, making any current interpretation speculative.",
+    "observations": [
+      "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested'.",
+      "**Hypothesis Status:** The hypothesis H-5D-001-A ('Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl') is 'untested'.",
+      "**Experimental Data:** The experiment EXP-5D-0001-v3 was run, but the manifest shows no reported results, data, or analysis outputs (e.g., `results: {}`, `data: null`, `evidence: []`).",
+      "**Known Limitations:** The system explicitly notes 'No registered scientific evidence is linked to this experiment.' and 'Git provenance is dirty or unavailable.'",
+      "**Literature Context:** The only literature source (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address 5D network topology or robustness.",
+      "**Research Question:** RQ-5D-001 is open, requiring measurable effects on network dynamics to be determined."
+    ],
+    "effect_direction": "N/A (Untested)",
+    "methodological_concerns": [
+      "The current experimental setup lacks sufficient statistical power and replication (minimum runs required: 30, actual runs: 15).",
+      "The lack of reported results or data analysis from EXP-5D-0001-v3 prevents any meaningful interpretation of the experiment's outcome.",
+      "The 'dirty' git status and missing provenance data (configuration SHA256, data digests) compromise the reproducibility and reliability of the experiment.",
+      "The comparison group (2D/3D network) and the specific mechanism of 'local structural loss' (damage type, scale, distribution) must be rigorously defined and controlled."
+    ],
+    "alternative_explanations": [
+      "Observed differences in network dynamics could be due to parameter tuning issues (e.g., learning rate, spiking threshold) rather than the dimensionality itself.",
+      "The observed effects might be artifacts of the simulation environment or the specific random seed used, rather than a fundamental property of 5D topology.",
+      "The perceived robustness might be correlated with the total number of neurons or connections, rather than the dimensionality of the embedding space."
+    ],
+    "recommended_experiments": [
+      "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled damage conditions (e.g., random node removal, targeted structural loss).",
+      "Implement a controlled damage experiment where the percentage and pattern of structural loss are systematically varied across dimensions (2D, 3D, 5D) to quantify the robustness metric.",
+      "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using multiple seeds and sufficient run counts (minimum 30) to establish statistical significance."
+    ],
+    "requested_evidence": [
+      "Quantitative metrics comparing network stability/recovery time/information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+      "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics.",
+      "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters and seeds."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "08cef4ab9d9951cefea691c2a97849d324581b6be3283e9d9c458ae3b15a90fb",
+    "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T21:12:48.541726+00:00"
+}

research/experiments/EXP-5D-0001-v3/analysis/AIAR-scientific_writer-20260903211319887624-fa63da1e.json

@@ -0,0 +1,75 @@
+{
+  "analysis_id": "AIAR-scientific_writer-20260903211319887624-fa63da1e",
+  "role": "scientific_writer",
+  "model": {
+    "provider": "ollama",
+    "model_name": "gemma4:latest",
+    "model_digest": "not_reported",
+    "quantization": "not_reported",
+    "context_length": "unknown",
+    "temperature": 0.0,
+    "top_p": 0.9,
+    "seed": "not_reported",
+    "backend_version": "unknown"
+  },
+  "inputs": {
+    "experiment_id": "EXP-5D-0001-v3",
+    "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+  },
+  "output": {
+    "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+    "observations": [
+      "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+      "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+      "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+      "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+      "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+      "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+      "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+    ],
+    "effect_direction": "N/A (Untested)",
+    "methodological_concerns": [
+      "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+      "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+      "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+      "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+      "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+    ],
+    "alternative_explanations": [
+      "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+      "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+      "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+    ],
+    "recommended_experiments": [
+      "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+      "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+      "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+    ],
+    "requested_evidence": [
+      "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+      "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+      "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+    ],
+    "confidence": 0.1
+  },
+  "provenance": {
+    "prompt_sha256": "93f3b45edc71f136f55f973d4086bd4111687a9d43a1ae67d66d793b1e2ceb81",
+    "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+    "prompt_protocol_version": "research_assistant_v1",
+    "assistant_schema_version": "1",
+    "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+    "model_self_confidence": "0.1"
+  },
+  "epistemic_status": {
+    "evidence": false,
+    "interpretation_only": true,
+    "human_review_required": true
+  },
+  "review": {
+    "status": "pending",
+    "reviewer": null,
+    "reviewed_at": null,
+    "disposition": null
+  },
+  "generated_at": "2026-09-03T21:13:19.887624+00:00"
+}

research/experiments/EXP-5D-0001-v3/manifest.json

@@ -0,0 +1,71 @@
+{
+  "ai_exposure": "none",
+  "ai_interactions": [],
+  "ai_protocol": {
+    "bump_reason": "initial registration",
+    "version": 1
+  },
+  "ai_reproducibility": "R0",
+  "artifacts": {
+    "ai_report_json": "experiments/EXP-5D-0001-v3/reports/AIRR-2026-0001.json",
+    "ai_report_markdown": "experiments/EXP-5D-0001-v3/reports/AIRR-2026-0001.md",
+    "data": "DATA/runs.json",
+    "report": "report.md",
+    "workflow": "workflow.json"
+  },
+  "causal_card": {
+    "classification": "PURE",
+    "interaction_ids": [],
+    "roles": []
+  },
+  "causal_taint": "PURE",
+  "config": {
+    "path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+    "sha256": ""
+  },
+  "data_partitions": [],
+  "experiment_id": "EXP-5D-0001-v3",
+  "experiment_status": "completed",
+  "git": {
+    "branch": "main",
+    "commit": "f0aca72400a08c247dbc535969332677fec305ad",
+    "dirty": true
+  },
+  "hypotheses": [
+    "H-5D-001-A"
+  ],
+  "network_mode": "OFFLINE",
+  "research_questions": [
+    "RQ-5D-001"
+  ],
+  "research_run_mode": "EXPLORATORY",
+  "results": {
+    "protocol": "science_suite_v1",
+    "run_count": 15
+  },
+  "runtime": {
+    "duration_seconds": 0.003703500027768314
+  },
+  "runtime_errors": [],
+  "simulation": {
+    "seed": 42,
+    "seeds": [
+      42,
+      43,
+      44
+    ],
+    "ticks": 8
+  },
+  "software": {
+    "brain5d_version": "unknown",
+    "os": "Windows 11",
+    "python": "3.13.14"
+  },
+  "timestamp": "2026-09-03T23:12:35.361754",
+  "validity": {
+    "fatal_error_count": 0,
+    "reason": null,
+    "runtime_error_count": 0,
+    "valid": true
+  }
+}

research/experiments/EXP-5D-0001-v3/report.md

@@ -0,0 +1,12 @@
+# EXP-5D-0001-v3: Dimensionalitaet und Signalpropagation v3
+
+## Protokoll
+science_suite_v1
+Runs: 15; Dauer: 0.003704 s
+
+## Evidenzstatus
+DATA und Manifest erzeugt. EVID wird erst nach Clean Freeze und Review erzeugt.
+
+## Hinweise
+Prediction Error, Memory, TIME, 5D, Regulation und Sensorverlust sind in diesem Lauf nicht behauptet.
+Post-hoc AIRR und Summary. KI beeinflusst weder Ausfuehrung noch Evidenz.

research/experiments/EXP-5D-0001-v3/reports/AIRR-2026-0001.json

@@ -0,0 +1,447 @@
+{
+  "report_id": "AIRR-2026-0001",
+  "experiment_id": "EXP-5D-0001-v3",
+  "research_question_id": "RQ-5D-001",
+  "hypothesis_ids": [
+    "H-5D-001-A"
+  ],
+  "claim_ids": [
+    "CLAIM-5D-001"
+  ],
+  "generated_at": "2026-09-03T21:13:19.888791+00:00",
+  "report_schema_version": "1.0",
+  "report_protocol_version": "PROTOCOL-AIRR-001",
+  "generated_by": "artificial_intelligence",
+  "ai_generated": true,
+  "scientific_evidence": false,
+  "human_review_required": true,
+  "status": "review_pending",
+  "content": {
+    "identification": {
+      "experiment": "EXP-5D-0001-v3",
+      "research_question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+      "hypotheses": [
+        {
+          "id": "H-5D-001-A",
+          "research_question": "RQ-5D-001",
+          "hypothesis": "Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl.",
+          "status": "untested",
+          "evidence": [],
+          "created": "2026-08-23",
+          "updated": "2026-08-23"
+        }
+      ],
+      "date": "2026-09-03",
+      "source_sha": "f0aca72400a08c247dbc535969332677fec305ad"
+    },
+    "executive_summary": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+    "research_question": {
+      "id": "RQ-5D-001",
+      "domain": "5D Topology",
+      "question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+      "relevance": "Kernfrage des gesamten Brain-5D-Projekts.",
+      "literature": [
+        "SRC-IZHIKEVICH-2003"
+      ],
+      "hypotheses": [
+        "H-5D-001-A"
+      ],
+      "evidence": [],
+      "status": "open",
+      "answer": {
+        "current": null,
+        "confidence": "none",
+        "limitations": null
+      },
+      "created": "2026-08-23",
+      "updated": "2026-08-23"
+    },
+    "experimental_design": {},
+    "data_basis": {
+      "data": {},
+      "evid": [],
+      "literature": [
+        {
+          "source_id": "SRC-IZHIKEVICH-2003",
+          "authors": [
+            "Eugene M. Izhikevich"
+          ],
+          "title": "Simple model of spiking neurons",
+          "year": 2003,
+          "journal": "IEEE Transactions on Neural Networks",
+          "doi": "10.1109/TNN.2003.820440",
+          "topic": [
+            "neuron_model",
+            "spiking",
+            "snn_dynamics"
+          ],
+          "claims": [
+            "Simple spiking models can reproduce multiple cortical firing patterns."
+          ],
+          "brain5d_questions": [
+            "RQ-SNN-001",
+            "RQ-SNN-002",
+            "RQ-DET-001"
+          ],
+          "brain5d_relevance": "Referenz für Neuronendynamik und Parametrisierung der Izhikevich-Neuronen in Brain-5D."
+        }
+      ]
+    },
+    "quantitative_results": {
+      "status": "NOT_AVAILABLE",
+      "source": "deterministic statistics.json"
+    },
+    "observations": [
+      "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested'.",
+      "**Hypothesis Status:** The hypothesis H-5D-001-A ('Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl') is 'untested'.",
+      "**Experimental Data:** The experiment EXP-5D-0001-v3 was run, but the manifest shows no reported results, data, or analysis outputs (e.g., `results: {}`, `data: null`, `evidence: []`).",
+      "**Known Limitations:** The system explicitly notes 'No registered scientific evidence is linked to this experiment.' and 'Git provenance is dirty or unavailable.'",
+      "**Literature Context:** The only literature source (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address 5D network topology or robustness.",
+      "**Research Question:** RQ-5D-001 is open, requiring measurable effects on network dynamics to be determined."
+    ],
+    "interpretation": {
+      "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+      "observations": [
+        "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+        "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+        "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+        "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+        "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+        "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+        "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+      ],
+      "effect_direction": "N/A (Untested)",
+      "methodological_concerns": [
+        "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+        "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+        "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+        "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+        "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+      ],
+      "alternative_explanations": [
+        "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+        "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+        "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+      ],
+      "recommended_experiments": [
+        "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+        "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+        "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+      ],
+      "requested_evidence": [
+        "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+        "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+        "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+      ],
+      "confidence": 0.1
+    },
+    "methodological_critique": [
+      "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+      "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+      "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+      "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+      "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+    ],
+    "alternative_explanations": [
+      "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+      "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+      "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+    ],
+    "reproducibility": {
+      "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+      "experiment_status": "completed",
+      "evidence_mode": "not_reported",
+      "git_dirty": "True",
+      "source_freeze_sha": "NOT_AVAILABLE",
+      "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+      "configuration_sha256": "NOT_AVAILABLE",
+      "experiment_manifest_digest": "b9ab6fd5aab5b4641d71ff28071f13b4b466a36798e3291d59206f2a2fdfbb3b",
+      "data_ids": "NOT_AVAILABLE",
+      "evid_ids": "NOT_AVAILABLE",
+      "protocol_id": "NOT_AVAILABLE",
+      "protocol_digest": "NOT_AVAILABLE",
+      "data_digest": "NOT_AVAILABLE",
+      "evid_digests": "[]"
+    },
+    "epistemic_status": {
+      "implementation": "NOT_DETERMINED",
+      "experiment_data": "ABSENT",
+      "registered_evidence": "ABSENT",
+      "ai_interpretation": "PRESENT",
+      "human_review": "PENDING",
+      "claim_support": "NOT_DETERMINED",
+      "rq_status": "IN_PROGRESS"
+    },
+    "ai_confidence": 0.1,
+    "missing_evidence": [
+      "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+      "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+      "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+    ],
+    "recommended_follow_up": [
+      "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+      "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+      "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+    ],
+    "conclusion": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+    "human_review": {
+      "status": "PENDING",
+      "reviewer": null,
+      "reviewed_at": null,
+      "disposition": null,
+      "comments": null
+    },
+    "aiar": {
+      "analyst": {
+        "analysis_id": "AIAR-scientific_analyst-20260903211248541726-fa63da1e",
+        "role": "scientific_analyst",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v3",
+          "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+        },
+        "output": {
+          "assessment": "The current evidence base is insufficient to evaluate the claim that five-dimensional organization improves robustness against local structural losses. The claim (CLAIM-5D-001) is currently untested, and the associated hypothesis (H-5D-001-A) and research question (RQ-5D-001) require dedicated experimental validation. The existing manifest indicates a completed experiment (EXP-5D-0001-v3) with limited run counts (15 runs) and no reported results or data analysis, making any current interpretation speculative.",
+          "observations": [
+            "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested'.",
+            "**Hypothesis Status:** The hypothesis H-5D-001-A ('Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl') is 'untested'.",
+            "**Experimental Data:** The experiment EXP-5D-0001-v3 was run, but the manifest shows no reported results, data, or analysis outputs (e.g., `results: {}`, `data: null`, `evidence: []`).",
+            "**Known Limitations:** The system explicitly notes 'No registered scientific evidence is linked to this experiment.' and 'Git provenance is dirty or unavailable.'",
+            "**Literature Context:** The only literature source (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address 5D network topology or robustness.",
+            "**Research Question:** RQ-5D-001 is open, requiring measurable effects on network dynamics to be determined."
+          ],
+          "effect_direction": "N/A (Untested)",
+          "methodological_concerns": [
+            "The current experimental setup lacks sufficient statistical power and replication (minimum runs required: 30, actual runs: 15).",
+            "The lack of reported results or data analysis from EXP-5D-0001-v3 prevents any meaningful interpretation of the experiment's outcome.",
+            "The 'dirty' git status and missing provenance data (configuration SHA256, data digests) compromise the reproducibility and reliability of the experiment.",
+            "The comparison group (2D/3D network) and the specific mechanism of 'local structural loss' (damage type, scale, distribution) must be rigorously defined and controlled."
+          ],
+          "alternative_explanations": [
+            "Observed differences in network dynamics could be due to parameter tuning issues (e.g., learning rate, spiking threshold) rather than the dimensionality itself.",
+            "The observed effects might be artifacts of the simulation environment or the specific random seed used, rather than a fundamental property of 5D topology.",
+            "The perceived robustness might be correlated with the total number of neurons or connections, rather than the dimensionality of the embedding space."
+          ],
+          "recommended_experiments": [
+            "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled damage conditions (e.g., random node removal, targeted structural loss).",
+            "Implement a controlled damage experiment where the percentage and pattern of structural loss are systematically varied across dimensions (2D, 3D, 5D) to quantify the robustness metric.",
+            "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using multiple seeds and sufficient run counts (minimum 30) to establish statistical significance."
+          ],
+          "requested_evidence": [
+            "Quantitative metrics comparing network stability/recovery time/information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+            "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics.",
+            "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters and seeds."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "08cef4ab9d9951cefea691c2a97849d324581b6be3283e9d9c458ae3b15a90fb",
+          "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T21:12:48.541726+00:00"
+      },
+      "reviewer": {
+        "analysis_id": "AIAR-critical_reviewer-20260903211303830974-fa63da1e",
+        "role": "critical_reviewer",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v3",
+          "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+        },
+        "output": {
+          "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+          "observations": [
+            "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+            "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+            "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+            "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+            "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+            "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+            "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+          ],
+          "effect_direction": "N/A (Untested)",
+          "methodological_concerns": [
+            "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+            "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+            "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+            "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+            "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+          ],
+          "alternative_explanations": [
+            "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+            "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+            "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+          ],
+          "recommended_experiments": [
+            "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+            "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+            "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+          ],
+          "requested_evidence": [
+            "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+            "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+            "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "487773c90dea6bd53dbee2d1e707e439138a61c4544411ac318d1cd4c7d81914",
+          "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T21:13:03.830974+00:00"
+      },
+      "writer": {
+        "analysis_id": "AIAR-scientific_writer-20260903211319887624-fa63da1e",
+        "role": "scientific_writer",
+        "model": {
+          "provider": "ollama",
+          "model_name": "gemma4:latest",
+          "model_digest": "not_reported",
+          "quantization": "not_reported",
+          "context_length": "unknown",
+          "temperature": 0.0,
+          "top_p": 0.9,
+          "seed": "not_reported",
+          "backend_version": "unknown"
+        },
+        "inputs": {
+          "experiment_id": "EXP-5D-0001-v3",
+          "packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c"
+        },
+        "output": {
+          "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+          "observations": [
+            "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+            "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+            "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+            "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+            "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+            "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+            "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+          ],
+          "effect_direction": "N/A (Untested)",
+          "methodological_concerns": [
+            "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+            "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+            "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+            "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+            "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+          ],
+          "alternative_explanations": [
+            "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+            "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+            "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+          ],
+          "recommended_experiments": [
+            "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+            "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+            "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+          ],
+          "requested_evidence": [
+            "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+            "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+            "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+          ],
+          "confidence": 0.1
+        },
+        "provenance": {
+          "prompt_sha256": "93f3b45edc71f136f55f973d4086bd4111687a9d43a1ae67d66d793b1e2ceb81",
+          "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+          "prompt_protocol_version": "research_assistant_v1",
+          "assistant_schema_version": "1",
+          "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+          "model_self_confidence": "0.1"
+        },
+        "epistemic_status": {
+          "evidence": false,
+          "interpretation_only": true,
+          "human_review_required": true
+        },
+        "review": {
+          "status": "pending",
+          "reviewer": null,
+          "reviewed_at": null,
+          "disposition": null
+        },
+        "generated_at": "2026-09-03T21:13:19.887624+00:00"
+      }
+    }
+  },
+  "provenance": {
+    "research_packet_digest": "fa63da1e742646995ecfe85598c247258c167da3e504b96d38336f929124d87c",
+    "source": {
+      "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+      "experiment_status": "completed",
+      "evidence_mode": "not_reported",
+      "git_dirty": "True",
+      "source_freeze_sha": "NOT_AVAILABLE",
+      "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+      "configuration_sha256": "NOT_AVAILABLE",
+      "experiment_manifest_digest": "b9ab6fd5aab5b4641d71ff28071f13b4b466a36798e3291d59206f2a2fdfbb3b",
+      "data_ids": "NOT_AVAILABLE",
+      "evid_ids": "NOT_AVAILABLE",
+      "protocol_id": "NOT_AVAILABLE",
+      "protocol_digest": "NOT_AVAILABLE",
+      "data_digest": "NOT_AVAILABLE",
+      "evid_digests": "[]"
+    }
+  },
+  "analysis_ids": [
+    "AIAR-scientific_analyst-20260903211248541726-fa63da1e",
+    "AIAR-critical_reviewer-20260903211303830974-fa63da1e",
+    "AIAR-scientific_writer-20260903211319887624-fa63da1e"
+  ],
+  "supersedes": null,
+  "content_digest": "12979faa8a4bc5c9f6c6440d144efecd1646b269b88c4319dae784e241d35dfb",
+  "markdown_digest": "4ffbb365b5cf90afeb157e61d9712b4228d86c18ad08d5e365da66f086871c48"
+}

research/experiments/EXP-5D-0001-v3/reports/AIRR-2026-0001.md

@@ -0,0 +1,255 @@
+============================================================
+BRAIN-5D - AI GENERATED SCIENTIFIC ANALYSIS
+============================================================
+
+Dieser Bericht wurde durch ein kuenstliches Intelligenzsystem erzeugt.
+
+Er stellt eine wissenschaftliche Interpretation von Brain-5D-Forschungsdaten dar.
+
+Er ist KEINE wissenschaftliche Evidenz.
+
+Er wurde zum Zeitpunkt der Erstellung noch nicht automatisch durch einen Menschen bestaetigt.
+
+Human Review Status: PENDING
+============================================================
+
+# AI Research Report AIRR-2026-0001
+
+Status: review_pending | Experiment: EXP-5D-0001-v3
+Research Question: RQ-5D-001
+## Executive Summary
+
+```json
+"The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn."
+```
+
+## Research Question
+
+```json
+{
+  "id": "RQ-5D-001",
+  "domain": "5D Topology",
+  "question": "Hat die fünfdimensionale Anordnung einen messbaren Effekt auf die Netzwerkdynamik?",
+  "relevance": "Kernfrage des gesamten Brain-5D-Projekts.",
+  "literature": [
+    "SRC-IZHIKEVICH-2003"
+  ],
+  "hypotheses": [
+    "H-5D-001-A"
+  ],
+  "evidence": [],
+  "status": "open",
+  "answer": {
+    "current": null,
+    "confidence": "none",
+    "limitations": null
+  },
+  "created": "2026-08-23",
+  "updated": "2026-08-23"
+}
+```
+
+## Experimental Design
+
+```json
+{}
+```
+
+## Data Basis
+
+```json
+{
+  "data": {},
+  "evid": [],
+  "literature": [
+    {
+      "source_id": "SRC-IZHIKEVICH-2003",
+      "authors": [
+        "Eugene M. Izhikevich"
+      ],
+      "title": "Simple model of spiking neurons",
+      "year": 2003,
+      "journal": "IEEE Transactions on Neural Networks",
+      "doi": "10.1109/TNN.2003.820440",
+      "topic": [
+        "neuron_model",
+        "spiking",
+        "snn_dynamics"
+      ],
+      "claims": [
+        "Simple spiking models can reproduce multiple cortical firing patterns."
+      ],
+      "brain5d_questions": [
+        "RQ-SNN-001",
+        "RQ-SNN-002",
+        "RQ-DET-001"
+      ],
+      "brain5d_relevance": "Referenz für Neuronendynamik und Parametrisierung der Izhikevich-Neuronen in Brain-5D."
+    }
+  ]
+}
+```
+
+## Quantitative Results
+
+```json
+{
+  "status": "NOT_AVAILABLE",
+  "source": "deterministic statistics.json"
+}
+```
+
+## Observations
+
+```json
+[
+  "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested'.",
+  "**Hypothesis Status:** The hypothesis H-5D-001-A ('Ein 5D-angeordnetes Netzwerk zeigt signifikant andere Dynamik als ein 2D/3D-Netzwerk gleicher Neuronenzahl') is 'untested'.",
+  "**Experimental Data:** The experiment EXP-5D-0001-v3 was run, but the manifest shows no reported results, data, or analysis outputs (e.g., `results: {}`, `data: null`, `evidence: []`).",
+  "**Known Limitations:** The system explicitly notes 'No registered scientific evidence is linked to this experiment.' and 'Git provenance is dirty or unavailable.'",
+  "**Literature Context:** The only literature source (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address 5D network topology or robustness.",
+  "**Research Question:** RQ-5D-001 is open, requiring measurable effects on network dynamics to be determined."
+]
+```
+
+## Interpretation
+
+```json
+{
+  "assessment": "The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.",
+  "observations": [
+    "**Claim Status:** The claim 'Die fünfdimensionale Organisation verbessert die Robustheit gegenüber lokalen Strukturverlusten' (CLAIM-5D-001) is marked as 'untested' and lacks supporting evidence.",
+    "**Hypothesis Status:** The hypothesis H-5D-001-A is 'untested' due to the absence of analyzed data.",
+    "**Experimental Data:** The manifest confirms the experiment EXP-5D-0001-v3 was run, but the provided data structure is empty (`data: null`, `evidence: []`, `results: {}`), indicating no reported results or analysis outputs.",
+    "**Methodological Flaw (Replication):** The experiment only ran 15 times (`run_count: 15`), which is statistically inadequate given the required minimum of 30 runs for robust claims.",
+    "**Methodological Flaw (Provenance):** The experiment suffers from critical provenance issues: the git status is dirty (`git_dirty: True`), and crucial configuration and data digests are missing (`configuration_sha256: \"NOT_AVAILABLE\"`, `data_digest: \"NOT_AVAILABLE\"`).",
+    "**Interpretation Flaw:** The system has not performed any data analysis or interpretation, making any potential conclusion purely speculative.",
+    "**Literature Context:** The only cited literature (SRC-IZHIKEVICH-2003) provides general context on spiking neuron models but does not address the specific topological or robustness claims of 5D networks."
+  ],
+  "effect_direction": "N/A (Untested)",
+  "methodological_concerns": [
+    "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+    "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+    "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+    "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+    "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+  ],
+  "alternative_explanations": [
+    "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+    "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+    "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+  ],
+  "recommended_experiments": [
+    "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+    "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+    "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+  ],
+  "requested_evidence": [
+    "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+    "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+    "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+  ],
+  "confidence": 0.1
+}
+```
+
+## Methodological Critique
+
+```json
+[
+  "**Insufficient Statistical Power:** The run count of 15 is below the required minimum of 30, severely limiting the ability to establish statistical significance for any observed effect.",
+  "**Lack of Comparative Control:** The current setup does not explicitly detail the control groups (e.g., 2D and 3D networks) or the systematic variation of the damage mechanism (type, scale, distribution) necessary for a robust ablation study.",
+  "**Missing Data Analysis:** The absence of reported results or analysis outputs means the experiment's outcome is unknown, preventing any assessment of the claim's validity.",
+  "**Poor Reproducibility/Provenance:** The dirty git status and missing configuration/data digests make the experiment non-reproducible and unreliable.",
+  "**Undefined Damage Model:** The specific mechanism of 'local structural loss' is not rigorously defined or controlled across different dimensions, which is critical for a fair comparison."
+]
+```
+
+## Alternative Explanations
+
+```json
+[
+  "The observed differences in network dynamics might be due to uncalibrated simulation parameters (e.g., spiking threshold, learning rate) rather than the intrinsic dimensionality of the network structure.",
+  "The perceived robustness might be correlated with the total number of connections or neurons, rather than the dimensionality of the embedding space, suggesting a confounding variable.",
+  "The results could be artifacts of the specific random seed (seed 42) or the simulation environment setup, rather than a fundamental property of 5D topology."
+]
+```
+
+## Reproducibility
+
+```json
+{
+  "git_commit": "f0aca72400a08c247dbc535969332677fec305ad",
+  "experiment_status": "completed",
+  "evidence_mode": "not_reported",
+  "git_dirty": "True",
+  "source_freeze_sha": "NOT_AVAILABLE",
+  "configuration_path": "F:\\Brain-5D\\configs\\learning_experiment.yaml",
+  "configuration_sha256": "NOT_AVAILABLE",
+  "experiment_manifest_digest": "b9ab6fd5aab5b4641d71ff28071f13b4b466a36798e3291d59206f2a2fdfbb3b",
+  "data_ids": "NOT_AVAILABLE",
+  "evid_ids": "NOT_AVAILABLE",
+  "protocol_id": "NOT_AVAILABLE",
+  "protocol_digest": "NOT_AVAILABLE",
+  "data_digest": "NOT_AVAILABLE",
+  "evid_digests": "[]"
+}
+```
+
+## Epistemic Status
+
+```json
+{
+  "implementation": "NOT_DETERMINED",
+  "experiment_data": "ABSENT",
+  "registered_evidence": "ABSENT",
+  "ai_interpretation": "PRESENT",
+  "human_review": "PENDING",
+  "claim_support": "NOT_DETERMINED",
+  "rq_status": "IN_PROGRESS"
+}
+```
+
+## Missing Evidence
+
+```json
+[
+  "Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.",
+  "Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.",
+  "Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used."
+]
+```
+
+## Recommended Follow-up Experiments
+
+```json
+[
+  "Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).",
+  "Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).",
+  "Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance."
+]
+```
+
+## Conclusion
+
+```json
+"The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn."
+```
+
+## Human Review
+
+```json
+{
+  "status": "PENDING",
+  "reviewer": null,
+  "reviewed_at": null,
+  "disposition": null,
+  "comments": null
+}
+```
+
+AI confidence: 0.1
+This value describes internal AI confidence and is not a statistical confidence interval.
+
+Content digest: 12979faa8a4bc5c9f6c6440d144efecd1646b269b88c4319dae784e241d35dfb
+JSON/Markdown link digest: 4ffbb365b5cf90afeb157e61d9712b4228d86c18ad08d5e365da66f086871c48

research/experiments/EXP-5D-0001-v3/reports/AIRR-2026-0001.review.json

@@ -0,0 +1,8 @@
+{
+  "report_id": "AIRR-2026-0001",
+  "review_status": "accepted_as_interpretation",
+  "reviewer": "Thomas Heisig",
+  "reviewed_at": "2026-09-08T09:14:15.429644+00:00",
+  "comments": "Restart in a later Run",
+  "report_content_digest": "12979faa8a4bc5c9f6c6440d144efecd1646b269b88c4319dae784e241d35dfb"
+}

research/experiments/EXP-5D-0001-v3/summary.md

@@ -0,0 +1,56 @@
+# EXP-5D-0001-v3: Zusammenfassung
+
+Diese Zusammenfassung wurde nach Abschluss des Laufs durch den internen Research Assistant aus den Experimentartefakten und dem AIRR erstellt. Sie beschreibt die Daten, die Berichte und deren wissenschaftliche Grenzen.
+
+## Versuchsuebersicht
+
+- Status: `completed`
+- Forschungsfragen: RQ-5D-001
+- Hypothesen: H-5D-001-A
+- Durchlaeufe: `15`
+- Laufmodus: `EXPLORATORY`
+- Netzwerkmodus: `OFFLINE`
+
+## Artefakte
+
+- [analysis/AIAR-critical_reviewer-20260903211303830974-fa63da1e.json](analysis/AIAR-critical_reviewer-20260903211303830974-fa63da1e.json)
+- [analysis/AIAR-scientific_analyst-20260903211248541726-fa63da1e.json](analysis/AIAR-scientific_analyst-20260903211248541726-fa63da1e.json)
+- [analysis/AIAR-scientific_writer-20260903211319887624-fa63da1e.json](analysis/AIAR-scientific_writer-20260903211319887624-fa63da1e.json)
+- [DATA/runs.json](DATA/runs.json)
+- [manifest.json](manifest.json)
+- [report.md](report.md)
+- [reports/AIRR-2026-0001.json](reports/AIRR-2026-0001.json)
+- [reports/AIRR-2026-0001.md](reports/AIRR-2026-0001.md)
+- [workflow.json](workflow.json)
+
+## AI-Bericht
+
+- AIRR-Status: `generated`
+- AIRR: [AIRR-2026-0001.md](reports/AIRR-2026-0001.md)
+- AIRR JSON: [AIRR-2026-0001.json](reports/AIRR-2026-0001.json)
+- Wissenschaftliche Evidenz: `false`
+- Human Review: `PENDING`
+
+### KI-Einschaetzung
+
+Die KI bewertet den vorliegenden Datensatz wie folgt:
+
+The current evidence base is critically insufficient and fundamentally flawed for evaluating the claim that five-dimensional organization improves robustness against local structural losses. The experiment (EXP-5D-0001-v3) is severely compromised by multiple methodological failures, including insufficient statistical power, lack of reported results, and poor provenance tracking. No scientific conclusion can be drawn.
+
+KI-Konfidenz: `0.1`
+
+Angeforderte zusaetzliche Nachweise:
+
+- Quantitative metrics comparing network stability, recovery time, and information flow in 5D vs. 3D/2D networks after controlled structural damage.
+- Detailed statistical analysis reports confirming the significance of any observed differences in network dynamics, including p-values and confidence intervals.
+- Full, reproducible data logs and analysis pipelines for EXP-5D-0001-v3, including all parameters, seeds, and the specific damage protocol used.
+
+Empfohlene Folgeexperimente:
+
+- Conduct a statistically powered ablation study comparing 5D, 3D, and 2D networks under controlled, systematically varied damage conditions (e.g., random node removal, targeted structural loss).
+- Systematically vary the structural loss parameters (e.g., percentage of nodes removed, pattern of removal) across all dimensions (2D, 3D, 5D) to quantify the robustness metric (e.g., recovery time, information flow decay).
+- Perform a comprehensive statistical replication of the core simulation (EXP-5D-0001-v3) using a minimum of 30 runs and multiple seeds to establish statistical significance and assess variance.
+
+## Wissenschaftliche Grenze
+
+Die KI-Auswertung ist post-hoc, steuert den Lauf nicht und ersetzt keine menschliche wissenschaftliche Pruefung oder Evidenzfreigabe.