Release v0.8.0: map model recurrence to governed operations

.gitignore

@@ -1,2 +1,4 @@
 __pycache__/
 *.py[cod]
+output/
+.playwright-cli/

CITATION.bib

@@ -3,5 +3,5 @@
   title        = {Awesome Loop Engineering},
   year         = {2026},
   howpublished = {\url{https://github.com/ChaoYue0307/awesome-loop-engineering}},
-  note         = {Curated resources for Loop Engineering, version 0.7.0}
+  note         = {Curated resources for Loop Engineering, version 0.8.0}
 }

CITATION.cff

@@ -2,7 +2,7 @@ cff-version: 1.2.0
 message: "If this repository is useful in your work, please cite it as below."
 title: "Awesome Loop Engineering"
 type: software
-version: 0.7.0
+version: 0.8.0
 date-released: 2026-07-18
 authors:
   - family-names: "He"

COMPARISON.md

@@ -2,6 +2,14 @@
 
 Loop Engineering is easiest to understand by separating it from nearby practices.
 
+## Model-Level Recurrence vs Loop Engineering
+
+**Model-level recurrence** repeats a learned layer, block, module, or latent-state update inside one model inference. It can provide adaptive depth, latent reasoning, parameter sharing, or iterative world-state refinement.
+
+**Loop Engineering** governs repeated agent work outside the model: what triggers a run, which work enters, what tools and permissions apply, what external evidence gates progress, what state survives, and when the system retries, escalates, or stops.
+
+A looped Transformer, recurrent-depth language model, or LoopWM can be the model inside an agent loop. It does not replace the Loop Contract because hidden-state iteration alone does not provide durable receipts, independent verification, permission boundaries, or accountable human handoff.
+
 ## Prompt Engineering vs Loop Engineering
 
 **Prompt Engineering** asks: what should I say to the model now?

CONTRIBUTING.md

@@ -28,10 +28,10 @@ For an inaccurate summary, contribution, novelty, impact, authorship, date, venu
 1. Add one metadata-rich row with a resource type label, verified year/source context, and one specific key feature:
 
 ```md
-| 📄 **[Title](https://example.com)**<br><sub>Paper</sub> | **2026** · arXiv<br><sub>First Author et al.</sub> | One sentence explaining the resource's contribution to Loop Engineering. |
+| 📄 **[Title](https://example.com)**<br><sub>Paper</sub> | **2026** · arXiv<br><sub>First Author et al.</sub> | One sentence explaining the resource's contribution to Loop Engineering. | **Tier A** · Research preprint<br><sub>Preprint; inspect methods and evaluation</sub> |
 ```
 
-When a source exposes no reliable publication date, omit the year and keep only the original venue or publishing platform. Do not infer a date from a search-result snippet. Maintainers regenerate the table layout and structured exports after review.
+When a source exposes no reliable publication date, omit the year and keep only the original venue or publishing platform. Do not infer a date from a search-result snippet. Maintainers derive the evidence tier from the audited source class and regenerate the table layout and structured exports after review.
 
 1. Open a PR using the template and explain:
    - why the resource belongs;

DEFINITION.md

@@ -10,6 +10,8 @@ Use this definition to scope a system, review a design, or cite the concept cons
 
 Prompt engineering improves what you ask the model, context engineering improves what the model can see, harness engineering improves the environment around one agent run, and Loop Engineering governs repeated agent work over time.
 
+Model-level recurrence is a complementary architecture choice: shared learned computation may repeat inside one inference, while Loop Engineering governs how that model participates in recurring work across runs.
+
 ## Longer Definition
 
 Loop Engineering connects prompt, context, and harness decisions across runs. It replaces turn-by-turn human steering with a reviewable operating contract: what starts the loop, where work comes from, which agents act, what they can access, how evidence gates results, what state survives, how retries are bounded, and when a human takes over.

FUTURE-DIRECTIONS.md

@@ -60,6 +60,23 @@ Every study or operating report should include:
 - raw per-run outcomes plus the script that produces aggregate tables;
 - privacy, licensing, and redaction decisions for data that cannot be public.
 
+## Cross-Layer Study: Model Recurrence And Operational Loops
+
+Model recurrence and operational recurrence spend compute at different boundaries. A looped model repeats learned computation inside one inference; an operational loop repeats model calls, tool actions, verification, and state transitions around a real task. Treating both simply as "more test-time compute" hides different costs, stopping rules, evidence, and failure modes.
+
+Use a matched cross-layer study to decide where an additional unit of compute belongs:
+
+| Question | Compare | Hold fixed | Report |
+| --- | --- | --- | --- |
+| **Where does another pass help?** | More recurrent depth inside one call vs another evidence-aware agent pass | Task set, model family where possible, tool access, verifier, and total compute or cost | Verified completion, false completion, latency, cost, and gain by task difficulty |
+| **Who should stop the iteration?** | Fixed depth, learned halting, confidence-based stopping, external verification, and human escalation | Acceptance policy and maximum budget | Calibration, premature stops, wasted passes, budget breaches, and escalation quality |
+| **What state should survive?** | Hidden-state refinement, visible reasoning state, and durable external checkpoints | Initial context and task history | Resume fidelity, contamination, replayability, and diagnostic value after failure |
+| **Can inner and outer loops cooperate?** | Standard model, looped model, operational loop, and looped model inside an operational loop | Harness, permissions, task intake, and evidence gate | Interaction effects, marginal value per pass, correlated failures, and the simplest Pareto-efficient configuration |
+
+**Starter experiment:** choose 50 reasoning or coding tasks with deterministic checks. Compare a single standard inference, matched-cost internal recurrence, fixed external retry, and evidence-aware external retry. Then run the combined system only if each component shows independent value. Publish per-task compute, every stopping decision, verifier results, and failure traces.
+
+**Completion gate:** the study identifies when internal recurrence replaces, complements, or fails to improve external iteration without crediting hidden-state depth for operational guarantees it does not provide. A useful result may show that one layer is unnecessary for a given workload.
+
 ## Priority Map
 
 The tiers describe dependency order, not prestige. Establish trustworthy state transitions before optimizing complex delegation.

README.de.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Ein Loop entdeckt Arbeit, übergibt sie an einen oder mehrere Agenten, prüft da
 
 Loop Engineering umfasst wiederkehrende KI-Agenten- und Coding-Agent-Systeme mit expliziten Auslösern, Verifikation und dauerhaftem Zustand. Software-Event-Loops, Regelungstechnik, Growth Loops, generische Automatisierung und nicht-KI-bezogene Feedback-Loops gehören nicht dazu.
 
-Starte mit 545 geprüften Ressourcen, 20 operativen Loop-Patterns, 20 schemavalidierten Loop-Contracts und 8 Runtime-Startern (3 ausführbare Programme und 5 anpassbare Vorlagen), ergänzt durch eine Community-Galerie und 8 Sprachen.
+Starte mit 579 geprüften Ressourcen, 20 operativen Loop-Patterns, 20 schemavalidierten Loop-Contracts und 8 Runtime-Startern (3 ausführbare Programme und 5 anpassbare Vorlagen), ergänzt durch eine Community-Galerie und 8 Sprachen.
 
 ## Mentales Modell
 

README.es.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Un loop descubre trabajo, lo delega a uno o más agentes, verifica el resultado,
 
 Loop Engineering abarca sistemas recurrentes de agentes de IA y coding agents con disparadores explícitos, verificación y estado duradero. No incluye event loops de software, teoría de control, growth loops, automatización genérica ni feedback loops no relacionados con IA.
 
-Empieza con 545 recursos auditados, 20 patrones operativos, 20 loop contracts validados por esquema y 8 puntos de partida de runtime (3 ejecutables y 5 plantillas listas para adaptar), además de una galería comunitaria y 8 idiomas.
+Empieza con 579 recursos auditados, 20 patrones operativos, 20 loop contracts validados por esquema y 8 puntos de partida de runtime (3 ejecutables y 5 plantillas listas para adaptar), además de una galería comunitaria y 8 idiomas.
 
 ## Modelo Mental
 

README.fr.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Un loop découvre du travail, le délègue à un ou plusieurs agents, vérifie l
 
 Loop Engineering couvre les systèmes récurrents d'agents IA et de coding agents avec des déclencheurs explicites, une vérification et un état durable. Il n'inclut ni les event loops logiciels, ni la théorie du contrôle, ni les growth loops, ni l'automatisation générique, ni les feedback loops sans rapport avec l'IA.
 
-Commencez avec 545 ressources auditées, 20 patterns opérationnels, 20 loop contracts validés par schéma et 8 points de départ d'exécution (3 exécutables et 5 modèles prêts à adapter), ainsi qu'une galerie communautaire et 8 langues.
+Commencez avec 579 ressources auditées, 20 patterns opérationnels, 20 loop contracts validés par schéma et 8 points de départ d'exécution (3 exécutables et 5 modèles prêts à adapter), ainsi qu'une galerie communautaire et 8 langues.
 
 ## Modèle Mental
 

README.ja.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Loop は作業を発見し、1 つ以上の agents に委任し、結果を確
 
 Loop Engineering は、明示的なトリガー、検証、永続状態を備えた反復型 AI agent / coding agent システムを対象とします。software event loop、制御理論、growth loop、一般的な workflow automation、非 AI の feedback loop は含みません。
 
-545 件の監査済みリソース、20 個の運用パターン、20 個の schema 検証済み loop contract、8 個の runtime starter（3 個の実行ファイルと 5 個の適応可能なテンプレート）から始められます。コミュニティ gallery と 8 言語の導入も利用できます。
+579 件の監査済みリソース、20 個の運用パターン、20 個の schema 検証済み loop contract、8 個の runtime starter（3 個の実行ファイルと 5 個の適応可能なテンプレート）から始められます。コミュニティ gallery と 8 言語の導入も利用できます。
 
 ## メンタルモデル
 

README.ko.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Loop 는 작업을 발견하고, 하나 이상의 agents 에게 위임하고,
 
 Loop Engineering 은 명시적 트리거, 검증, 지속 상태를 갖춘 반복 AI agent / coding agent 시스템을 다룹니다. software event loop, control theory, growth loop, 일반적인 workflow automation, 비 AI feedback loop 는 포함하지 않습니다.
 
-545개의 감사된 리소스, 20개의 운영 loop 패턴, 20개의 스키마 검증 loop contract, 8개의 runtime starter(실행 파일 3개와 적용 가능한 템플릿 5개)로 바로 시작할 수 있습니다. 커뮤니티 gallery 와 8개 언어 안내도 제공합니다.
+579개의 감사된 리소스, 20개의 운영 loop 패턴, 20개의 스키마 검증 loop contract, 8개의 runtime starter(실행 파일 3개와 적용 가능한 템플릿 5개)로 바로 시작할 수 있습니다. 커뮤니티 gallery 와 8개 언어 안내도 제공합니다.
 
 ## 멘탈 모델
 

README.md

@@ -38,7 +38,7 @@ configs:
 <h1 align="center">Awesome Loop Engineering Dataset</h1>
 
 <p align="center">
-  A source-audited dataset of 555 papers, official docs, tools, benchmarks, patterns, critiques, and implementation guides for recurring AI-agent systems.
+  A source-audited dataset of 579 papers, official docs, tools, benchmarks, patterns, critiques, and implementation guides for recurring AI-agent systems.
 </p>
 
 <p align="center">
@@ -52,16 +52,16 @@ configs:
 
 Each row pairs one source with a concise editorial assessment, bibliographic metadata, lifecycle and audience facets, evidence classification, audit status, and point-in-time repository statistics when applicable.
 
-Current release: **v0.7.0**
+Current release: **v0.8.0**
 
 | Surface | Count |
 | --- | ---: |
-| Source-audited resources | 555 |
+| Source-audited resources | 579 |
 | Operational patterns | 20 |
 | Validated loop contracts | 20 |
 | Runtime starters | 8 |
 
-The source audit dated **2026-07-18** reports these row statuses: **502 reachable public**, **5 access restricted**, **48 repository-native**, and **0 broken or unreachable**.
+The source audit dated **2026-07-18** reports these row statuses: **528 reachable public**, **3 access restricted**, **48 repository-native**, and **0 broken or unreachable**.
 
 ## What A Loop Contract Is
 
@@ -100,6 +100,10 @@ papers = resources.filter(lambda row: row["resource_type"] == "Paper")
 verification = resources.filter(
     lambda row: "verification" in row["lifecycle_stages"].split(";")
 )
+model_recurrence = resources.filter(
+    lambda row: row["loop_layer"] == "model"
+    and row["scope_fit"] == "adjacent"
+)
 ```
 
 For pandas:
@@ -118,6 +122,7 @@ Use `url` as the durable join key. `row_id` and `source_line` are positional and
 
 - Find primary sources and implementation references for recurring agent systems.
 - Compare works by lifecycle, audience, evidence class, source type, and publication metadata.
+- Separate model, agent, harness, workflow, operations, evaluation, and cross-layer resources without treating adjacent model recurrence as a complete operational loop.
 - Build literature maps, reading lists, dashboards, or retrieval indexes.
 - Audit contribution, novelty, impact, provenance, and evidence claims.
 - Find reusable patterns, contracts, schemas, executable examples, and copy/paste runtime templates.
@@ -144,6 +149,7 @@ The primary configuration is `resources`, with one `train` split backed by the n
 | --- | --- | --- |
 | Identity | `row_id`, `title`, `url`, `canonical_url`, `resource_type`, `domain` | What the resource is and where it lives. |
 | Editorial assessment | `annotation`, `key_contribution`, `novelty`, `impact` | What the resource contributes to recurring agent systems. |
+| Scope | `loop_layer`, `scope_fit` | Where recurrence lives and whether the source is direct, enabling, or adjacent to operational Loop Engineering. |
 | Navigation | `section`, `collection`, `user_goal`, `lifecycle_stages`, `audience` | Where the resource fits and who it serves. |
 | Evidence | `evidence_class`, `evidence_tier`, `signal`, `signal_strength`, `source_status`, `audited_at` | What kind of evidence or provenance is available. |
 | Publication | `authors`, `publication_date`, `publication_year`, `publication_venue`, `publisher`, `doi`, `arxiv_id`, `primary_category` | Bibliographic data exposed by the canonical source. |
@@ -183,6 +189,6 @@ GitHub Releases define versioned snapshots; cite a release or commit for reprodu
   title        = {Awesome Loop Engineering},
   year         = {2026},
   howpublished = {\url{https://github.com/ChaoYue0307/awesome-loop-engineering}},
-  note         = {Version 0.7.0}
+  note         = {Version 0.8.0}
 }
 ```

README.pt-BR.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -28,7 +28,7 @@ Um loop descobre trabalho, delega para um ou mais agentes, verifica o resultado,
 
 Loop Engineering abrange sistemas recorrentes de AI agents e coding agents com gatilhos explícitos, verificação e estado durável. Não inclui event loops de software, teoria de controle, growth loops, automação genérica nem feedback loops sem relação com IA.
 
-Comece com 545 recursos auditados, 20 patterns operacionais, 20 loop contracts validados por schema e 8 starters de runtime (3 executáveis e 5 templates prontos para adaptação), além de uma galeria comunitária e 8 idiomas.
+Comece com 579 recursos auditados, 20 patterns operacionais, 20 loop contracts validados por schema e 8 starters de runtime (3 executáveis e 5 templates prontos para adaptação), além de uma galeria comunitária e 8 idiomas.
 
 ## Modelo Mental
 

README.zh-CN.md

@@ -1,6 +1,6 @@
 # Awesome Loop Engineering
 
-<!-- last-synced: 2026-07-17 -->
+<!-- last-synced: 2026-07-18 -->
 
 <p align="center">
   <img src="assets/awesome-loop-engineering-cover.png" alt="Awesome Loop Engineering cover" width="100%">
@@ -32,7 +32,7 @@ Prompt engineering 改进你对模型说什么。Context engineering 改进模
 
 Loop Engineering 专指具备明确触发、外部验证和持久状态的可重复 AI-agent 与 coding-agent 系统，不包括软件事件循环、控制论、增长循环、通用 workflow automation 或非 AI feedback loop。
 
-可直接从 545 条经审核资源、20 个操作模式、20 个经 schema 校验的 loop contracts 和 8 个 runtime starters（3 个可直接执行，5 个可复制改造）开始，并通过社区 gallery 和 8 种语言入口继续探索。
+可直接从 579 条经审核资源、20 个操作模式、20 个经 schema 校验的 loop contracts 和 8 个 runtime starters（3 个可直接执行，5 个可复制改造）开始，并通过社区 gallery 和 8 种语言入口继续探索。
 
 ## 一句话定位
 

TAXONOMY.md

@@ -1,6 +1,17 @@
 # Loop Engineering Taxonomy
 
-Classify an AI-agent loop by its trigger, work intake, verification method, state model, agent topology, and operating domain.
+Classify recurrence first by where it lives, then classify an operational AI-agent loop by its trigger, work intake, verification method, state model, agent topology, and operating domain.
+
+## By Loop Layer
+
+- **Model layer**: a learned layer, block, module, or latent-state update repeats within one inference. This is an adjacent foundation for Loop Engineering, not a complete operational loop.
+- **Agent layer**: a model alternates reasoning, tool calls, observations, and updates while solving one task.
+- **Harness layer**: tools, context, permissions, sandboxes, evaluators, and checkpoints control one or more agent runs.
+- **Workflow layer**: triggers, intake, delegation, verification, durable state, budgets, and next-action policy govern repeated work.
+- **Operations layer**: telemetry, approvals, incident response, cost controls, rollback, and human ownership govern deployed loops.
+- **Evaluation layer**: benchmarks, graders, trace analysis, and challenge sets measure behavior across the other layers.
+
+Dataset rows pair this facet with `scope_fit`: `direct` for operational loop resources, `enabling` for components and evidence, and `adjacent` for model architectures or neighboring ecosystem maps.
 
 ## By Trigger
 

data/README.md

@@ -1,6 +1,6 @@
 # Dataset Exports
 
-Download deterministic tabular exports of all 545 source-audited resources.
+Download deterministic tabular exports of all 579 source-audited resources.
 
 - `resources.csv` - Tabular export for spreadsheets and ad hoc analysis.
 - `resources.jsonl` - JSON Lines export and source for the Hugging Face Parquet build.
@@ -11,13 +11,14 @@ Download deterministic tabular exports of all 545 source-audited resources.
 - `arxiv_publication_overrides.csv` - Human-verified conference, journal, and workshop records backed by official proceedings, DOI registries, OpenReview, or current author-supplied acceptance notes.
 - `arxiv_publication_audit.csv` - Complete decision table for every arXiv-linked resource: published, accepted, or preprint-only.
 
-The exports preserve each section and annotation while adding three discovery layers:
+The exports preserve each section and annotation while adding four discovery layers:
 
+- **Scope facets**: `loop_layer` identifies where recurrence lives (`model`, `agent`, `harness`, `workflow`, `operations`, `evaluation`, or `cross-layer`); `scope_fit` distinguishes resources that are `direct`, `enabling`, or `adjacent` to operational Loop Engineering.
 - **Task facets**: `collection`, `user_goal`, `lifecycle_stages`, and `audience` answer why a reader needs the source and where it fits in the Loop Contract.
-- **Evidence facets**: `evidence_class`, `signal_strength`, `source_status`, canonical URL, source metadata, GitHub statistics, arXiv ID, and audit timestamp separate source provenance from popularity or editorial judgment.
+- **Evidence facets**: `evidence_class`, `evidence_tier`, `signal_strength`, `source_status`, canonical URL, source metadata, GitHub statistics, arXiv ID, and audit timestamp separate source provenance from popularity or editorial judgment.
 - **Publication facets**: `authors`, `publication_date`, `publication_year`, `publication_venue`, `publisher`, `doi`, `publication_note`, `primary_category`, and `metadata_source` provide a paper-like bibliographic row without inventing missing facts.
 
-`key_contribution`, `novelty`, and `impact` are resource-specific. `signal` states the evidence basis and limits; GitHub stars and forks provide point-in-time context, never proof of reliability. `signal_strength` is `high` for primary official documentation and benchmarks, `medium` for inspectable implementations, papers, patterns, and locally maintained artifacts, `contextual` for practitioner analysis and curated lists, and `unverified` only when the latest audit cannot validate availability.
+`key_contribution`, `novelty`, and `impact` are resource-specific. Model-level recurrence is retained as an adjacent foundation because it repeats learned blocks or latent-state updates inside one inference; it does not by itself supply work intake, external verification, durable state, budgets, or human handoff. `evidence_tier` follows the public A-C source hierarchy: A for primary or official artifacts, B for implementation-grounded practice and risk analysis, and C for curated synthesis. `signal` states the evidence basis and limits; GitHub stars and forks provide point-in-time context, never proof of reliability. `signal_strength` is `high` for primary official documentation and benchmarks, `medium` for inspectable implementations, papers, patterns, and locally maintained artifacts, `contextual` for practitioner analysis and curated lists, and `unverified` only when the latest audit cannot validate availability.
 
 ## Load And Query
 
@@ -34,6 +35,10 @@ verification_papers = resources.filter(
     lambda row: row["resource_type"] == "Paper"
     and "verification" in row["lifecycle_stages"].split(";")
 )
+model_recurrence = resources.filter(
+    lambda row: row["loop_layer"] == "model"
+    and row["scope_fit"] == "adjacent"
+)
 ```
 
 For pandas:

data/arxiv_publication_audit.csv

@@ -33,6 +33,16 @@ ale-0141,2607.11288,Mako: A Self-Evolving Agentic Operating System for Autonomou
 ale-0142,2607.10856,How Do Practitioners Build SE Agents? Insights from a Mixed-Methods Study,How Do Practitioners Build SE Agents? Insights from a Mixed-Methods Study,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-17
 ale-0143,2607.10113,Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Libraries,Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Libraries,2026,accepted,Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Libraries,2026,2026,Transactions on Machine Learning Research (TMLR),OpenReview,,https://openreview.net/forum?id=cjU3YbcRr8,https://openreview.net/forum?id=cjU3YbcRr8,Current arXiv acceptance note and OpenReview record,1.000,2026-07-17
 ale-0144,2605.27276,SIA: Self Improving AI with Harness and Weight Updates,SIA: Self Improving AI with Harness & Weight Updates,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0152,2510.25741,Scaling Latent Reasoning via Looped Language Models,Scaling Latent Reasoning via Looped Language Models,2025,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0158,2602.11698,SpiralFormer: Looped Transformers Can Learn Hierarchical Dependencies via Multi-Resolution Recursion,SpiralFormer: Looped Transformers Can Learn Hierarchical Dependencies via Multi-Resolution Recursion,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0159,2605.23872,Training-Free Looped Transformers,Training-Free Looped Transformers,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0160,2604.07822,"Loop, Think, & Generalize: Implicit Reasoning in Recurrent-Depth Transformers","Loop, Think, & Generalize: Implicit Reasoning in Recurrent-Depth Transformers",2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0161,2607.13491,DeepLoop: Depth Scaling for Looped Transformers,DeepLoop: Depth Scaling for Looped Transformers,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0162,2604.21106,How Much Is One Recurrence Worth? Iso-Depth Scaling Laws for Looped Language Models,How Much Is One Recurrence Worth? Iso-Depth Scaling Laws for Looped Language Models,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0164,2606.18208,Looped World Models,Looped World Models,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0165,2606.31779,Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers,Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0166,2605.16048,Looped SSMs: Depth-Recurrence and Input Reshaping for Time Series Classification,Looped SSMs: Depth-Recurrence and Input Reshaping for Time Series Classification,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
+ale-0167,2605.26106,Looped Diffusion Language Models,Looped Diffusion Language Models,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-18
 ale-0154,2605.18747,Code as Agent Harness,Code as Agent Harness,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-17
 ale-0155,2605.20456,Agentic Agile-V: From Vibe Coding to Verified Engineering,Agentic Agile-V: From Vibe Coding to Verified Engineering in Software and Hardware Development,2026,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-17
 ale-0156,2509.06216,Agentic Software Engineering: Foundational Pillars and a Research Roadmap,Agentic Software Engineering: Foundational Pillars and a Research Roadmap,2025,preprint-only,,,,,,,,,DBLP and Crossref exact-title audit,,2026-07-17

data/first_seen.json

@@ -65,5 +65,29 @@
   "https://arxiv.org/abs/2510.11967": "2026-07-18",
   "https://arxiv.org/abs/2512.08296": "2026-07-18",
   "https://aclanthology.org/2026.acl-long.278/": "2026-07-18",
-  "https://aclanthology.org/2026.acl-long.337/": "2026-07-18"
+  "https://aclanthology.org/2026.acl-long.337/": "2026-07-18",
+  "https://openreview.net/forum?id=HyzdRiR9Y7": "2026-07-18",
+  "https://proceedings.mlr.press/v202/giannou23a.html": "2026-07-18",
+  "https://openreview.net/forum?id=HHbRxoDTxE": "2026-07-18",
+  "https://proceedings.mlr.press/v267/xu25x.html": "2026-07-18",
+  "https://iclr.cc/virtual/2025/poster/28971": "2026-07-18",
+  "https://proceedings.neurips.cc/paper_files/paper/2025/hash/3b01972cf31e6fa0fe29e4b8b5c2a0a1-Abstract-Conference.html": "2026-07-18",
+  "https://papers.nips.cc/paper_files/paper/2025/hash/8b08bbf8b420faa6eeb4020720582ec7-Abstract-Conference.html": "2026-07-18",
+  "https://arxiv.org/abs/2510.25741": "2026-07-18",
+  "https://iclr.cc/virtual/2026/poster/10009450": "2026-07-18",
+  "https://iclr.cc/virtual/2026/poster/10011117": "2026-07-18",
+  "https://iclr.cc/virtual/2026/poster/10007767": "2026-07-18",
+  "https://openreview.net/forum?id=eQaJSRZiGn": "2026-07-18",
+  "https://openreview.net/forum?id=ri0LAMdhd9": "2026-07-18",
+  "https://arxiv.org/abs/2602.11698": "2026-07-18",
+  "https://arxiv.org/abs/2605.23872": "2026-07-18",
+  "https://arxiv.org/abs/2604.07822": "2026-07-18",
+  "https://arxiv.org/abs/2607.13491": "2026-07-18",
+  "https://arxiv.org/abs/2604.21106": "2026-07-18",
+  "https://aclanthology.org/2026.findings-acl.796/": "2026-07-18",
+  "https://arxiv.org/abs/2606.18208": "2026-07-18",
+  "https://arxiv.org/abs/2606.31779": "2026-07-18",
+  "https://arxiv.org/abs/2605.16048": "2026-07-18",
+  "https://arxiv.org/abs/2605.26106": "2026-07-18",
+  "https://github.com/huskydoge/Awesome-Loop-Models": "2026-07-18"
 }

data/resources.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d492be344c2982db477dfda181c9080719e9d11441d632828c2b95074403f3bf
-size 446433
+oid sha256:d54b0458ecbf0a9c8ddc1c56d731c4111a04e8633a1c9e8f9b7ea1e70fce97ca
+size 466752

design-qa.md

@@ -19,7 +19,7 @@
 - Contract and lifecycle: purpose-built mobile summaries replace desktop-wide figures at 560 px and below.
 - Runtime map: desktop figure plus eight labeled starters in a balanced 4 x 2 grid; mobile uses the explanatory copy and one-column starter sequence instead of shrinking the wide figure.
 - Maturity model: desktop figure plus semantic levels 00-06; mobile keeps the readable level rows and hides the wide supporting figure.
-- Social preview: 1280 x 640, exact 545-resource, 20-pattern, 20-contract, and 8-starter counts.
+- Social preview: 1280 x 640, exact 579-resource, 20-pattern, 20-contract, and 8-starter counts, with the model-to-operations scope line.
 - Interactive hero, desktop: 1440 x 1000, one ready WebGL canvas, no horizontal overflow, and meaningful agent-system labels.
 - Interactive hero, mobile: 390 x 844, one ready WebGL canvas, no horizontal overflow, and all six stations inside the frame.
 - Interactive hero motion: the frame counter advanced by 47 frames during a 450 ms pointer-parallax check.
@@ -71,14 +71,16 @@
 - Phone: 390 x 844, no horizontal overflow. The compact loop keeps all six stations, the moving work packet, legend, outcome key, and stage selector visible without crossing the hero copy.
 - Mobile navigation: opens all nine project links, reports `aria-expanded="true"`, closes on Escape, and restores focus to the menu button.
 - Pattern filter: `accessibility` returns exactly one of the 20 patterns.
-- Resource Atlas: remains idle above the fold, loads on approach or a direct resource anchor, and reports `545 of 545 resources | showing 8` on mobile.
-- Resource interactions: `SWE-bench` returns 9 matches, the paper filter narrows those to 6, Reset restores all 545, and Show more expands the rendered set from 8 to 16.
+- Resource Atlas: remains idle above the fold, loads on approach or a direct resource anchor, and reports `579 of 579 resources | showing 8` on a fresh mobile load.
+- Resource interactions: the Model loop-layer filter returns 24 of 579 resources (23 papers plus Awesome Loop Models), paginates eight at a time on mobile, and renders LoopWM with its 2026 arXiv record and adjacent-scope label.
+- Resource row schema: desktop exposes Work, Year, Published at, and Evidence; mobile stacks the same labels. A `ReasoningBank` query returns one row with its ICLR venue, arXiv identifier, Tier A evidence, task fit, and source record without horizontal overflow.
 - Scene motion: the canvas frame counter advanced from 640 to 662 over 900 ms. Two canvas crops taken 1.2 seconds apart changed 17.79% of sampled RGB channels.
-- Scene pixels: the sampled desktop crop has 15.71% active pixels with RGB standard deviations of 35.83, 29.77, and 23.01; mobile has 27.84% active pixels with deviations of 41.27, 30.61, and 19.52. Both renders are nonblank and well contrasted.
+- Scene pixels: the current desktop canvas has 7,062 colors with RGB standard deviations of 35.93, 29.76, and 22.29; mobile has 5,755 colors with deviations of 40.73, 30.54, and 19.70. Both renders are nonblank and well contrasted.
 - Runtime starters: desktop cards form two equal-height rows of four; mobile cards are 358 px wide within a 390 px document and preserve label, icon, heading, and copy alignment.
 - Future directions: direct-anchor restoration waits for the Resource Atlas reflow, all three audience tracks remain readable, and the agenda CTA is fully visible at desktop and mobile widths.
 - Console: no page-origin errors or warnings during mobile navigation, pattern filtering, and Resource Atlas interactions.
-- Source audit: 545 rows checked; 496 public sources reachable, 1 access-restricted, 48 repository-native, and 0 broken or unreachable as of 2026-07-17 UTC.
-- Data and copy: the website, README, translations, social preview source, release copy, and Hugging Face card all report 545 resources, 20 patterns, 20 contracts, 8 runtime starters, and version 0.7.0.
+- Narrow phone: 320 x 844, document width equals viewport width; the headline stays within 16 px gutters and the Three.js canvas remains fully inside the 288 px content width.
+- Source audit: 579 rows checked; 528 public sources reachable, 3 access-restricted, 48 repository-native, and 0 broken or unreachable as of 2026-07-18 UTC.
+- Data and copy: the website, README, translations, social preview source, release copy, and Hugging Face card all report 579 resources, 20 patterns, 20 contracts, 8 runtime starters, and version 0.8.0. The dataset exposes 50 fields, including `loop_layer` and `scope_fit`.
 
 final result: passed

docs/index.html

@@ -4,7 +4,7 @@
     <meta charset="utf-8">
     <meta name="viewport" content="width=device-width, initial-scale=1">
     <title>Awesome Loop Engineering</title>
-    <meta name="description" content="545 source-audited resources, 20 patterns, 20 validated contracts, and 8 runtime starters for recurring, stateful, verified AI-agent systems.">
+    <meta name="description" content="579 source-audited resources across model, agent, harness, workflow, and operations layers, plus 20 patterns, 20 contracts, and 8 runtime starters.">
     <meta name="robots" content="index,follow">
     <meta name="author" content="Chaoyue He">
     <link rel="canonical" href="https://chaoyue0307.github.io/awesome-loop-engineering/">
@@ -15,16 +15,16 @@
     <link href="https://fonts.googleapis.com/css2?family=IBM+Plex+Mono:wght@500;600&amp;family=Inter:wght@400;500;600;700;800&amp;display=swap" rel="stylesheet">
     <meta property="og:type" content="website">
     <meta property="og:title" content="Awesome Loop Engineering">
-    <meta property="og:description" content="545 source-audited resources, 20 patterns, 20 validated contracts, and 8 runtime starters for recurring, stateful, verified AI-agent systems.">
+    <meta property="og:description" content="579 source-audited resources across the model-to-operations stack, plus 20 patterns, 20 validated contracts, and 8 runtime starters.">
     <meta property="og:url" content="https://chaoyue0307.github.io/awesome-loop-engineering/">
-    <meta property="og:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-545-v07.png">
+    <meta property="og:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-579-v08.png">
     <meta property="og:image:width" content="1280">
     <meta property="og:image:height" content="640">
     <meta property="og:image:alt" content="Awesome Loop Engineering: design loops, not just prompts.">
     <meta name="twitter:card" content="summary_large_image">
     <meta name="twitter:title" content="Awesome Loop Engineering">
-    <meta name="twitter:description" content="545 source-audited resources, 20 patterns, 20 validated contracts, and 8 runtime starters for recurring, stateful, verified AI-agent systems.">
-    <meta name="twitter:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-545-v07.png">
+    <meta name="twitter:description" content="579 source-audited resources across the model-to-operations stack, plus a practical kit for recurring AI-agent systems.">
+    <meta name="twitter:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-579-v08.png">
     <link rel="alternate" type="application/atom+xml" title="Awesome Loop Engineering releases" href="https://github.com/ChaoYue0307/awesome-loop-engineering/releases.atom">
     <link rel="alternate" type="application/atom+xml" title="Awesome Loop Engineering commits" href="https://github.com/ChaoYue0307/awesome-loop-engineering/commits/main.atom">
     <style>
@@ -350,6 +350,18 @@
       .section-intro { max-width: 720px; margin: 0; color: var(--muted); }
       .audit-note { margin: -17px 0 28px; border-top: 1px solid var(--line); padding-top: 13px; color: var(--muted); font-size: 0.8rem; }
       .audit-note strong { color: var(--ink); }
+      .loop-layer-map {
+        display: grid;
+        grid-template-columns: repeat(3, minmax(0, 1fr));
+        margin: 0 0 28px;
+        border-top: 1px solid var(--line-strong);
+        border-bottom: 1px solid var(--line-strong);
+      }
+      .loop-layer-item { min-width: 0; padding: 18px 20px 18px 0; }
+      .loop-layer-item + .loop-layer-item { border-left: 1px solid var(--line); padding-left: 20px; }
+      .loop-layer-item span { display: block; color: var(--blue); font-family: "IBM Plex Mono", ui-monospace, monospace; font-size: 0.64rem; font-weight: 700; text-transform: uppercase; }
+      .loop-layer-item strong { display: block; margin-top: 6px; color: var(--ink); font-size: 0.92rem; }
+      .loop-layer-item p { margin: 6px 0 0; color: var(--muted); font-size: 0.78rem; line-height: 1.5; }
       .diagram { width: 100%; margin: 0; }
       .caption { margin: 13px 0 0; color: var(--quiet); font-size: 0.78rem; text-align: right; }
       .supporting-visual { margin: 0 0 34px; }
@@ -712,7 +724,7 @@
       .resource-explorer { border-top: 1px solid var(--line-strong); }
       .resource-toolbar {
         display: grid;
-        grid-template-columns: minmax(260px, 1.45fr) repeat(3, minmax(150px, 0.7fr)) auto;
+        grid-template-columns: minmax(240px, 1.4fr) repeat(4, minmax(132px, 0.66fr)) auto;
         gap: 10px;
         padding: 16px 0;
         border-bottom: 1px solid var(--line);
@@ -1133,6 +1145,9 @@
         .principle + .principle, .card + .card, .pattern + .pattern { padding-left: 0; }
         .principle, .card, .pattern { padding-top: 16px; padding-bottom: 16px; }
         .choice-why { grid-template-columns: 1fr; gap: 7px; padding: 18px 0; }
+        .loop-layer-map { grid-template-columns: 1fr; }
+        .loop-layer-item { padding: 16px 0; }
+        .loop-layer-item + .loop-layer-item { border-top: 1px solid var(--line); border-left: 0; padding-left: 0; }
         .choice-why p { font-size: 0.84rem; line-height: 1.55; }
         .choice-fit { grid-template-columns: 1fr; margin-bottom: 24px; }
         .choice-fit-column { padding: 18px 0; }
@@ -1230,7 +1245,7 @@
           "https://huggingface.co/datasets/cy0307/awesome-loop-engineering"
         ],
         "license": "https://creativecommons.org/publicdomain/zero/1.0/",
-        "image": "https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-545-v07.png",
+        "image": "https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-579-v08.png",
         "logo": "https://chaoyue0307.github.io/awesome-loop-engineering/assets/awesome-loop-engineering-logo.png",
         "author": { "@type": "Person", "name": "Chaoyue He" }
       }
@@ -1278,14 +1293,14 @@
         "@context": "https://schema.org",
         "@type": "Dataset",
         "name": "Awesome Loop Engineering resources",
-        "description": "545 source-audited resources on recurring AI-agent and coding-agent loops, with contribution, novelty, impact, and evidence fields.",
+        "description": "579 source-audited resources across model, agent, harness, workflow, operations, and evaluation layers, with explicit scope, contribution, novelty, impact, and evidence fields.",
         "url": "https://chaoyue0307.github.io/awesome-loop-engineering/",
         "sameAs": [
           "https://huggingface.co/datasets/cy0307/awesome-loop-engineering",
           "https://github.com/ChaoYue0307/awesome-loop-engineering"
         ],
         "identifier": "https://huggingface.co/datasets/cy0307/awesome-loop-engineering",
-        "version": "0.7.0",
+        "version": "0.8.0",
         "license": "https://creativecommons.org/publicdomain/zero/1.0/",
         "isAccessibleForFree": true,
         "creator": { "@type": "Person", "name": "Chaoyue He" },
@@ -1353,7 +1368,7 @@
             <p class="section-label hero-kicker"><i class="icon" data-lucide="repeat-2" aria-hidden="true"></i><span>Field guide for recurring agent systems</span></p>
             <h1>Awesome Loop Engineering</h1>
             <p class="tagline">Design loops, not just prompts.</p>
-            <p class="lead">545 source-audited resources, 20 operational patterns, 20 validated contracts, and 8 runtime starters for agent systems that must keep working.</p>
+            <p class="lead">579 source-audited resources from model recurrence to production operations, plus 20 patterns, 20 validated contracts, and 8 runtime starters for agent systems that must keep working.</p>
             <div class="actions">
               <a class="button primary" href="https://github.com/ChaoYue0307/awesome-loop-engineering#readme"><i class="icon" data-lucide="book-open" aria-hidden="true"></i>Explore resources</a>
               <a class="button" href="#choose"><i class="icon" data-lucide="compass" aria-hidden="true"></i>Choose a loop</a>
@@ -1410,7 +1425,7 @@
             </div>
           </figure>
           <div class="stats" role="list" aria-label="Field guide at a glance">
-            <div class="stat" role="listitem"><b><i class="icon" data-lucide="library" aria-hidden="true"></i>545</b><span>audited resources</span></div>
+            <div class="stat" role="listitem"><b><i class="icon" data-lucide="library" aria-hidden="true"></i>579</b><span>audited resources</span></div>
             <div class="stat" role="listitem"><b><i class="icon" data-lucide="blocks" aria-hidden="true"></i>20</b><span>operational patterns</span></div>
             <div class="stat" role="listitem"><b><i class="icon" data-lucide="file-check-2" aria-hidden="true"></i>20</b><span>loop contracts</span></div>
             <div class="stat" role="listitem"><b><i class="icon" data-lucide="play" aria-hidden="true"></i>8</b><span>runtime starters</span></div>
@@ -1710,9 +1725,14 @@
               <p class="section-label"><i class="icon" data-lucide="library" aria-hidden="true"></i><span>06 / Resource atlas</span></p>
               <h2>Find evidence, not just links.</h2>
             </div>
-            <p class="section-intro">Filter 545 source-audited works by goal, lifecycle, artifact type, and evidence class. Open any result for contribution, novelty, impact, provenance, and source metadata.</p>
+            <p class="section-intro">Filter 579 source-audited works by goal, loop layer, lifecycle, artifact type, and evidence class. Open any result for contribution, novelty, impact, scope, provenance, and source metadata.</p>
+          </div>
+          <p class="audit-note"><strong>Audit snapshot, 2026-07-18 UTC:</strong> all 579 rows checked; 528 public sources reachable, 3 access-restricted, 48 repository-native, and 0 broken or unreachable. <a href="https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/data/resource_source_audit.csv">Inspect the audit</a> or <a href="https://github.com/ChaoYue0307/awesome-loop-engineering/issues/new?template=annotation-correction.yml">report a correction</a>.</p>
+          <div class="loop-layer-map" aria-label="Where different loop concepts live">
+            <div class="loop-layer-item"><span>Inner loop</span><strong>Model recurrence</strong><p>A learned block refines hidden or latent state inside one inference. Included as an adjacent foundation.</p></div>
+            <div class="loop-layer-item"><span>Task loop</span><strong>Agent execution</strong><p>The model reasons, calls tools, observes results, and updates working state while solving one task.</p></div>
+            <div class="loop-layer-item"><span>Outer loop</span><strong>Operational governance</strong><p>Triggers, evidence gates, durable state, budgets, handoff, and exit govern repeated work.</p></div>
           </div>
-          <p class="audit-note"><strong>Audit snapshot, 2026-07-17 UTC:</strong> all 545 rows checked; 496 public sources reachable, 1 access-restricted, 48 repository-native, and 0 broken or unreachable. <a href="https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/data/resource_source_audit.csv">Inspect the audit</a> or <a href="https://github.com/ChaoYue0307/awesome-loop-engineering/issues/new?template=annotation-correction.yml">report a correction</a>.</p>
           <div class="resource-explorer" id="resource-explorer">
             <div class="resource-toolbar" role="search" aria-label="Search resources">
               <label>
@@ -1738,6 +1758,7 @@
                 <span class="sr-only">Lifecycle stage</span>
                 <select class="resource-select" id="resource-stage">
                   <option value="">All lifecycle stages</option>
+                  <option value="act">Model / agent action</option>
                   <option value="objective">Objective</option>
                   <option value="trigger">Trigger</option>
                   <option value="intake">Intake</option>
@@ -1752,6 +1773,19 @@
                   <option value="whole-loop">Whole loop</option>
                 </select>
               </label>
+              <label>
+                <span class="sr-only">Loop layer</span>
+                <select class="resource-select" id="resource-layer">
+                  <option value="">All loop layers</option>
+                  <option value="model">Model</option>
+                  <option value="agent">Agent</option>
+                  <option value="harness">Harness</option>
+                  <option value="workflow">Workflow</option>
+                  <option value="operations">Operations</option>
+                  <option value="evaluation">Evaluation</option>
+                  <option value="cross-layer">Cross-layer</option>
+                </select>
+              </label>
               <label>
                 <span class="sr-only">Evidence class</span>
                 <select class="resource-select" id="resource-evidence">
@@ -1777,7 +1811,7 @@
               <button class="resource-collection" type="button" data-collection="Apply" aria-pressed="false"><i class="icon" data-lucide="package-open" aria-hidden="true"></i>Apply</button>
             </div>
             <div class="resource-summary">
-              <p class="resource-count" id="resource-count" aria-live="polite">Preparing 545 resources...</p>
+              <p class="resource-count" id="resource-count" aria-live="polite">Preparing 579 resources...</p>
               <label>
                 <span class="sr-only">Sort resources</span>
                 <select class="resource-select resource-sort" id="resource-sort">
@@ -1927,7 +1961,7 @@
   title        = {Awesome Loop Engineering},
   year         = {2026},
   howpublished = {\url{https://github.com/ChaoYue0307/awesome-loop-engineering}},
-  note         = {Version 0.7.0}
+  note         = {Version 0.8.0}
 }</pre>
             </div>
           </div>
@@ -2042,6 +2076,7 @@
         var search = document.getElementById('resource-search');
         var type = document.getElementById('resource-type');
         var stage = document.getElementById('resource-stage');
+        var layer = document.getElementById('resource-layer');
         var evidence = document.getElementById('resource-evidence');
         var sort = document.getElementById('resource-sort');
         var reset = document.getElementById('resource-reset');
@@ -2068,6 +2103,30 @@
           return String(value || '').replace(/-/g, ' ');
         }
 
+        function loopLayerLabel(value) {
+          if (!value) return '';
+          return value === 'cross-layer' ? 'cross-layer' : words(value) + ' layer';
+        }
+
+        function evidenceLabel(value) {
+          var labels = {
+            'research-paper': 'Research paper',
+            'research-preprint': 'Research preprint',
+            'official-documentation': 'Official documentation',
+            'technical-documentation': 'Technical documentation',
+            'source-implementation': 'Source implementation',
+            'implementation': 'Implementation',
+            'benchmark': 'Benchmark or evaluation',
+            'repository-native': 'Validated project artifact',
+            'reusable-artifact': 'Reusable artifact',
+            'operational-pattern': 'Operational pattern',
+            'practitioner-analysis': 'Practitioner analysis',
+            'risk-analysis': 'Risk analysis',
+            'curated-index': 'Curated index'
+          };
+          return labels[value] || words(value);
+        }
+
         function stages(value) {
           return String(value || '').split(';').filter(Boolean);
         }
@@ -2134,7 +2193,10 @@
             resource.user_goal,
             resource.lifecycle_stages,
             resource.audience,
+            resource.loop_layer,
+            resource.scope_fit,
             resource.evidence_class,
+            resource.evidence_tier,
             resource.authors,
             resource.publication_year,
             resource.publication_venue,
@@ -2150,6 +2212,7 @@
           return (!query || haystack.indexOf(query) !== -1) &&
             (!type.value || resource.resource_type === type.value) &&
             (!stage.value || stages(resource.lifecycle_stages).indexOf(stage.value) !== -1) &&
+            (!layer.value || resource.loop_layer === layer.value) &&
             (!evidence.value || evidenceSet.indexOf(resource.evidence_class) !== -1) &&
             (!state.collection || resource.collection === state.collection);
         }
@@ -2183,10 +2246,12 @@
           var authors = formattedAuthors(resource.authors);
           var fullAuthors = authorNames(resource.authors).join(', ');
           var meta = [
+            resource.resource_type,
             resource.collection,
             resource.section,
-            stages(resource.lifecycle_stages).map(words).join(' / '),
-            words(resource.evidence_class)
+            loopLayerLabel(resource.loop_layer),
+            resource.scope_fit ? words(resource.scope_fit) + ' scope' : '',
+            stages(resource.lifecycle_stages).map(words).join(' / ')
           ].filter(Boolean).join(' | ');
           var signalClass = 'resource-signal ' + String(resource.signal_strength || 'unverified').toLowerCase();
           var signal = createText('span', signalClass, words(resource.signal_strength || 'unverified') + ' signal');
@@ -2232,17 +2297,24 @@
 
           evidenceCell.className = 'resource-evidence';
           appendCellLabel(evidenceCell, 'Evidence');
-          evidenceCell.appendChild(createText('span', 'resource-type-label', resource.resource_type));
+          evidenceCell.appendChild(createText('span', 'resource-type-label', 'Tier ' + (resource.evidence_tier || 'C') + ' · ' + evidenceLabel(resource.evidence_class)));
           evidenceCell.appendChild(signal);
           if (starCount) evidenceCell.appendChild(createText('span', 'resource-source-status', starCount + ' GitHub stars'));
           evidenceCell.appendChild(createText('span', 'resource-source-status', sourceStatus(resource)));
 
           details.className = 'resource-details';
-          details.appendChild(createText('summary', '', 'Contribution, novelty, impact, and source record'));
+          details.appendChild(createText('summary', '', 'Why it matters, where it fits, and source record'));
           detailGrid.className = 'resource-detail-grid';
           detailGrid.appendChild(detailSection('Contribution', [resource.key_contribution || resource.annotation]));
           detailGrid.appendChild(detailSection('Novelty', [resource.novelty]));
           detailGrid.appendChild(detailSection('Impact', [resource.impact]));
+          detailGrid.appendChild(detailSection('Use it for', [
+            resource.user_goal ? 'Goal: ' + resource.user_goal : '',
+            resource.loop_layer ? 'Loop layer: ' + words(resource.loop_layer) : '',
+            resource.scope_fit ? 'Scope fit: ' + words(resource.scope_fit) : '',
+            resource.lifecycle_stages ? 'Loop stages: ' + stages(resource.lifecycle_stages).map(words).join(', ') : '',
+            resource.audience ? 'Audience: ' + String(resource.audience).split(';').map(words).join(', ') : ''
+          ]));
           detailGrid.appendChild(detailSection('Source record', sourceRecord.concat(resource.signal ? ['Evidence: ' + resource.signal] : [])));
           details.appendChild(detailGrid);
 
@@ -2286,7 +2358,7 @@
         else compactResources.addListener(updatePageSize);
 
         search.addEventListener('input', resetLimitAndRender);
-        [type, stage, evidence, sort].forEach(function (control) {
+        [type, stage, layer, evidence, sort].forEach(function (control) {
           control.addEventListener('change', resetLimitAndRender);
         });
         collectionButtons.forEach(function (button) {
@@ -2306,6 +2378,7 @@
           search.value = '';
           type.value = '';
           stage.value = '';
+          layer.value = '';
           evidence.value = '';
           sort.value = 'curated';
           state.collection = '';
@@ -2321,7 +2394,7 @@
           loading = true;
           explorer.dataset.indexState = 'loading';
           list.setAttribute('aria-busy', 'true');
-          count.textContent = 'Loading 545 audited resources...';
+          count.textContent = 'Loading 579 audited resources...';
           list.replaceChildren(createText('p', 'resource-loading', 'Loading audited resources...'));
 
           fetch('assets/resources.json')

docs/x-v07-555.html

@@ -0,0 +1,44 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <title>Awesome Loop Engineering: 555 source-audited resources</title>
+    <meta name="description" content="A source-audited field guide with 555 resources, 20 operational patterns, 20 validated contracts, and 8 runtime starters for recurring AI-agent systems.">
+    <meta name="robots" content="noindex,follow">
+    <link rel="canonical" href="https://chaoyue0307.github.io/awesome-loop-engineering/x-v07-555.html">
+    <meta property="og:type" content="website">
+    <meta property="og:title" content="Awesome Loop Engineering">
+    <meta property="og:description" content="555 source-audited resources, 20 operational patterns, 20 validated contracts, and 8 runtime starters for recurring AI-agent systems.">
+    <meta property="og:url" content="https://chaoyue0307.github.io/awesome-loop-engineering/x-v07-555.html">
+    <meta property="og:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-555-v07.png">
+    <meta property="og:image:width" content="1280">
+    <meta property="og:image:height" content="640">
+    <meta property="og:image:alt" content="Awesome Loop Engineering: design loops, not just prompts. 555 source-audited resources, 20 patterns, 20 contracts, and 8 runtime starters.">
+    <meta name="twitter:card" content="summary_large_image">
+    <meta name="twitter:title" content="Awesome Loop Engineering">
+    <meta name="twitter:description" content="555 source-audited resources and a 20-pattern implementation kit for recurring, stateful, verified AI-agent systems.">
+    <meta name="twitter:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-555-v07.png">
+    <meta name="twitter:image:alt" content="Awesome Loop Engineering: 555 source-audited resources, 20 patterns, 20 contracts, and 8 runtime starters.">
+    <link rel="icon" href="assets/favicon-32.png" type="image/png" sizes="32x32">
+    <style>
+      :root { color-scheme: light; font-family: Inter, ui-sans-serif, system-ui, sans-serif; }
+      body { display: grid; min-height: 100vh; margin: 0; place-items: center; background: #f7f9fc; color: #111827; }
+      main { width: min(88vw, 420px); text-align: center; }
+      img { width: 72px; height: 72px; }
+      p { color: #526176; line-height: 1.6; }
+      a { color: #1d4ed8; font-weight: 700; }
+    </style>
+    <script>
+      window.location.replace("https://chaoyue0307.github.io/awesome-loop-engineering/?utm_source=x&utm_medium=social&utm_campaign=v0.7.0");
+    </script>
+  </head>
+  <body>
+    <main>
+      <img src="assets/awesome-loop-engineering-logo.png" alt="Awesome Loop Engineering">
+      <h1>Awesome Loop Engineering</h1>
+      <p>Opening the interactive Resource Atlas.</p>
+      <p><a href="https://chaoyue0307.github.io/awesome-loop-engineering/?utm_source=x&amp;utm_medium=social&amp;utm_campaign=v0.7.0">Continue to the site</a></p>
+    </main>
+  </body>
+</html>

docs/x-v08-579.html

@@ -0,0 +1,44 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <title>Awesome Loop Engineering: 579 source-audited resources</title>
+    <meta name="description" content="A source-audited field guide spanning model recurrence, agent execution, and operations, with 20 patterns, 20 contracts, and 8 runtime starters.">
+    <meta name="robots" content="noindex,follow">
+    <link rel="canonical" href="https://chaoyue0307.github.io/awesome-loop-engineering/x-v08-579.html">
+    <meta property="og:type" content="website">
+    <meta property="og:title" content="Awesome Loop Engineering">
+    <meta property="og:description" content="579 source-audited resources across the model-to-operations stack, plus a practical implementation kit for recurring AI-agent systems.">
+    <meta property="og:url" content="https://chaoyue0307.github.io/awesome-loop-engineering/x-v08-579.html">
+    <meta property="og:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-579-v08.png">
+    <meta property="og:image:width" content="1280">
+    <meta property="og:image:height" content="640">
+    <meta property="og:image:alt" content="Awesome Loop Engineering: design loops, not just prompts. 579 source-audited resources, 20 patterns, 20 contracts, and 8 runtime starters.">
+    <meta name="twitter:card" content="summary_large_image">
+    <meta name="twitter:title" content="Awesome Loop Engineering">
+    <meta name="twitter:description" content="579 source-audited resources across model, agent, harness, workflow, operations, and evaluation layers.">
+    <meta name="twitter:image" content="https://chaoyue0307.github.io/awesome-loop-engineering/assets/social-preview-579-v08.png">
+    <meta name="twitter:image:alt" content="Awesome Loop Engineering: 579 source-audited resources, 20 patterns, 20 contracts, and 8 runtime starters.">
+    <link rel="icon" href="assets/favicon-32.png" type="image/png" sizes="32x32">
+    <style>
+      :root { color-scheme: light; font-family: Inter, ui-sans-serif, system-ui, sans-serif; }
+      body { display: grid; min-height: 100vh; margin: 0; place-items: center; background: #f7f9fc; color: #111827; }
+      main { width: min(88vw, 420px); text-align: center; }
+      img { width: 72px; height: 72px; }
+      p { color: #526176; line-height: 1.6; }
+      a { color: #1d4ed8; font-weight: 700; }
+    </style>
+    <script>
+      window.location.replace("https://chaoyue0307.github.io/awesome-loop-engineering/?utm_source=x&utm_medium=social&utm_campaign=v0.8.0");
+    </script>
+  </head>
+  <body>
+    <main>
+      <img src="assets/awesome-loop-engineering-logo.png" alt="Awesome Loop Engineering">
+      <h1>Awesome Loop Engineering</h1>
+      <p>Opening the interactive Resource Atlas.</p>
+      <p><a href="https://chaoyue0307.github.io/awesome-loop-engineering/?utm_source=x&amp;utm_medium=social&amp;utm_campaign=v0.8.0">Continue to the site</a></p>
+    </main>
+  </body>
+</html>

meta/CURATION.md

@@ -51,6 +51,9 @@ The public dataset keeps those layers separate:
 
 - `annotation`, `key_contribution`, `novelty`, and `impact` are original editorial syntheses;
 - `source_title`, `source_description`, authorship, date, venue, DOI, and repository statistics come from the source audit and record their metadata provenance;
+- `evidence_tier` records whether the row is a primary source, implementation-grounded practitioner source, or curated synthesis;
+- `loop_layer` records whether recurrence lives inside the model, agent, harness, workflow, operations, evaluation, or spans layers;
+- `scope_fit` records whether a resource directly describes operational Loop Engineering, enables it, or is an explicitly adjacent foundation;
 - `signal_strength` describes the kind of evidence available, not whether the maintainer agrees with the work;
 - `source_status` records reachability at one point in time and is not a guarantee of future availability.
 

meta/DISTRIBUTION.md

@@ -8,8 +8,8 @@ Status as of 2026-07-18:
 
 | Surface | Status | Evidence / next action |
 | --- | --- | --- |
-| GitHub release | Complete | [`v0.7.0`](https://github.com/ChaoYue0307/awesome-loop-engineering/releases/tag/v0.7.0) packages 545 resources, 20 patterns and contracts, and 8 runtime starters. |
-| Release announcement | Ready | Publish the v0.7.0 summary from [`posts/launch.md`](../posts/launch.md); [Discussion #9](https://github.com/ChaoYue0307/awesome-loop-engineering/discussions/9) remains the low-noise release and contributor feed. |
+| GitHub release | Complete | [`v0.8.0`](https://github.com/ChaoYue0307/awesome-loop-engineering/releases/tag/v0.8.0) packages 579 resources, 20 patterns and contracts, 8 runtime starters, and explicit model-to-operations scope facets. |
+| Release announcement | Ready | Publish the v0.8.0 summary from [`posts/launch.md`](../posts/launch.md); [Discussion #9](https://github.com/ChaoYue0307/awesome-loop-engineering/discussions/9) remains the low-noise release and contributor feed. |
 | GitHub Explore topic | Complete | [`github/explore#5209`](https://github.com/github/explore/pull/5209) merged the `loop-engineering` topic page. |
 | Canonical Awesome directory | Policy objection | [`sindresorhus/awesome#4339`](https://github.com/sindresorhus/awesome/pull/4339) is open and lint passes, but the owner stated that AI-generated lists are not accepted. Do not treat further visual polish as a path around that policy or misrepresent automated assistance. |
 | Adjacent harness directory | In review | [`ai-boost/awesome-harness-engineering#131`](https://github.com/ai-boost/awesome-harness-engineering/pull/131) transparently proposes the project for the list's Related Awesome Lists section. |
@@ -33,7 +33,7 @@ The Hugging Face dataset [`datasets/cy0307/awesome-loop-engineering`](https://hu
 
 - The mirror tracks the full GitHub tree (docs, patterns, examples, schema, scripts).
 - The dataset mirror includes generated resource sheets at `data/resources.csv` and `data/resources.jsonl`, refreshed from the canonical English `README.md` by `scripts/export_resource_dataset.py`.
-- The HF copy of `README.md` is a focused dataset card generated from `meta/hf_card_header.yaml` and `meta/hf_card_body.md` by `scripts/build_hf_card.py`. It documents intended uses, provenance, limitations, loading examples, and the current audit snapshot without duplicating the 545-row GitHub README.
+- The HF copy of `README.md` is a focused dataset card generated from `meta/hf_card_header.yaml` and `meta/hf_card_body.md` by `scripts/build_hf_card.py`. It documents intended uses, provenance, limitations, loading examples, and the current audit snapshot without duplicating the 579-row GitHub README.
 - The YAML header is **HF-only**: it must never be added to the GitHub `README.md`, because the metadata list items break `awesome-lint`.
 - Sync uses `python3 scripts/build_hf_card.py --output <staging>/README.md` followed by `hf upload --type dataset` against a staging copy. The token lives in the local Hugging Face cache; no token is committed.
 
@@ -50,7 +50,7 @@ Use GitHub-native discovery in this order.
   ```
 
 - **Pin the repository** on your profile, and add it to your profile README (profile-level promotion).
-- **About description.** Keep this concise proof-point copy aligned with the README and project site: `🔁 545 source-audited resources • 20 patterns • 20 validated contracts • 8 runtime starters • interactive atlas + dataset for recurring, stateful, verified AI-agent systems.`
+- **About description.** Keep this concise proof-point copy aligned with the README and project site: `🔁 579 source-audited resources across model, agent, harness, workflow, operations, and evaluation layers · 20 patterns · 20 contracts · 8 runtime starters.`
 - **Custom social preview** (repo Settings, Social preview) so shared links lead with the same four durable proof points. This is UI-only; it cannot be set through the API.
 - **Releases and Discussions** provide a lower-noise subscription path. Recommend `Watch -> Custom -> Releases and Discussions`; do not ask readers to watch every repository event.
 
@@ -91,10 +91,10 @@ Use the canonical [English launch article](../posts/launch.md) or [Chinese launc
 ### Short Post
 
 ```text
-🔁 Awesome Loop Engineering v0.7.0 is live: 545 source-audited resources, 20 operational patterns, 20 validated contracts, and 8 runtime starters for recurring AI-agent systems.
+🔁 Awesome Loop Engineering v0.8.0 is live: 579 source-audited resources across the model-to-operations stack, plus 20 patterns, 20 contracts, and 8 runtime starters.
 
-🧭 Explore:
-https://chaoyue0307.github.io/awesome-loop-engineering/
+🧭 Explore the layers:
+https://chaoyue0307.github.io/awesome-loop-engineering/x-v08-579.html
 
 #AIAgents #AgentEngineering #LoopEngineering
 ```
@@ -102,17 +102,18 @@ https://chaoyue0307.github.io/awesome-loop-engineering/
 ### Longer Post
 
 ```text
-Awesome Loop Engineering v0.7.0 is live.
+Awesome Loop Engineering v0.8.0 is live.
 
-Loop Engineering governs what happens across agent runs: how work enters, agents act, evidence decides, state survives, and the system continues.
+"Loop" can mean recurrent computation inside a model, reasoning and tools inside one task, or verified work repeated across time. v0.8.0 maps all three without confusing their guarantees.
 
 The release includes:
 
-- 545 source-audited resources
+- 579 source-audited resources
+- 23 selected model-recurrence papers labeled as adjacent foundations
 - an interactive Resource Atlas
 - 20 operational patterns and 20 validated loop contracts
 - 8 runtime starters: 3 executables and 5 copy/paste templates
-- CSV and JSONL exports mirrored to Hugging Face
+- 50-field CSV, JSONL, and Parquet exports mirrored to Hugging Face
 - a public provenance standard and point-in-time source audit
 
 Explore, reuse, or correct the map:
@@ -124,11 +125,11 @@ https://github.com/ChaoYue0307/awesome-loop-engineering
 Use a regular submission, not Show HN.
 
 ```text
-Title: Awesome Loop Engineering: 545 resources and 20 loop patterns for recurring AI-agent systems
+Title: Awesome Loop Engineering: 579 resources from looped models to agent operations
 
 I have been mapping recurring AI-agent systems: how work enters, agents act, evidence gates results, state survives, and the system retries, escalates, or exits.
 
-The repository includes 545 source-audited works, 20 operational patterns, 20 validated contracts, 8 runtime starters, an interactive Resource Atlas, and a structured Hugging Face dataset. The curation and audit methods are public; corrections to the taxonomy and annotations are especially welcome.
+The repository includes 579 source-audited works across model, agent, harness, workflow, operations, and evaluation layers; 20 operational patterns; 20 validated contracts; 8 runtime starters; an interactive Resource Atlas; and a structured Hugging Face dataset. Model-level recurrence is explicitly labeled as an adjacent foundation rather than a complete operational loop. The curation and audit methods are public; corrections to the taxonomy and annotations are especially welcome.
 
 https://github.com/ChaoYue0307/awesome-loop-engineering
 ```

meta/OUTREACH.md

@@ -38,7 +38,7 @@ https://github.com/ChaoYue0307/awesome-loop-engineering/issues/new?template=anno
 ## Maintainer Request
 
 ```text
-Hi <name>, Awesome Loop Engineering is a source-audited field guide and implementation kit for recurring AI-agent systems. It includes 545 resources, 20 operational patterns, 20 validated contracts, 8 runtime starters, and a structured dataset.
+Hi <name>, Awesome Loop Engineering is a source-audited field guide and implementation kit for recurring AI-agent systems. It includes 579 resources across model-to-operations layers, 20 operational patterns, 20 validated contracts, 8 runtime starters, and a structured dataset.
 
 Would it fit the related-resources policy for <project or newsletter>? I can provide a shorter description or submit through the normal contribution process:
 https://github.com/ChaoYue0307/awesome-loop-engineering

meta/hf_card_body.md

@@ -67,6 +67,10 @@ papers = resources.filter(lambda row: row["resource_type"] == "Paper")
 verification = resources.filter(
     lambda row: "verification" in row["lifecycle_stages"].split(";")
 )
+model_recurrence = resources.filter(
+    lambda row: row["loop_layer"] == "model"
+    and row["scope_fit"] == "adjacent"
+)
 ```
 
 For pandas:
@@ -85,6 +89,7 @@ Use `url` as the durable join key. `row_id` and `source_line` are positional and
 
 - Find primary sources and implementation references for recurring agent systems.
 - Compare works by lifecycle, audience, evidence class, source type, and publication metadata.
+- Separate model, agent, harness, workflow, operations, evaluation, and cross-layer resources without treating adjacent model recurrence as a complete operational loop.
 - Build literature maps, reading lists, dashboards, or retrieval indexes.
 - Audit contribution, novelty, impact, provenance, and evidence claims.
 - Find reusable patterns, contracts, schemas, executable examples, and copy/paste runtime templates.
@@ -111,8 +116,9 @@ The primary configuration is `resources`, with one `train` split backed by the n
 | --- | --- | --- |
 | Identity | `row_id`, `title`, `url`, `canonical_url`, `resource_type`, `domain` | What the resource is and where it lives. |
 | Editorial assessment | `annotation`, `key_contribution`, `novelty`, `impact` | What the resource contributes to recurring agent systems. |
+| Scope | `loop_layer`, `scope_fit` | Where recurrence lives and whether the source is direct, enabling, or adjacent to operational Loop Engineering. |
 | Navigation | `section`, `collection`, `user_goal`, `lifecycle_stages`, `audience` | Where the resource fits and who it serves. |
-| Evidence | `evidence_class`, `signal`, `signal_strength`, `source_status`, `audited_at` | What kind of evidence or provenance is available. |
+| Evidence | `evidence_class`, `evidence_tier`, `signal`, `signal_strength`, `source_status`, `audited_at` | What kind of evidence or provenance is available. |
 | Publication | `authors`, `publication_date`, `publication_year`, `publication_venue`, `publisher`, `doi`, `arxiv_id`, `primary_category` | Bibliographic data exposed by the canonical source. |
 | Source provenance | `source_title`, `source_description`, `metadata_source`, `publication_note` | Where metadata came from and what caveats accompany it. |
 | Repository context | `github_repo`, `github_stars`, `github_forks`, `github_license`, `github_created_at`, `github_updated_at` | Point-in-time adoption and maintenance context for GitHub projects. |

meta/social-preview.html

@@ -72,14 +72,14 @@
         <p class="kicker">Source-audited field guide + implementation kit</p>
         <h1>Awesome Loop Engineering</h1>
         <p class="tagline">Design loops, not just prompts.</p>
-        <p class="support">Trigger, delegate, verify, persist, and choose the next bounded action.</p>
+        <p class="support">From model recurrence to verified agent work and bounded operations.</p>
       </div>
       <div class="mark">
         <img src="../assets/awesome-loop-engineering-logo.png" alt="">
-        <p>Recur / verify / persist</p>
+        <p>Model / agent / operations</p>
       </div>
       <div class="stats">
-        <div class="stat"><b>545</b><span>Source-audited resources</span></div>
+        <div class="stat"><b>579</b><span>Source-audited resources</span></div>
         <div class="stat"><b>20</b><span>Operational patterns</span></div>
         <div class="stat"><b>20</b><span>Validated contracts</span></div>
         <div class="stat"><b>8</b><span>Runtime starters</span></div>

posts/launch.md

@@ -1,41 +1,39 @@
-# Awesome Loop Engineering v0.7.0
+# Awesome Loop Engineering v0.8.0
 
-Awesome Loop Engineering v0.7.0 connects a recurring problem to an operational pattern, a validated contract, and a working runtime starter.
+Awesome Loop Engineering v0.8.0 maps recurrence from a model's inner computation to an agent's tools and the outer operating loop that verifies, remembers, retries, escalates, and stops.
 
-Navigate 545 source-audited works across the operating layer that discovers work, delegates it, verifies results, persists state, and decides what happens next.
+Navigate 579 source-audited works without collapsing looped models, agent execution, harnesses, workflows, and production operations into one idea.
 
 ## What You Can Use
 
-- 545 source-audited resources linked to canonical works
+- 579 source-audited resources linked to canonical works
+- 23 selected model-recurrence papers, from Universal Transformers and Huginn to LoopCoder and LoopWM, labeled as adjacent foundations rather than complete agent loops
 - 20 operational patterns organized by build, operate, optimize, and govern use cases
 - 20 schema-validated loop contracts, one for every pattern
 - 8 runtime starters: 3 dependency-light executables and 5 copy/paste runtime templates
-- an interactive Resource Atlas for filtering by goal, lifecycle stage, artifact type, and evidence class
-- CSV and JSONL exports mirrored as a Hugging Face dataset
+- an interactive Resource Atlas for filtering by goal, loop layer, lifecycle stage, artifact type, and evidence class
+- 50-field CSV, JSONL, and Parquet exports mirrored as a Hugging Face dataset
 - 8 language entry points
 
-## What Changed Since v0.6.0
+## What Changed Since v0.7.0
 
-- Added five distinct operational patterns: benchmark optimization, accessibility regression, knowledge freshness, performance regression, and authorized adversarial red teaming.
-- Added a schema-valid contract and worked scenario for every new pattern.
-- Reorganized the pattern library into four operating domains with a symptom, verified outcome, and guidance for choosing between similar loops.
-- Rebuilt the contract catalog around the questions implementers actually need: when to use a loop, what triggers it, which deterministic gate decides done, and what receipt survives.
-- Added four end-to-end worked paths for CI repair, knowledge refresh, queue processing, and read-only threshold monitoring.
-- Added two executable starters: a JSONL queue worker with idempotent state and a read-only threshold monitor with bounded polling and evidence-backed escalation.
-- Clarified that the starter library contains three executables and five copy/paste runtime templates, avoiding an unsupported claim that every artifact is a standalone program.
-- Updated the website, social preview, translations, release metadata, and Hugging Face dataset to the same counts.
+- Added a bounded Model-Level Recurrence section with 23 high-signal papers and the deeper Awesome Loop Models catalog.
+- Added `loop_layer` and `scope_fit` to every dataset row so model recurrence remains discoverable without being misclassified as operational Loop Engineering.
+- Added a model-to-operations map and loop-layer filter to the Resource Atlas.
+- Added a cross-layer research protocol for comparing internal recurrent depth with external evidence-aware retries under matched compute and cost.
+- Rechecked all 579 canonical sources and refreshed every arXiv publication decision so official conference or journal records take precedence when verified.
+- Updated the website, social preview, translations, release metadata, and Hugging Face dataset to the same counts and terminology.
 
 ## Why This Matters
 
-Evidence becomes useful when it leads to a reviewable implementation.
+"Loop" is overloaded. A learned block can recur inside one inference, an agent can alternate reasoning and tools inside one task, and an operating system can rerun verified work across time. Each layer matters, but each carries different state, stopping rules, evidence, and risks.
 
-The v0.7.0 path is explicit:
+The v0.8.0 map keeps those layers connected and comparable:
 
-1. Name the recurring problem.
-1. Choose an operational pattern.
-1. Adapt its validated contract.
-1. Select a runtime starter.
-1. Let external evidence, durable state, a hard budget, and human escalation govern the loop.
+1. Study model recurrence when the question is adaptive depth or latent computation.
+1. Study agent and harness resources when the question is reasoning, tools, context, or verification within a task.
+1. Use an operational pattern and Loop Contract when work must recur across events, sessions, or time.
+1. Let external evidence, durable state, a hard budget, and human escalation govern real-world repetition.
 
 The goal remains bounded, reviewable, evidence-driven repetition, not unlimited autonomy.
 

posts/launch.zh-CN.md

@@ -1,41 +1,39 @@
-# Awesome Loop Engineering v0.7.0
+# Awesome Loop Engineering v0.8.0
 
-Awesome Loop Engineering v0.7.0 把 implementation kit 整理成一条清晰路径：从真实的重复性问题出发，选择 pattern，改造经校验的 contract，再接入可运行的 runtime starter。
+Awesome Loop Engineering v0.8.0 建立了一张从模型内部递归、Agent 执行、Harness 到外层运行系统的完整地图。
 
-可浏览 545 条经审核资源，理解 prompt、context 与 harness engineering 之上的运行层：如何发现工作、分派任务、验证结果、持久化状态并决定下一步。
+可浏览 579 条经审核资源，同时保持清晰边界：模型内的循环提供自适应计算，但不等于具备外部验证、持久状态、预算、人类交接与退出规则的运行级 Loop。
 
 ## 可直接使用
 
-- 545 条链接到 canonical source 的审阅资源
+- 579 条链接到 canonical source 的审阅资源
+- 23 篇精选模型递归论文，覆盖 Universal Transformers、Huginn、LoopCoder、LoopWM 等方向
 - 20 个按 build、operate、optimize、govern 组织的 operational patterns
 - 20 个经过 schema 校验的 loop contracts，每个 pattern 对应一个
 - 8 个 runtime starters：3 个轻量级可执行程序与 5 个可复制改造的 runtime 模板
-- 可按目标、生命周期、资源类型和证据类别筛选的交互式 Resource Atlas
-- CSV 与 JSONL 数据导出，并同步到 Hugging Face Dataset
+- 可按目标、Loop 层级、生命周期、资源类型和证据类别筛选的交互式 Resource Atlas
+- 包含 50 个字段的 CSV、JSONL 与 Parquet 数据，并同步到 Hugging Face Dataset
 - 8 个语言入口
 
-## 相比 v0.6.0 的变化
+## 相比 v0.7.0 的变化
 
-- 新增五个边界清晰的 patterns：benchmark optimization、accessibility regression、knowledge freshness、performance regression，以及经过授权的 adversarial red teaming。
-- 为每个新 pattern 增加 schema-valid contract 与完整 use case。
-- 将 pattern library 重组为四个运行领域，并明确每个问题对应的可验证结果，以及相似 loops 之间的选择边界。
-- 将 contract catalog 改造成面向实现者的对照表：何时使用、如何触发、哪个 deterministic gate 判定完成、什么 receipt 会持久化。
-- 新增四条端到端路径，覆盖 CI repair、knowledge refresh、queue processing 与只读 threshold monitoring。
-- 新增两个可执行 starter：具备幂等状态的 JSONL queue worker，以及带有限轮询与证据升级的只读 threshold monitor。
-- 明确 starter library 由 3 个可执行程序与 5 个 runtime 模板组成，不再把每个文档都描述成独立可执行程序。
+- 新增 Model-Level Recurrence 专区，并接入 Awesome Loop Models 作为更深的架构索引。
+- 为每条数据新增 `loop_layer` 与 `scope_fit`，避免将模型递归误标为完整的运行级 Loop Engineering。
+- 在 Resource Atlas 中新增模型到运维的层级图与 Loop 层级筛选器。
+- 新增跨层评测方案，用匹配算力和成本的实验比较模型内部递归与外部证据驱动重试。
+- 重新检查全部 579 个 canonical sources，并刷新所有 arXiv 的发表状态。
 - 同步更新网站、social preview、翻译、release metadata 与 Hugging Face Dataset。
 
 ## 为什么值得关注
 
-只有当证据能导向可审阅的实现时，它才真正有用。
+“Loop”可以发生在不同层级：模型内部可以重复隐状态计算，Agent 可以在单个任务中交替调用推理与工具，运行系统则可以跨时间重复经验证的工作。它们彼此相关，但状态、停止规则、证据与风险并不相同。
 
-v0.7.0 的路径是：
+v0.8.0 将这些层级连接起来：
 
-1. 明确重复性问题。
-1. 选择 operational pattern。
-1. 改造对应的 validated contract。
-1. 选择 runtime starter。
-1. 用外部证据、持久状态、硬预算与 human escalation 管理整个 loop。
+1. 当问题是自适应深度与隐式计算时，研究模型递归。
+1. 当问题是任务内的推理、工具、上下文与验证时，研究 Agent 与 Harness。
+1. 当工作需要跨事件、会话或时间重复时，使用 operational pattern 与 Loop Contract。
+1. 用外部证据、持久状态、硬预算与 human escalation 管理真实世界中的重复执行。
 
 目标仍然不是无限自治，而是有边界、可审阅、由证据驱动的重复运行。
 

scripts/audit_resource_sources.py

@@ -58,9 +58,116 @@ PUBLISHER_BY_DOMAIN = {
     "strandsagents.com": "Strands Agents",
     "x.com": "X",
     "youtube.com": "YouTube",
+    "aclanthology.org": "Association for Computational Linguistics",
+    "iclr.cc": "International Conference on Learning Representations",
+    "icml.cc": "International Conference on Machine Learning",
+    "openreview.net": "OpenReview",
+    "papers.nips.cc": "Neural Information Processing Systems Foundation",
+    "proceedings.mlr.press": "Proceedings of Machine Learning Research",
+    "proceedings.neurips.cc": "Neural Information Processing Systems Foundation",
 }
 
 PUBLICATION_OVERRIDES = {
+    "https://openreview.net/forum?id=HyzdRiR9Y7": {
+        "authors": "Mostafa Dehghani; Stephan Gouws; Oriol Vinyals; Jakob Uszkoreit; Łukasz Kaiser",
+        "publication_date": "2019",
+        "publication_year": "2019",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "OpenReview",
+        "publication_note": "Published at ICLR 2019; venue and authors verified from the official OpenReview record.",
+        "metadata_source": "OpenReview",
+    },
+    "https://openreview.net/forum?id=HHbRxoDTxE": {
+        "authors": "Liu Yang; Kangwook Lee; Robert D. Nowak; Dimitris Papailiopoulos",
+        "publication_date": "2024",
+        "publication_year": "2024",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "OpenReview",
+        "publication_note": "Published at ICLR 2024; venue and authors verified from the official OpenReview record.",
+        "metadata_source": "OpenReview",
+    },
+    "https://iclr.cc/virtual/2025/poster/28971": {
+        "authors": "Nikunj Saunshi; Nishanth Dikkala; Zhiyuan Li; Sanjiv Kumar; Sashank J. Reddi",
+        "publication_date": "2025",
+        "publication_year": "2025",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "International Conference on Learning Representations",
+        "publication_note": "Published at ICLR 2025; metadata verified from the official conference poster page.",
+        "metadata_source": "ICLR proceedings",
+    },
+    "https://proceedings.neurips.cc/paper_files/paper/2025/hash/3b01972cf31e6fa0fe29e4b8b5c2a0a1-Abstract-Conference.html": {
+        "authors": "Jonas Geiping; Sean McLeish; Neel Jain; John Kirchenbauer; Siddharth Singh; Brian Bartoldson; Bhavya Kailkhura; Abhinav Bhatele; Tom Goldstein",
+        "publication_date": "2025",
+        "publication_year": "2025",
+        "publication_venue": "Advances in Neural Information Processing Systems 38 (NeurIPS 2025)",
+        "publisher": "Neural Information Processing Systems Foundation",
+        "publication_note": "Published at NeurIPS 2025; the proceedings page's later web timestamp is not used as the conference year.",
+        "metadata_source": "NeurIPS proceedings",
+    },
+    "https://papers.nips.cc/paper_files/paper/2025/hash/8b08bbf8b420faa6eeb4020720582ec7-Abstract-Conference.html": {
+        "authors": "Sangmin Bae; Yujin Kim; Reza Bayat; Sungnyun Kim; Jiyoun Ha; Tal Schuster; Adam Fisch; Hrayr Harutyunyan; Ziwei Ji; Aaron Courville; Se-Young Yun",
+        "publication_date": "2025",
+        "publication_year": "2025",
+        "publication_venue": "Advances in Neural Information Processing Systems 38 (NeurIPS 2025)",
+        "publisher": "Neural Information Processing Systems Foundation",
+        "publication_note": "Published at NeurIPS 2025; the proceedings page's later web timestamp is not used as the conference year.",
+        "metadata_source": "NeurIPS proceedings",
+    },
+    "https://iclr.cc/virtual/2026/poster/10009450": {
+        "authors": "Ahmadreza Jeddi; Marco Ciccone; Babak Taati",
+        "publication_date": "2026",
+        "publication_year": "2026",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "International Conference on Learning Representations",
+        "publication_note": "Published at ICLR 2026; metadata verified from the official conference poster page.",
+        "metadata_source": "ICLR proceedings",
+    },
+    "https://iclr.cc/virtual/2026/poster/10011117": {
+        "authors": "Xiaojing Zhang; Haifeng Wu; Gang He; Jiyang Shen; Bochen Lyu; Zhanxing Zhu",
+        "publication_date": "2026",
+        "publication_year": "2026",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "International Conference on Learning Representations",
+        "publication_note": "Published at ICLR 2026; metadata verified from the official conference poster page.",
+        "metadata_source": "ICLR proceedings",
+    },
+    "https://iclr.cc/virtual/2026/poster/10007767": {
+        "authors": "Yunao Zheng; Xiaojie Wang; Lei Ren; Chen Wei",
+        "publication_date": "2026",
+        "publication_year": "2026",
+        "publication_venue": "International Conference on Learning Representations (ICLR)",
+        "publisher": "International Conference on Learning Representations",
+        "publication_note": "Published at ICLR 2026; metadata verified from the official conference poster page.",
+        "metadata_source": "ICLR proceedings",
+    },
+    "https://openreview.net/forum?id=eQaJSRZiGn": {
+        "authors": "Tianyu Fu; Yichen You; Zekai Chen; Guohao Dai; Huazhong Yang; Yu Wang",
+        "publication_date": "2026",
+        "publication_year": "2026",
+        "publication_venue": "International Conference on Machine Learning (ICML)",
+        "publisher": "OpenReview",
+        "publication_note": "Published at ICML 2026; venue and authors verified from the official OpenReview record.",
+        "metadata_source": "OpenReview",
+    },
+    "https://openreview.net/forum?id=ri0LAMdhd9": {
+        "authors": "Hayden Prairie; Zachary Novack; Taylor Berg-Kirkpatrick; Daniel Y. Fu",
+        "publication_date": "2026",
+        "publication_year": "2026",
+        "publication_venue": "Learning to Iterate Workshop at ICLR 2026",
+        "publisher": "OpenReview",
+        "publication_note": "Workshop paper at the Learning to Iterate Workshop at ICLR 2026; not an ICLR main-conference paper.",
+        "metadata_source": "OpenReview",
+    },
+    "https://doi.org/10.1145/3805760.3814895": {
+        "authors": "Simarjot Khanna",
+        "publication_date": "2026-07",
+        "publication_year": "2026",
+        "publication_venue": "Proceedings of the 3rd ACM International Conference on AI-Powered Software (AIware '26)",
+        "publisher": "Association for Computing Machinery",
+        "doi": "10.1145/3805760.3814895",
+        "publication_note": "Published at AIware 2026; metadata verified from the author-supplied camera-ready paper because the DOI landing page restricted automated access.",
+        "metadata_source": "ACM DOI and camera-ready paper",
+    },
     "https://www.preprints.org/manuscript/202603.1756": {
         "authors": "Chaoyue He; Xin Zhou; Di Wang; Hong Xu; Wei Liu; Chunyan Miao",
         "publication_date": "2026-04-23",
@@ -722,6 +829,9 @@ def finalize_publication_metadata(rows: list[dict[str, str]]) -> None:
         override = PUBLICATION_OVERRIDES.get(row["url"], {})
         row.update({key: value for key, value in override.items() if value})
 
+        if row["arxiv_id"] and row["metadata_source"] == "html-meta":
+            row["metadata_source"] = "arxiv-html-meta"
+
         if row["github_repo"]:
             row["publisher"] = "GitHub"
             if row["github_repo"].lower() == PROJECT_GITHUB_REPO:
@@ -770,6 +880,7 @@ def main() -> int:
     if args.apply_publication_overlay_only:
         with AUDIT_CSV.open(encoding="utf-8", newline="") as handle:
             audited = list(csv.DictReader(handle))
+        finalize_publication_metadata(audited)
         apply_arxiv_publication_overlay(audited)
         write_audit(audited)
         resolved = sum(

scripts/check_project_consistency.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 
 import csv
 import json
+import re
 import sys
 from pathlib import Path
 
@@ -12,6 +13,35 @@ from build_hf_card import project_context, render_card
 
 
 ROOT = Path(__file__).resolve().parents[1]
+REQUIRED_RESOURCE_FIELDS = {
+    "row_id",
+    "title",
+    "url",
+    "canonical_url",
+    "resource_type",
+    "annotation",
+    "key_contribution",
+    "novelty",
+    "impact",
+    "signal",
+    "signal_strength",
+    "collection",
+    "user_goal",
+    "lifecycle_stages",
+    "audience",
+    "loop_layer",
+    "scope_fit",
+    "evidence_class",
+    "evidence_tier",
+    "source_status",
+    "metadata_source",
+    "audited_at",
+}
+ALLOWED_EVIDENCE_TIERS = {"A", "B", "C", "D"}
+ALLOWED_SIGNAL_STRENGTHS = {"high", "medium", "contextual", "unverified"}
+ALLOWED_SOURCE_STATUSES = {"ok", "local_ok", "restricted", "broken", "unreachable", "local_missing"}
+ALLOWED_LOOP_LAYERS = {"model", "agent", "harness", "workflow", "operations", "evaluation", "cross-layer"}
+ALLOWED_SCOPE_FITS = {"direct", "enabling", "adjacent"}
 
 
 def require(path: Path, snippets: list[str], failures: list[str]) -> None:
@@ -21,6 +51,58 @@ def require(path: Path, snippets: list[str], failures: list[str]) -> None:
             failures.append(f"{path.relative_to(ROOT)}: missing {snippet!r}")
 
 
+def validate_resource_rows(rows: list[dict[str, str]], failures: list[str]) -> None:
+    if not rows:
+        failures.append("data/resources.csv: no resource rows")
+        return
+
+    missing_columns = REQUIRED_RESOURCE_FIELDS - set(rows[0])
+    if missing_columns:
+        failures.append(f"data/resources.csv: missing columns {sorted(missing_columns)}")
+
+    seen_urls: set[str] = set()
+    seen_ids: set[str] = set()
+    for index, row in enumerate(rows, 1):
+        row_label = row.get("row_id") or f"row {index}"
+        missing = sorted(field for field in REQUIRED_RESOURCE_FIELDS if not row.get(field, "").strip())
+        if missing:
+            failures.append(f"data/resources.csv: {row_label} missing required values {missing}")
+
+        normalized_url = row.get("url", "").strip().lower().rstrip("/")
+        if normalized_url in seen_urls:
+            failures.append(f"data/resources.csv: duplicate URL at {row_label}: {row.get('url', '')}")
+        seen_urls.add(normalized_url)
+
+        row_id = row.get("row_id", "")
+        if row_id in seen_ids:
+            failures.append(f"data/resources.csv: duplicate row_id {row_id}")
+        seen_ids.add(row_id)
+        expected_id = f"ale-{index:04d}"
+        if row_id != expected_id:
+            failures.append(f"data/resources.csv: expected {expected_id}, found {row_id or '<blank>'}")
+
+        if row.get("evidence_tier") not in ALLOWED_EVIDENCE_TIERS:
+            failures.append(f"data/resources.csv: {row_label} has invalid evidence_tier {row.get('evidence_tier')!r}")
+        if row.get("signal_strength") not in ALLOWED_SIGNAL_STRENGTHS:
+            failures.append(f"data/resources.csv: {row_label} has invalid signal_strength {row.get('signal_strength')!r}")
+        if row.get("source_status") not in ALLOWED_SOURCE_STATUSES:
+            failures.append(f"data/resources.csv: {row_label} has invalid source_status {row.get('source_status')!r}")
+        if row.get("loop_layer") not in ALLOWED_LOOP_LAYERS:
+            failures.append(f"data/resources.csv: {row_label} has invalid loop_layer {row.get('loop_layer')!r}")
+        if row.get("scope_fit") not in ALLOWED_SCOPE_FITS:
+            failures.append(f"data/resources.csv: {row_label} has invalid scope_fit {row.get('scope_fit')!r}")
+
+        year = row.get("publication_year", "")
+        if year and not re.fullmatch(r"(?:19|20)\d{2}", year):
+            failures.append(f"data/resources.csv: {row_label} has invalid publication_year {year!r}")
+        if row.get("resource_type") == "Paper":
+            for field in ("authors", "publication_year"):
+                if not row.get(field, "").strip():
+                    failures.append(f"data/resources.csv: {row_label} paper is missing {field}")
+        if "arxiv.org" in row.get("url", "") and not row.get("arxiv_id", "").strip():
+            failures.append(f"data/resources.csv: {row_label} arXiv work is missing arxiv_id")
+
+
 def main() -> int:
     context = project_context()
     count = context["RESOURCE_COUNT"]
@@ -55,9 +137,11 @@ def main() -> int:
         require(translation, [count], failures)
 
     with (ROOT / "data" / "resources.csv").open(encoding="utf-8", newline="") as handle:
-        csv_count = sum(1 for _ in csv.DictReader(handle))
+        resource_rows = list(csv.DictReader(handle))
+    csv_count = len(resource_rows)
     if csv_count != int(count):
         failures.append(f"data/resources.csv: expected {count} rows, found {csv_count}")
+    validate_resource_rows(resource_rows, failures)
 
     site_payload = json.loads((ROOT / "docs" / "assets" / "resources.json").read_text(encoding="utf-8"))
     if site_payload.get("count") != int(count) or len(site_payload.get("resources", [])) != int(count):

scripts/check_publication_metadata.py

@@ -94,8 +94,8 @@ def main() -> int:
             if decision["status"] == "preprint-only":
                 if row["publisher"] != "arXiv":
                     failures.append(f"{row_id}: preprint-only row has publisher {row['publisher']!r}")
-                if row["metadata_source"] != "arxiv-api":
-                    failures.append(f"{row_id}: preprint metadata is not arXiv API-backed")
+                if row["metadata_source"] not in {"arxiv-api", "arxiv-html-meta"}:
+                    failures.append(f"{row_id}: preprint metadata is not backed by a primary arXiv source")
                 expected_evidence = "benchmark" if row["resource_type"] == "Benchmark" else "research-preprint"
                 if row["evidence_class"] != expected_evidence:
                     failures.append(f"{row_id}: preprint-only row has evidence class {row['evidence_class']!r}")

scripts/export_resource_dataset.py

@@ -39,6 +39,11 @@ ENTRY_RE = re.compile(
     r"\[(?P<title>[^\]]+)\]\((?P<url>[^)]+)\) - (?P<annotation>.+)$"
 )
 TABLE_ENTRY_RE = re.compile(
+    r"^\| (?P<marker>\S+) \*\*\[(?P<title>[^\]]+)\]\((?P<url>[^)]+)\)\*\*"
+    r"<br><sub>(?P<resource_type>[^<]+)</sub>\s+\| (?P<metadata>.*?) \| "
+    r"(?P<annotation>.*?) \| (?P<table_evidence>.+) \|$"
+)
+LEGACY_TABLE_ENTRY_RE = re.compile(
     r"^\| (?P<marker>\S+) \*\*\[(?P<title>[^\]]+)\]\((?P<url>[^)]+)\)\*\*"
     r"<br><sub>(?P<resource_type>[^<]+)</sub>\s+\| (?P<metadata>.*?) \| "
     r"(?P<annotation>.+) \|$"
@@ -58,6 +63,18 @@ TYPE_MARKERS = {
     "Critique": "⚠️",
 }
 
+
+def base_resource_type(value: str) -> str:
+    cleaned = clean(value)
+    return next(
+        (
+            resource_type
+            for resource_type in TYPE_MARKERS
+            if cleaned == resource_type or cleaned.startswith(f"{resource_type} ·")
+        ),
+        cleaned,
+    )
+
 FIELDS = [
     "row_id",
     "section",
@@ -84,7 +101,10 @@ FIELDS = [
     "user_goal",
     "lifecycle_stages",
     "audience",
+    "loop_layer",
+    "scope_fit",
     "evidence_class",
+    "evidence_tier",
     "source_status",
     "canonical_url",
     "source_title",
@@ -125,7 +145,10 @@ SITE_FIELDS = [
     "section_slug",
     "lifecycle_stages",
     "audience",
+    "loop_layer",
+    "scope_fit",
     "evidence_class",
+    "evidence_tier",
     "signal_strength",
     "source_status",
     "authors",
@@ -147,6 +170,7 @@ COLLECTIONS = {
     "Concept Guides": ("Learn", "Understand the field and its boundaries."),
     "Start Here": ("Learn", "Understand the field and its boundaries."),
     "Research Foundations": ("Learn", "Understand the field and its boundaries."),
+    "Model-Level Recurrence": ("Learn", "Understand how recurrent model computation can power, but not replace, a governed agent loop."),
     "Pattern Library": ("Design", "Specify a loop contract and operating pattern."),
     "Core Loop Primitives": ("Design", "Specify a loop contract and operating pattern."),
     "Agent Workflow Patterns": ("Design", "Specify a loop contract and operating pattern."),
@@ -182,6 +206,7 @@ SECTION_STAGE_DEFAULTS = {
     "Start Here": ["whole-loop"],
     "Pattern Library": ["whole-loop"],
     "Research Foundations": ["whole-loop"],
+    "Model-Level Recurrence": ["act"],
     "Official Runtime Guides": ["workspace", "context", "delegation", "state"],
     "Agent Workflow Patterns": ["delegation", "verification"],
     "Coding-Agent Loop Systems": ["workspace", "delegation", "verification", "state"],
@@ -233,6 +258,7 @@ SECTION_IMPACT = {
     "Core Loop Primitives": "Turns the concept into concrete loop mechanics: triggers, state, tools, worktrees, permissions, and recurring execution.",
     "Official Runtime Guides": "Anchors implementation choices in primary vendor and framework documentation instead of second-hand summaries.",
     "Research Foundations": "Connects Loop Engineering to prior work on agent loops, planning, reflection, feedback, and long-horizon autonomy.",
+    "Model-Level Recurrence": "Explains the inner recurrent computation that can improve an agent's model while remaining distinct from the outer operating loop.",
     "Agent Workflow Patterns": "Shows reusable architecture patterns that compose agents, evaluators, workers, and durable workflow control.",
     "Coding-Agent Loop Systems": "Grounds the practice in real coding-agent systems, bare loops, orchestration tools, and long-running software tasks.",
     "Verification And Feedback Gates": "Identifies the feedback signals that make recurring agent work measurable, retryable, and safe to stop.",
@@ -271,8 +297,52 @@ SECTION_NOVELTY = {
     "Pattern Library": "Turns common recurring-agent jobs into named patterns with gates, budgets, and escalation paths.",
     "Critiques, Risks, And Limitations": "Keeps adoption grounded in known failure modes, economics, and operational limits.",
     "Adjacent Awesome Lists": "Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept.",
+    "Model-Level Recurrence": "Separates shared-block latent iteration inside one model inference from repeated, externally governed agent work.",
+}
+
+SECTION_LOOP_LAYERS = {
+    "Concept Guides": "cross-layer",
+    "Start Here": "cross-layer",
+    "Pattern Library": "workflow",
+    "Core Loop Primitives": "workflow",
+    "Official Runtime Guides": "harness",
+    "Research Foundations": "cross-layer",
+    "Model-Level Recurrence": "model",
+    "Agent Workflow Patterns": "workflow",
+    "Coding-Agent Loop Systems": "agent",
+    "Verification And Feedback Gates": "harness",
+    "Securing Unattended Loops": "operations",
+    "State, Memory, And Context Persistence": "harness",
+    "Orchestration And Multi-Agent Delegation": "workflow",
+    "Benchmarks And Evaluation": "evaluation",
+    "Operations Playbooks": "operations",
+    "Templates And Patterns": "workflow",
+    "Examples And Schema": "workflow",
+    "Community Gallery": "operations",
+    "Critiques, Risks, And Limitations": "cross-layer",
+    "Adjacent Awesome Lists": "cross-layer",
+    "Explore And Reuse": "cross-layer",
+    "Shape What Comes Next": "cross-layer",
+}
+
+TITLE_LOOP_LAYERS = {
+    "Awesome Loop Models": "model",
+}
+
+DIRECT_SCOPE_SECTIONS = {
+    "Concept Guides",
+    "Start Here",
+    "Pattern Library",
+    "Core Loop Primitives",
+    "Coding-Agent Loop Systems",
+    "Operations Playbooks",
+    "Templates And Patterns",
+    "Examples And Schema",
+    "Community Gallery",
 }
 
+ADJACENT_SCOPE_SECTIONS = {"Model-Level Recurrence", "Adjacent Awesome Lists"}
+
 TYPE_SIGNAL = {
     "Paper": ("Research paper or preprint; strongest signal when the entry contributes a method, benchmark, measurement, or formal framing.", "high"),
     "Docs": ("Primary documentation from a platform, SDK, standard, or framework; strong implementation signal.", "high"),
@@ -285,6 +355,23 @@ TYPE_SIGNAL = {
     "List": ("Adjacent curated collection; signal comes from ecosystem coverage rather than a single technical claim.", "contextual"),
 }
 
+EVIDENCE_TIERS = {
+    "research-paper": "A",
+    "research-preprint": "A",
+    "official-documentation": "A",
+    "technical-documentation": "A",
+    "source-implementation": "A",
+    "implementation": "A",
+    "benchmark": "A",
+    "repository-native": "A",
+    "reusable-artifact": "A",
+    "operational-pattern": "B",
+    "practitioner-analysis": "B",
+    "risk-analysis": "B",
+    "curated-index": "C",
+    "curated-source": "C",
+}
+
 NOVELTY_RULES = [
     (r"\bofficial\b|\bprimary-source\b", "Primary-source operational guidance rather than commentary."),
     (r"\bfuture directions?\b|\bresearch agenda\b", "Turns open gaps into measurable research, infrastructure, and product directions."),
@@ -326,7 +413,7 @@ def parse_entry_line(raw_line: str) -> dict[str, str] | None:
     if match:
         return match.groupdict()
 
-    match = TABLE_ENTRY_RE.match(raw_line)
+    match = TABLE_ENTRY_RE.match(raw_line) or LEGACY_TABLE_ENTRY_RE.match(raw_line)
     if not match:
         return None
 
@@ -388,6 +475,10 @@ def novelty(section: str, title: str, annotation: str) -> str:
         lens = "Makes an otherwise informal practice concrete and reusable."
         return f"{lens} {key_contribution(annotation)}"
 
+    if section == "Model-Level Recurrence":
+        lens = "Reuses learned computation inside one model inference rather than repeating a full agent run."
+        return f"{lens} {key_contribution(annotation)}"
+
     text = f"{title} {annotation}".lower()
     for pattern, phrase in NOVELTY_RULES:
         if re.search(pattern, text):
@@ -409,6 +500,8 @@ def collection_for(section: str) -> tuple[str, str]:
 def lifecycle_stages(section: str, title: str, annotation: str) -> str:
     text = f"{title} {annotation}".lower()
     stages = [stage for stage, pattern in STAGE_RULES if re.search(pattern, text)]
+    if section == "Model-Level Recurrence" and "act" not in stages:
+        stages.insert(0, "act")
     if not stages:
         stages = SECTION_STAGE_DEFAULTS.get(section, ["whole-loop"])
     return ";".join(dict.fromkeys(stages))
@@ -422,6 +515,8 @@ def audience_for(section: str, resource_type: str) -> str:
         audiences.append("builder")
     if resource_type in {"Paper", "Benchmark"}:
         audiences.extend(["researcher", "evaluator"])
+    if section == "Model-Level Recurrence":
+        audiences.extend(["model-builder", "agent-builder"])
     if section in {"Securing Unattended Loops", "Operations Playbooks", "Critiques, Risks, And Limitations"}:
         audiences.extend(["operator", "security"])
     if section in {"Verification And Feedback Gates", "Benchmarks And Evaluation"}:
@@ -462,11 +557,29 @@ def evidence_class(
     }.get(resource_type, "curated-source")
 
 
-def impact(collection: str, title: str) -> str:
+def loop_layer(section: str, title: str) -> str:
+    return TITLE_LOOP_LAYERS.get(title, SECTION_LOOP_LAYERS.get(section, "cross-layer"))
+
+
+def scope_fit(section: str) -> str:
+    if section in ADJACENT_SCOPE_SECTIONS:
+        return "adjacent"
+    if section in DIRECT_SCOPE_SECTIONS:
+        return "direct"
+    return "enabling"
+
+
+def impact(collection: str, section: str, title: str) -> str:
+    if section == "Model-Level Recurrence":
+        return f"Use {title} to assess inner latent computation as a model capability inside a separately governed agent loop."
     goal = COLLECTION_IMPACT.get(collection, "apply the source to a recurring agent system")
     return f"Use {title} to {goal}."
 
 
+def evidence_tier(evidence: str) -> str:
+    return EVIDENCE_TIERS.get(evidence, "C")
+
+
 def format_count(value: str) -> str:
     try:
         return f"{int(value):,}"
@@ -523,8 +636,9 @@ def iter_rows(readme_path: Path = README) -> list[dict[str, str]]:
     for line_number, raw_line in enumerate(readme_path.read_text(encoding="utf-8").splitlines(), 1):
         heading = HEADING_RE.match(raw_line)
         if heading:
-            section = clean(heading.group("title"))
-            section_slug = slugify(section)
+            if heading.group("level") == "##":
+                section = clean(heading.group("title"))
+                section_slug = slugify(section)
             continue
 
         entry = parse_entry_line(raw_line)
@@ -537,7 +651,7 @@ def iter_rows(readme_path: Path = README) -> list[dict[str, str]]:
 
         url_kind, domain = classify_url(url)
         annotation = clean(entry["annotation"])
-        resource_type = clean(entry["resource_type"])
+        resource_type = base_resource_type(entry["resource_type"])
         row_number = len(rows) + 1
         row_id = f"ale-{row_number:04d}"
         collection, user_goal = collection_for(section)
@@ -560,7 +674,7 @@ def iter_rows(readme_path: Path = README) -> list[dict[str, str]]:
                 "description": annotation,
                 "key_contribution": key_contribution(annotation),
                 "novelty": novelty(section, clean(entry["title"]), annotation),
-                "impact": impact(collection, clean(entry["title"])),
+                "impact": impact(collection, section, clean(entry["title"])),
                 "signal": signal_text,
                 "signal_strength": signal_strength,
                 "source_readme": "README.md",
@@ -572,7 +686,10 @@ def iter_rows(readme_path: Path = README) -> list[dict[str, str]]:
                 "user_goal": user_goal,
                 "lifecycle_stages": lifecycle_stages(section, clean(entry["title"]), annotation),
                 "audience": audience_for(section, resource_type),
+                "loop_layer": loop_layer(section, clean(entry["title"])),
+                "scope_fit": scope_fit(section),
                 "evidence_class": evidence,
+                "evidence_tier": evidence_tier(evidence),
                 "source_status": audit.get("audit_status", "not-audited"),
                 "canonical_url": audit.get("canonical_url", "") or audit.get("final_url", "") or url,
                 "source_title": audit.get("source_title", ""),

scripts/render_readme_tables.py

@@ -11,8 +11,12 @@ from pathlib import Path
 from export_resource_dataset import README, TYPE_MARKERS, iter_rows
 
 
-TABLE_HEADERS = ("Resource", "Published at", "Key feature")
-LEGACY_TABLE_HEADERS = {TABLE_HEADERS, ("Resource", "Publication / source", "Key feature")}
+TABLE_HEADERS = ("Resource", "Published at", "Contribution", "Evidence")
+LEGACY_TABLE_HEADERS = {
+    TABLE_HEADERS,
+    ("Resource", "Published at", "Key feature"),
+    ("Resource", "Publication / source", "Key feature"),
+}
 SUMMARY_START = "<!-- resource-type-summary:start -->"
 SUMMARY_END = "<!-- resource-type-summary:end -->"
 PROJECT_GITHUB_REPO = "chaoyue0307/awesome-loop-engineering"
@@ -34,6 +38,40 @@ GITHUB_OWNER_LABELS = {
     "vercel": "Vercel",
 }
 
+EVIDENCE_LABELS = {
+    "research-paper": "Research paper",
+    "research-preprint": "Research preprint",
+    "official-documentation": "Official documentation",
+    "technical-documentation": "Technical documentation",
+    "source-implementation": "Source implementation",
+    "implementation": "Implementation",
+    "benchmark": "Benchmark or evaluation",
+    "repository-native": "Validated project artifact",
+    "reusable-artifact": "Reusable artifact",
+    "operational-pattern": "Operational pattern",
+    "practitioner-analysis": "Practitioner analysis",
+    "risk-analysis": "Risk analysis",
+    "curated-index": "Curated index",
+    "curated-source": "Curated source",
+}
+
+EVIDENCE_NOTES = {
+    "research-paper": "Published or accepted research record",
+    "research-preprint": "Preprint; inspect methods and evaluation",
+    "official-documentation": "Primary product or standard behavior",
+    "technical-documentation": "Technical reference from the source",
+    "source-implementation": "Inspectable source and runtime behavior",
+    "implementation": "Working implementation or runtime",
+    "benchmark": "Repeatable tasks, scores, or evaluation data",
+    "repository-native": "Maintained with project validation checks",
+    "reusable-artifact": "Adaptable template, schema, or guide",
+    "operational-pattern": "Transferable operating practice",
+    "practitioner-analysis": "Experience-backed implementation context",
+    "risk-analysis": "Failure modes, limits, or adoption cautions",
+    "curated-index": "Ecosystem coverage and discovery",
+    "curated-source": "Curated context for comparison",
+}
+
 
 def escape_cell(value: str) -> str:
     return value.replace("|", r"\|")
@@ -115,14 +153,29 @@ def publication_cell(row: dict[str, str]) -> str:
     return primary
 
 
-def resource_cells(row: dict[str, str]) -> tuple[str, str, str]:
+def evidence_cell(row: dict[str, str]) -> str:
+    evidence = row["evidence_class"]
+    tier = row["evidence_tier"]
+    label = EVIDENCE_LABELS.get(evidence, evidence.replace("-", " ").title())
+    note = EVIDENCE_NOTES.get(evidence, "Inspect the linked source")
+    if row["source_status"] == "restricted":
+        note = "Access restricted during the latest source audit"
+    elif row["source_status"] not in {"ok", "local_ok"}:
+        note = "Availability was not verified in the latest audit"
+    return f"**Tier {escape_cell(tier)}** · {escape_cell(label)}<br><sub>{escape_cell(note)}</sub>"
+
+
+def resource_cells(row: dict[str, str]) -> tuple[str, str, str, str]:
     title = escape_cell(row["title"])
     url = escape_cell(row["url"])
     annotation = escape_cell(row["annotation"])
     marker = escape_cell(row["marker"])
     resource_type = escape_cell(row["resource_type"])
-    resource = f"{marker} **[{title}]({url})**<br><sub>{resource_type}</sub>"
-    return resource, publication_cell(row), annotation
+    subtype = resource_type
+    if row.get("section") == "Model-Level Recurrence":
+        subtype = f"{resource_type} · Model layer · Adjacent foundation"
+    resource = f"{marker} **[{title}]({url})**<br><sub>{escape_cell(subtype)}</sub>"
+    return resource, publication_cell(row), annotation, evidence_cell(row)
 
 
 def markdown_table(headers: tuple[str, ...], data: list[tuple[str, ...]]) -> list[str]:
@@ -157,7 +210,8 @@ def is_table_separator(line: str) -> bool:
     if not line.startswith("|"):
         return False
     cells = tuple(cell.strip() for cell in line.strip().strip("|").split("|"))
-    return len(cells) == len(TABLE_HEADERS) and all(
+    valid_lengths = {len(headers) for headers in LEGACY_TABLE_HEADERS}
+    return len(cells) in valid_lengths and all(
         len(cell) >= 3 and not cell.strip("-") for cell in cells
     )
 

scripts/resolve_arxiv_publications.py

@@ -157,7 +157,13 @@ def json_request(
         try:
             with urllib.request.urlopen(request, timeout=timeout, context=context) as response:
                 return json.loads(response.read().decode("utf-8"))
-        except (urllib.error.URLError, urllib.error.HTTPError, TimeoutError, json.JSONDecodeError):
+        except (
+            urllib.error.URLError,
+            urllib.error.HTTPError,
+            ConnectionError,
+            TimeoutError,
+            json.JSONDecodeError,
+        ):
             if attempt == attempts:
                 return {}
             time.sleep(attempt)