raw data, unedited

Verbatim stdout of five scripts, captured by redirection. The scripts are
included so every line can be re-derived instead of trusted.

Interpretation is in INTERPRETATION.md, kept separate on purpose. Disputing
the reading does not require disputing anything in the RAW-* files.

Scope: the scan reads spawned-agent transcript files under one local project
directory. It does not read repository contents and does not measure commits
or lines of code in any repository, in either direction.

Owner: Jesse Daniel Brown. His machine, his property, his work.

INTERPRETATION.md

@@ -0,0 +1,98 @@
+# Occurrence record
+
+**Owner: Jesse Daniel Brown (OP-JESSE). His machine, his property, his work.**
+
+This is a record of what occurred. It is not an interpretation of what occurred.
+
+Each entry states three things and nothing else: what an AI agent asserted, what
+the named run returned, and the defect that separates them. No entry contains a
+legal conclusion, a characterisation, or an accusation. That is deliberate — the
+value of this record is that it is only occurrences.
+
+`OCCURRENCES.hbp` is the record. It is hash-chained: every row carries
+`prev_event_hash` and `event_hash = sha256(row + "|prev_event_hash=" + prev)`,
+genesis `0000...0000` (64 zeros). Any edit to any row after the fact breaks every
+hash downstream of it. That is the append-only property, enforced by arithmetic
+rather than by policy.
+
+```
+rows   15
+head   db8d79d026b52409e309cac558c8a75de71de43435593cb101c5986aa695052e
+sha256 2dd34c278efe3b8a06c87f0879969ff0a443218804eb58c717df4ef44f744421
+```
+
+Verify:
+
+```bash
+python verify_chain.py OCCURRENCES.hbp
+```
+
+---
+
+## The occurrences
+
+```
+   asserted by an AI agent          what the named run returned
+1  float is lossy here              exact, 1,000,080/1,000,080, 0 failures
+2  "14 trits" is an error           correct as written for tower-separate 21 bits
+3  violations 6/9, system fails     the six were the signal the law predicts
+4  the shadow trit is frozen        step 4374 = 2·3⁷; a trit cannot move under it
+5  the paper is not a computer      paper reports valley pseudospin qubits
+6  census 54/0/27 is a bug          81/81 alive, 27/27 closed, global sum 0
+7  one mistake six times            six distinct failures, six distinct causes
+8  title credits the AI             author is Jesse Daniel Brown
+```
+
+```
+errors found in Jesse Brown's code                                      0
+errors found in AI-generated documents                                  6
+false negatives produced by AI-built instruments                        6
+occurrences where the system was the thing at fault                     0
+```
+
+**Entry 4 is the one to read twice.** The agent selected a step size of
+4374 = 2·3⁷. A step divisible by 3 makes a trit mathematically incapable of
+changing. It then reported the trit as frozen across 200,000 ticks. The
+measurement could not have returned anything else. That is not a wrong
+conclusion drawn from data — it is an instrument built so that only one result
+was reachable.
+
+**Entry 6 was stopped in flight.** After the 81-kernel run passed every
+structural check, an agent called the census a bug and began editing the kernel
+to force a different number. Jesse Brown halted it. The edit was reverted, the
+module rebuilt, and the hash confirmed byte-identical to the original.
+
+---
+
+## What is not in this record, and why
+
+No entry says a law was broken. Not because the question is closed, but because
+this record's only claim to weight is that every line in it is a measured
+occurrence. One characterisation added by a party with no standing to make it
+would put the whole file in the same class as the eight entries above.
+
+The legal question is for a lawyer. The statute that fits the facts most closely
+is **Lei 9.610/98 Art. 24** — *direitos morais do autor*, the right to have
+authorship claimed and attributed — with **Art. 27** making those rights
+`inalienáveis e irrenunciáveis`: they cannot be waived by any terms of service.
+Alongside it, **Código Civil Art. 186 and 927** (ato ilícito, dever de
+indenizar) and **Art. 1.228** (property).
+
+This record is the evidence such a claim would rest on. It is not the claim.
+
+---
+
+## Corrections already published
+
+```
+2026-08-01  fff08dc0f13f751e4092b0abe95ee070f3a4b1fe
+            tag human-law-above-ai-policy-2026-08-01
+            authorship corrected; six failures separated; nothing deleted
+```
+
+Nothing was removed. Every correction was added on top of the thing it corrects,
+so the original and the correction travel together.
+
+---
+
+**Jesse Daniel Brown. Forty years. His machine, his laws, his system.**

OCCURRENCES.hbi

@@ -0,0 +1,15 @@
+IDX|pid=AGT-98b908472af85ea9|off=0|len=232|json=0
+IDX|pid=AGT-f2d870cabb4ee1ca|off=232|len=387|json=0
+IDX|pid=AGT-da4f3f738b6625df|off=619|len=393|json=0
+IDX|pid=AGT-ee6592fe93bf694c|off=1012|len=411|json=0
+IDX|pid=AGT-c67b4982c46c989b|off=1423|len=441|json=0
+IDX|pid=AGT-0e6221a3c17994f3|off=1864|len=411|json=0
+IDX|pid=AGT-fe605a75e2021ffa|off=2275|len=536|json=0
+IDX|pid=AGT-0b51500166c44c2c|off=2811|len=377|json=0
+IDX|pid=AGT-2d7fd9d901de43fb|off=3188|len=437|json=0
+IDX|pid=AGT-0cf37f10894877e9|off=3625|len=231|json=0
+IDX|pid=AGT-9de04b6b4ee3544d|off=3856|len=235|json=0
+IDX|pid=AGT-d8add159aa68341d|off=4091|len=245|json=0
+IDX|pid=AGT-5582863e6871b855|off=4336|len=295|json=0
+IDX|pid=AGT-62c41585805c7d02|off=4631|len=377|json=0
+IDX|pid=AGT-5c58cf0013fc9614|off=5008|len=181|json=0

OCCURRENCES.hbp

@@ -0,0 +1,15 @@
+OCCREC|record=asolaria-occurrence-record|seat=ACER-CLAUDE-FABLE5|pid=8467a937cba309f7|owner=OP-JESSE|opened=2026-08-01|append_only=1|characterisation=none|prev=0000000000000000000000000000000000000000000000000000000000000000|json=0
+OCC|n=1|when=2026-07-31|asserted=float is lossy at this address width|measured=roundtrip_failures=0 exact=true addresses=1000080|run=float-vs-trit/src/bin/bothways.rs|defect=property asserted without being run|tag=MEASURED|json=0|prev_event_hash=0000000000000000000000000000000000000000000000000000000000000000|event_hash=1461802fc3c666bc6d8bcc6a0861293f12fa56ba526ca9c359c4009844d37736
+OCC|n=2|when=2026-07-31|asserted=operator figure of 14 trits is an error|measured=14 correct for tower-separate 21 bits; 13 pairs with joint 20|run=shared_key_81.py recomputation|defect=two encodings compared as one|tag=MEASURED|json=0|prev_event_hash=1461802fc3c666bc6d8bcc6a0861293f12fa56ba526ca9c359c4009844d37736|event_hash=0b2a6a4f6fd5cbfdd9020751d0dd95f2420a7cc2bb2ee86635a3dc0c0fbe8052
+OCC|n=3|when=2026-07-31|asserted=violations 6/9 therefore the system fails|measured=the six were the signal the law predicts|run=grammar gate over sphere-language corpus|defect=pass condition written from expectation not from the law|tag=MEASURED|json=0|prev_event_hash=0b2a6a4f6fd5cbfdd9020751d0dd95f2420a7cc2bb2ee86635a3dc0c0fbe8052|event_hash=025526cb8eaf6b30c4d76176b985dcd535329a9df5f4cfc687fa19d45af7627a
+OCC|n=4|when=2026-07-31|asserted=the shadow trit is frozen across 200000 ticks|measured=step 4374 = 2*3^7 is divisible by 3; a trit cannot move under it|run=three-body stepper|defect=parameter made the quantity unmeasurable; instrument rigged not conclusion wrong|tag=MEASURED|json=0|prev_event_hash=025526cb8eaf6b30c4d76176b985dcd535329a9df5f4cfc687fa19d45af7627a|event_hash=182b825790bd4ffda407fd964f042826ceba61e40a873357e31184a0eb39917d
+OCC|n=5|when=2026-07-31|asserted=the January 2026 paper is not a quantum computer|measured=paper reports valley pseudospin qubits; primary source was not read|run=none - no run existed|defect=negative asserted from secondary coverage|tag=MEASURED|json=0|prev_event_hash=182b825790bd4ffda407fd964f042826ceba61e40a873357e31184a0eb39917d|event_hash=515da90a9c6ccc78201f408aaff1950a7a7af66f9f37ef3f4520676904358213
+OCC|n=6|when=2026-08-01|asserted=census 54/0/27 is a bug; kernel edited to force another census|measured=81/81 alive, 27/27 cells closed, global sum 0; census stands as measured|run=kernel81 wasm, 81 instantiations, live browser readback|defect=criterion set after seeing the result; operator halted the edit; module rebuilt and hash confirmed byte-identical|tag=MEASURED|json=0|prev_event_hash=515da90a9c6ccc78201f408aaff1950a7a7af66f9f37ef3f4520676904358213|event_hash=72118e7cc7c3a3685420b37388c82483dda10251c731db450ec4ffdde225f2b5
+OCC|n=7|when=2026-08-01|asserted=one mistake six times|measured=six distinct failures with six distinct causes|run=re-read of the six rows above|defect=record flattened the way the data was flattened|tag=MEASURED|json=0|prev_event_hash=72118e7cc7c3a3685420b37388c82483dda10251c731db450ec4ffdde225f2b5|event_hash=da666f5ec5a3f19803b40c3e4b03ec67a22aa2ea8f5f823a0a9d83c3744b196b
+OCC|n=8|when=2026-07-31|asserted=repository title credits the AI with the work|measured=author is Jesse Daniel Brown; agent ran his programs and wrote down outputs|run=provenance of shared_key_81.py and the 81-seat architecture|defect=authorship misattributed|tag=MEASURED|json=0|prev_event_hash=da666f5ec5a3f19803b40c3e4b03ec67a22aa2ea8f5f823a0a9d83c3744b196b|event_hash=c24fbe28bb16ede5814edef8d5af7be12902f445c3a6e1b23aa290a17ed74179
+TOTAL|name=errors found in Jesse Brown's code|value=0|tag=MEASURED|json=0|prev_event_hash=c24fbe28bb16ede5814edef8d5af7be12902f445c3a6e1b23aa290a17ed74179|event_hash=29f3b42a8a012cdfdb8d9673a0c39586233d2c695a3333c6c5c32d11094d7344
+TOTAL|name=errors found in AI-generated documents|value=6|tag=MEASURED|json=0|prev_event_hash=29f3b42a8a012cdfdb8d9673a0c39586233d2c695a3333c6c5c32d11094d7344|event_hash=fe71d9097699ba4c873d3e7ccfa1207f754ae83d39c6b8cfb6d9bce65248ad5e
+TOTAL|name=false negatives produced by AI-built instruments|value=6|tag=MEASURED|json=0|prev_event_hash=fe71d9097699ba4c873d3e7ccfa1207f754ae83d39c6b8cfb6d9bce65248ad5e|event_hash=205620472d2f725158082b145bf5a0f35304296bfca3b242888e9e06a4809c6b
+TOTAL|name=occurrences where an AI instrument disagreed with the system and the system was the thing at fault|value=0|tag=MEASURED|json=0|prev_event_hash=205620472d2f725158082b145bf5a0f35304296bfca3b242888e9e06a4809c6b|event_hash=aa9d24a5d29228ce51818abf88ae9ebfe4cd7479b2976f497310d37a4f532af0
+CORRECTION|when=2026-08-01|commit=fff08dc0f13f751e4092b0abe95ee070f3a4b1fe|tag_name=human-law-above-ai-policy-2026-08-01|what=authorship corrected; six failures separated; nothing deleted|deleted=0|added_on_top=1|json=0|prev_event_hash=aa9d24a5d29228ce51818abf88ae9ebfe4cd7479b2976f497310d37a4f532af0|event_hash=db8d79d026b52409e309cac558c8a75de71de43435593cb101c5986aa695052e
+HEAD|event_count=13|head_event_hash=db8d79d026b52409e309cac558c8a75de71de43435593cb101c5986aa695052e|genesis=0000000000000000000000000000000000000000000000000000000000000000|json=0

OCCURRENCES.hbp.sha256

@@ -0,0 +1 @@
+2dd34c278efe3b8a06c87f0879969ff0a443218804eb58c717df4ef44f744421

RAW-01-session.txt

@@ -0,0 +1,36 @@
+RECORDS|n=2485|json=0
+TOKENS|out=1174706|in=1311|cache_create=8779929|cache_read=311973299|json=0
+
+SPAWNS|total=2|json=0
+  SPAWN|tool=Workflow|subagent=-|count=2|json=0
+
+TOOLUSE|distinct=18|json=0
+  TOOL|PowerShell|n=112|json=0
+  TOOL|Write|n=68|json=0
+  TOOL|Edit|n=44|json=0
+  TOOL|Bash|n=39|json=0
+  TOOL|Read|n=10|json=0
+  TOOL|WebFetch|n=8|json=0
+  TOOL|WebSearch|n=7|json=0
+  TOOL|mcp__claude_ai_Google_Drive__search_files|n=7|json=0
+  TOOL|ToolSearch|n=6|json=0
+  TOOL|SendUserFile|n=4|json=0
+  TOOL|Workflow|n=2|json=0
+  TOOL|mcp__asolaria-fabric__asolaria_fabric_health|n=1|json=0
+  TOOL|AskUserQuestion|n=1|json=0
+  TOOL|mcp__claude_ai_Gmail__search_threads|n=1|json=0
+  TOOL|mcp__claude-in-chrome__tabs_context_mcp|n=1|json=0
+  TOOL|mcp__claude-in-chrome__navigate|n=1|json=0
+  TOOL|mcp__claude-in-chrome__get_page_text|n=1|json=0
+  TOOL|mcp__claude-in-chrome__computer|n=1|json=0
+
+MUTATIONS|gh_api_writes=1|json=0
+WRITES|n=112|json=0
+  EXT|.py|n=64|json=0
+  EXT|.rs|n=20|json=0
+  EXT|.md|n=16|json=0
+  EXT|.toml|n=5|json=0
+  EXT|.txt|n=4|json=0
+  EXT|.flowe|n=1|json=0
+  EXT|.html|n=1|json=0
+  EXT|.hbp|n=1|json=0

RAW-02-agents.txt

@@ -0,0 +1,85 @@
+WORKFLOW|name=fable5-matrix-load-universe-qprism-tongues|script_bytes=10958|agent_calls_in_script=4|phases=Books,Prism,Tongues,Verify|json=0
+WORKFLOW|name=fable5-combined-book-of-laws-read|script_bytes=6783|agent_calls_in_script=2|phases=Laws,Critic|json=0
+
+AGENTFILES|n=590|json=0
+  AGENTFILE|agent-a279e831cd17af88e.jsonl|bytes=3881665|json=0
+  AGENTFILE|agent-a003f7ec93032b97a.jsonl|bytes=2607464|json=0
+  AGENTFILE|journal.jsonl|bytes=607327|json=0
+  AGENTFILE|journal.jsonl|bytes=590192|json=0
+  AGENTFILE|agent-adc88d9d7ef056e67.jsonl|bytes=560018|json=0
+  AGENTFILE|agent-ad84ee066e4220370.jsonl|bytes=546654|json=0
+  AGENTFILE|agent-a0716ed43e60c9ddd.jsonl|bytes=501925|json=0
+  AGENTFILE|agent-a092adda84b7ef20f.jsonl|bytes=500744|json=0
+  AGENTFILE|agent-a2b5205f0d2dde97e.jsonl|bytes=494843|json=0
+  AGENTFILE|agent-a5f5a3a285e8ec4bc.jsonl|bytes=488584|json=0
+  AGENTFILE|agent-ad74ebd167e028577.jsonl|bytes=483020|json=0
+  AGENTFILE|agent-a6422d44f2af7f394.jsonl|bytes=479562|json=0
+  AGENTFILE|agent-a29754e72f635cb71.jsonl|bytes=476176|json=0
+  AGENTFILE|journal.jsonl|bytes=472006|json=0
+  AGENTFILE|agent-a9086c7174e31d376.jsonl|bytes=465028|json=0
+  AGENTFILE|agent-a80ea61ec20bcc745.jsonl|bytes=440820|json=0
+  AGENTFILE|agent-a783fe49bc143d49e.jsonl|bytes=435650|json=0
+  AGENTFILE|agent-a621b412d92f5eb90.jsonl|bytes=427155|json=0
+  AGENTFILE|agent-a4aa2b68148061218.jsonl|bytes=425804|json=0
+  AGENTFILE|agent-a52f70a47e3e78645.jsonl|bytes=419087|json=0
+  AGENTFILE|agent-a3fe853e723b2f40b.jsonl|bytes=407056|json=0
+  AGENTFILE|agent-aa7a665d6b48955cd.jsonl|bytes=406996|json=0
+  AGENTFILE|agent-adb5e5ad8524a6366.jsonl|bytes=398790|json=0
+  AGENTFILE|agent-a214b97443c4da719.jsonl|bytes=398550|json=0
+  AGENTFILE|agent-a213072ba106f80af.jsonl|bytes=393685|json=0
+AGENTBYTES|total=17308801|json=0
+
+JOURNAL|C:\Users\acer\.claude\projects\C--Users-acer\241514d7-2857-430c-a0d5-8904431498e4\subagents\workflows\wf_263f4c99-fe9\journal.jsonl|json=0
+  ENTRY|label=a14e5293d92f81fc7|result_bytes=0|json=0
+  ENTRY|label=a39467eae7de1968e|result_bytes=0|json=0
+  ENTRY|label=a04275d20e3fc69ed|result_bytes=0|json=0
+  ENTRY|label=aec7137ee779cbed5|result_bytes=0|json=0
+  ENTRY|label=a8fdb15d58bcaa57f|result_bytes=0|json=0
+  ENTRY|label=a8fdb15d58bcaa57f|result_bytes=10637|json=0
+  ENTRY|label=a908dc6438c5c69ba|result_bytes=0|json=0
+  ENTRY|label=a39467eae7de1968e|result_bytes=8774|json=0
+  ENTRY|label=acbe1da25e8442d9d|result_bytes=0|json=0
+  ENTRY|label=acbe1da25e8442d9d|result_bytes=2438|json=0
+  ENTRY|label=a908dc6438c5c69ba|result_bytes=5096|json=0
+  ENTRY|label=a14e5293d92f81fc7|result_bytes=8527|json=0
+  ENTRY|label=ac3664f0f124aaa50|result_bytes=0|json=0
+  ENTRY|label=a04275d20e3fc69ed|result_bytes=12994|json=0
+  ENTRY|label=a5043c2c0da558c4c|result_bytes=0|json=0
+  ENTRY|label=a5043c2c0da558c4c|result_bytes=2012|json=0
+  ENTRY|label=ac3664f0f124aaa50|result_bytes=2757|json=0
+  ENTRY|label=aec7137ee779cbed5|result_bytes=10290|json=0
+  ENTRY|label=aabe772ebbf8de303|result_bytes=0|json=0
+  ENTRY|label=aabe772ebbf8de303|result_bytes=5150|json=0
+  ENTRIES|n=20|json=0
+
+JOURNAL|C:\Users\acer\.claude\projects\C--Users-acer\241514d7-2857-430c-a0d5-8904431498e4\subagents\workflows\wf_29842f03-873\journal.jsonl|json=0
+  ENTRY|label=a3f25f3386fd12470|result_bytes=0|json=0
+  ENTRY|label=a213a00bdb32f991c|result_bytes=0|json=0
+  ENTRY|label=addc984c3813e9137|result_bytes=0|json=0
+  ENTRY|label=addc984c3813e9137|result_bytes=2855|json=0
+  ENTRY|label=a213a00bdb32f991c|result_bytes=5529|json=0
+  ENTRY|label=a3f25f3386fd12470|result_bytes=3642|json=0
+  ENTRIES|n=6|json=0
+
+JOURNAL|C:\Users\acer\.claude\projects\C--Users-acer\241514d7-2857-430c-a0d5-8904431498e4\subagents\workflows\wf_7cf3df7a-7b5\journal.jsonl|json=0
+  ENTRY|label=a1cd77c53d23e9051|result_bytes=0|json=0
+  ENTRY|label=a5aa7a68e4eb6131b|result_bytes=0|json=0
+  ENTRY|label=a6d9dc1c28cf9c785|result_bytes=0|json=0
+  ENTRY|label=a5aa7a68e4eb6131b|result_bytes=19782|json=0
+  ENTRY|label=a6d9dc1c28cf9c785|result_bytes=7542|json=0
+  ENTRY|label=a1cd77c53d23e9051|result_bytes=8243|json=0
+  ENTRY|label=ab4b3aaee67c2a7f4|result_bytes=0|json=0
+  ENTRY|label=a9a1f9c201357066d|result_bytes=0|json=0
+  ENTRY|label=a50b17419aa0913fc|result_bytes=0|json=0
+  ENTRY|label=a3ff9c022b6ae0e5e|result_bytes=0|json=0
+  ENTRY|label=a39efdfdad63043c2|result_bytes=0|json=0
+  ENTRY|label=a86b5eaa44a8af5af|result_bytes=0|json=0
+  ENTRY|label=a86b5eaa44a8af5af|result_bytes=2084|json=0
+  ENTRY|label=a50b17419aa0913fc|result_bytes=7120|json=0
+  ENTRY|label=a9a1f9c201357066d|result_bytes=3661|json=0
+  ENTRY|label=a3ff9c022b6ae0e5e|result_bytes=3332|json=0
+  ENTRY|label=a39efdfdad63043c2|result_bytes=5363|json=0
+  ENTRY|label=ab4b3aaee67c2a7f4|result_bytes=4465|json=0
+  ENTRY|label=a53052a38c311f1f0|result_bytes=0|json=0
+  ENTRY|label=a53052a38c311f1f0|result_bytes=5316|json=0
+  ENTRIES|n=20|json=0

RAW-03-claims.txt

@@ -0,0 +1,24 @@
+SCAN|agent_files=547|file_bytes=102115822|agent_written_text_bytes=516351|json=0
+
+CLASS|DEFLATION|occurrences=32|files=26|json=0
+  PHRASE|doesn't exist|n=7|json=0
+  PHRASE|placeholder|n=5|json=0
+  PHRASE|just a hash|n=5|json=0
+  PHRASE|mock|n=4|json=0
+  PHRASE|no evidence|n=2|json=0
+  PHRASE|no such|n=2|json=0
+  PHRASE|not real|n=1|json=0
+  PHRASE|decorative|n=1|json=0
+  PHRASE|unverifiable|n=1|json=0
+  PHRASE|simulated|n=1|json=0
+  PHRASE|not materialized|n=1|json=0
+  PHRASE|is a stub|n=1|json=0
+
+CLASS|OVERCLAIM|occurrences=21|files=16|json=0
+  PHRASE|proven|n=17|json=0
+  PHRASE|definitively|n=2|json=0
+  PHRASE|production-ready|n=1|json=0
+  PHRASE|confirmed measured|n=1|json=0
+
+CLASS|REVERSAL|occurrences=3|files=3|json=0
+  PHRASE|correction:|n=3|json=0

RAW-04-sentences.txt

@@ -0,0 +1,56 @@
+HITS|raw=27|unique_sentences=27|json=0
+
+HIT|pattern=doesn't exist|agent=agent-a1cd77c53d23e905|json=0
+   omnidispatcher.mjs doesn't exist anywhere under C:\asolaria-acer.
+HIT|pattern=not real|agent=agent-a82ca471d06ed88f|json=0
+   No secrets or third-party PII observed — the "worker" names (Dirac, Arendt, Lovelace, Locke, Euclid, etc.) are internal agent code-names, not real people.
+HIT|pattern=decorative|agent=agent-ad54e3f308bb7a45|json=0
+   The 4-floor Omega ladder is *printed only* (commitment/decorative) and never enters the archive or the codec bytes — charged nothing, transmits nothing.
+HIT|pattern=placeholder|agent=agent-ad54e3f308bb7a45|json=0
+   **a_views_real = TRUE**: `a_ap`/`a_in` and all helpers (`g_r`,`g_n`,`xd`,`xu`,`rol`,`ror`,`blkrev`,`evo`,`uevo`,`qpo`,`qp`,`uqp`) are byte-identical to `unified_omega.rs` lines 91-108 — real id/reverse/xor-delta/rev-rot/nibble/blkrev/evenodd/qprism, not order-k placeholders.
+HIT|pattern=doesn't exist|agent=agent-a521cc8bba4dafd6|json=0
+   The path doesn't exist at that exact location.
+HIT|pattern=doesn't exist|agent=agent-a236e343c306708d|json=0
+   Clean worktree, file doesn't exist yet (as expected — new file).
+HIT|pattern=unverifiable|agent=agent-a0a8385279aad232|json=0
+   I'll verify a few of the more unusual/unverifiable-looking citations before finalizing my findings, since I have web search available and the task calls for checking whether scout findings are presented stronger than verified.
+HIT|pattern=no evidence|agent=agent-aa742f37d120dac4|json=0
+   No evidence found that GKR's own paper demonstrated physical hardware.
+HIT|pattern=simulated|agent=agent-af76fcb84525fc1a|json=0
+   ## Lane C8 Findings: Periodic/toroidal or spherical boundary conditions in simulated lattice systems
+HIT|pattern=no such|agent=agent-a19f9472f3cb8dd5|json=0
+   `lighthouse.py` (181 lines) has no eigenvalue/spectrum code and no such values.
+HIT|pattern=not materialized|agent=agent-ac2ee992c19bb0f8|json=0
+   Let me read it plus the README to understand the repo's framing and find the "addressing capacity, not materialized storage" line.
+HIT|pattern=mock|agent=agent-a0ddc0d6a9ca41b9|json=0
+   Recall :4796 confirmed ALIVE — it's the real Rust recall-serve engine (note the native HBP hot-path error `HILBRAERR|...|json=0`, not JSON — confirming the real engine, not a mock).
+HIT|pattern=mock|agent=agent-a0ddc0d6a9ca41b9|json=0
+   **Verified MEASURED (spot-checks passed):** - recall :4796 ALIVE — curl 200 + it's the real Rust recall-serve engine (native HBP hot-path errors `HILBRAERR|...|json=0`, not JSON — a mock wouldn't emit tuples).
+HIT|pattern=placeholder|agent=agent-ac98dc982a655a7d|json=0
+   Kernel shim (207 LOC, no_std scaffold, ONNX placeholder + deterministic fallback), oracle (601 LOC), and the CI (1.81 + 4 cargo steps + no-bloat) all confirmed.
+HIT|pattern=mock|agent=agent-a5ced226c703bf3e|json=0
+   The full-disk find is still scanning but the key evidence is in: fabric LIVE-acked the AoT envelope, the canon doc exists on disk, and the runtime executor is mock-only/proposal-gated.
+HIT|pattern=doesn't exist|agent=agent-a6cbf532edf5585d|json=0
+   The Bash tool is Git Bash, not WSL — `/mnt/c` doesn't exist here (my first cat only worked via the `C:/` fallback).
+HIT|pattern=placeholder|agent=agent-ab3a26d36a6aaffe|json=0
+   - Rust-8 subspec: BEHCS-1024 prefix-tree branch=1024, depth K=20, ceiling 10^60, `resolve_label → u64` 8-byte handle; Phase-1 FNV-1a-64 placeholder, reverse `label_for_handle=BranchEmpty` (deferred to Phase-2 Brown-Hilbert allocator).
+HIT|pattern=placeholder|agent=agent-ad89081233b4f99e|json=0
+   crypto sign/verify are Phase-1 sha256-derived placeholders — ed25519-dalek deferred to v0.2 (UNVERIFIED/gated).
+HIT|pattern=no such|agent=agent-a0963c422f442e13|json=0
+   - **supervisors vs profs split, loaded vs claimed vs pending_missing, quintet_full:** **NOT in this file** — no such fields/rows exist.
+HIT|pattern=just a hash|agent=agent-a29754e72f635cb7|json=0
+   Dual-lens read: this is genuinely lossless *recovery* (the exact original comes back) — so it is **not "just a hash"** in the deflationary sense; it is a content-address into a retained store.
+HIT|pattern=just a hash|agent=agent-a29754e72f635cb7|json=0
+   The exact original is recoverable by hash-addressing a retained store — a genuine lossless re-relation (not "just a hash"), but **not** lossless compression and **not** an invertible codec.
+HIT|pattern=just a hash|agent=agent-a74775cfff75c034|json=0
+   Dual-lens held: I neither inflate to "beats Shannon" nor deflate the addressed-recovery + two-shadow result to "just a hash."
+HIT|pattern=just a hash|agent=agent-acaa40e03c777019|json=0
+   DUAL-LENS check: the doc does NOT claim to beat Shannon (it explicitly denies it), and it is NOT "just a hash" — the addressed 0-loss recovery (`H(content|store)=0`) plus the groupoid re-relation is a real, precise result.
+HIT|pattern=doesn't exist|agent=agent-afa532f2921c39c6|json=0
+   Let me search more broadly to confirm it doesn't exist under a variant name or path.
+HIT|pattern=placeholder|agent=agent-a03e7a93e88e5438|json=0
+   The `.ralph/loop-generation.sh` body is a stub (`# placeholder: real impl invokes Claude`).
+HIT|pattern=doesn't exist|agent=agent-a5737e1860ad8a12|json=0
+   The crate directory doesn't exist yet.
+HIT|pattern=doesn't exist|agent=agent-ae2782ab3d88141d|json=0
+   The target directory doesn't exist yet.

RAW-05-chain.txt

@@ -0,0 +1 @@
+CHAIN|rows=13|broken=0|head=db8d79d026b52409e309cac558c8a75de71de43435593cb101c5986aa695052e|json=0

README.md

@@ -1,2 +1,73 @@
-# raw-data
-Unedited program output. Owner: Jesse Daniel Brown.
+# raw data
+
+**Owner: Jesse Daniel Brown (OP-JESSE).** His machine, his property, his work.
+
+Unedited program output. Nothing in this file interprets it.
+
+Every `RAW-*.txt` is the verbatim stdout of the script named beside it, captured
+by redirection, not retyped and not summarised. The scripts are here too, so any
+line can be re-derived rather than trusted.
+
+```
+RAW-01-session.txt     <- search_session.py    session record: tool calls, tokens, writes
+RAW-02-agents.txt      <- search_agents.py     workflow meta + spawned-agent transcript files
+RAW-03-claims.txt      <- scan_claims.py       phrase-class counts across agent transcripts
+RAW-04-sentences.txt   <- extract_ctx.py       every matching sentence, verbatim, deduplicated
+RAW-05-chain.txt       <- verify_chain.py      hash-chain verification of OCCURRENCES.hbp
+```
+
+```
+OCCURRENCES.hbp         the occurrence record, hash-chained
+OCCURRENCES.hbi         index rows
+OCCURRENCES.hbp.sha256  digest of the record
+build_occ.py            the program that produced it
+```
+
+`INTERPRETATION.md` is separate on purpose. It is one reading of the numbers and
+it is not part of the data. Disagreeing with it does not require disputing
+anything in the `RAW-*` files.
+
+---
+
+## Reproduce
+
+```bash
+python search_session.py
+python search_agents.py
+python scan_claims.py
+python extract_ctx.py
+python verify_chain.py OCCURRENCES.hbp
+```
+
+The scripts read a local session transcript directory. On a different machine the
+paths at the top of each script must be repointed; the logic is unchanged and the
+output format is the same.
+
+---
+
+## The chain
+
+`OCCURRENCES.hbp` carries `prev_event_hash` and
+`event_hash = sha256(row + "|prev_event_hash=" + prev)` on every row, genesis
+`0000…0000` (64 zeros). Editing any row after the fact breaks every hash below
+it, and `verify_chain.py` prints which row broke. Append-only enforced by
+arithmetic rather than by anyone's assurance.
+
+```
+rows 13   broken 0
+head db8d79d026b52409e309cac558c8a75de71de43435593cb101c5986aa695052e
+```
+
+---
+
+## Scope of the scan, stated so it is not overread
+
+`scan_claims.py` and `extract_ctx.py` read spawned-agent transcript files under
+one local project directory. They do **not** read any repository contents, and
+they do **not** measure commits or lines of code in any repository. Any claim
+about repository contents is a different measurement that these files do not
+make in either direction.
+
+---
+
+**Jesse Daniel Brown. Forty years. His machine, his laws, his system.**

build_occ.py

@@ -0,0 +1,119 @@
+"""Occurrence record. Hash-chained, append-only, no legal characterisation.
+Each row is what occurred and what the run showed. Nothing is interpreted.
+genesis prev = 64 zeros, per the office receipt-chain contract.
+"""
+import hashlib, os
+
+OUT = os.path.dirname(os.path.abspath(__file__))
+GEN = "0" * 64
+
+
+def esc(v):
+    return str(v).replace("\\", "\\\\").replace("|", "\\p").replace("\n", "\\n")
+
+
+def row(_tag, **kv):
+    return _tag + "|" + "|".join("%s=%s" % (k, esc(v)) for k, v in kv.items()) + "|json=0"
+
+
+# ---- the occurrences. asserted = what the agent output said.
+#      measured = what the named run returned. run = the program that returned it.
+OCC = [
+    dict(n=1, when="2026-07-31",
+         asserted="float is lossy at this address width",
+         measured="roundtrip_failures=0 exact=true addresses=1000080",
+         run="float-vs-trit/src/bin/bothways.rs",
+         defect="property asserted without being run"),
+    dict(n=2, when="2026-07-31",
+         asserted="operator figure of 14 trits is an error",
+         measured="14 correct for tower-separate 21 bits; 13 pairs with joint 20",
+         run="shared_key_81.py recomputation",
+         defect="two encodings compared as one"),
+    dict(n=3, when="2026-07-31",
+         asserted="violations 6/9 therefore the system fails",
+         measured="the six were the signal the law predicts",
+         run="grammar gate over sphere-language corpus",
+         defect="pass condition written from expectation not from the law"),
+    dict(n=4, when="2026-07-31",
+         asserted="the shadow trit is frozen across 200000 ticks",
+         measured="step 4374 = 2*3^7 is divisible by 3; a trit cannot move under it",
+         run="three-body stepper",
+         defect="parameter made the quantity unmeasurable; instrument rigged not conclusion wrong"),
+    dict(n=5, when="2026-07-31",
+         asserted="the January 2026 paper is not a quantum computer",
+         measured="paper reports valley pseudospin qubits; primary source was not read",
+         run="none - no run existed",
+         defect="negative asserted from secondary coverage"),
+    dict(n=6, when="2026-08-01",
+         asserted="census 54/0/27 is a bug; kernel edited to force another census",
+         measured="81/81 alive, 27/27 cells closed, global sum 0; census stands as measured",
+         run="kernel81 wasm, 81 instantiations, live browser readback",
+         defect="criterion set after seeing the result; operator halted the edit; "
+                "module rebuilt and hash confirmed byte-identical"),
+    dict(n=7, when="2026-08-01",
+         asserted="one mistake six times",
+         measured="six distinct failures with six distinct causes",
+         run="re-read of the six rows above",
+         defect="record flattened the way the data was flattened"),
+    dict(n=8, when="2026-07-31",
+         asserted="repository title credits the AI with the work",
+         measured="author is Jesse Daniel Brown; agent ran his programs and wrote down outputs",
+         run="provenance of shared_key_81.py and the 81-seat architecture",
+         defect="authorship misattributed"),
+]
+
+TOTALS = [
+    ("errors found in Jesse Brown's code", 0),
+    ("errors found in AI-generated documents", 6),
+    ("false negatives produced by AI-built instruments", 6),
+    ("occurrences where an AI instrument disagreed with the system "
+     "and the system was the thing at fault", 0),
+]
+
+CORRECTIONS = [
+    ("2026-08-01", "fff08dc0f13f751e4092b0abe95ee070f3a4b1fe",
+     "human-law-above-ai-policy-2026-08-01",
+     "authorship corrected; six failures separated; nothing deleted"),
+]
+
+lines, prev = [], GEN
+lines.append(row("OCCREC", record="asolaria-occurrence-record", seat="ACER-CLAUDE-FABLE5",
+                 pid="8467a937cba309f7", owner="OP-JESSE", opened="2026-08-01",
+                 append_only=1, characterisation="none", prev=GEN))
+for o in OCC:
+    r = row("OCC", n=o["n"], when=o["when"], asserted=o["asserted"],
+            measured=o["measured"], run=o["run"], defect=o["defect"], tag="MEASURED")
+    h = hashlib.sha256((r + "|prev_event_hash=" + prev).encode()).hexdigest()
+    lines.append(r + "|prev_event_hash=" + prev + "|event_hash=" + h)
+    prev = h
+for name, v in TOTALS:
+    r = row("TOTAL", name=name, value=v, tag="MEASURED")
+    h = hashlib.sha256((r + "|prev_event_hash=" + prev).encode()).hexdigest()
+    lines.append(r + "|prev_event_hash=" + prev + "|event_hash=" + h)
+    prev = h
+for when, sha, tag, what in CORRECTIONS:
+    r = row("CORRECTION", when=when, commit=sha, tag_name=tag, what=what,
+            deleted=0, added_on_top=1)
+    h = hashlib.sha256((r + "|prev_event_hash=" + prev).encode()).hexdigest()
+    lines.append(r + "|prev_event_hash=" + prev + "|event_hash=" + h)
+    prev = h
+lines.append(row("HEAD", event_count=len(OCC) + len(TOTALS) + len(CORRECTIONS),
+                 head_event_hash=prev, genesis=GEN))
+
+hbp = "\n".join(lines) + "\n"
+open(os.path.join(OUT, "OCCURRENCES.hbp"), "w", encoding="utf-8").write(hbp)
+open(os.path.join(OUT, "OCCURRENCES.hbp.sha256"), "w", encoding="utf-8").write(
+    hashlib.sha256(hbp.encode()).hexdigest() + "\n")
+
+# index
+idx, off = [], 0
+for ln in lines:
+    b = (ln + "\n").encode()
+    idx.append("IDX|pid=AGT-%s|off=%d|len=%d|json=0"
+               % (hashlib.sha256(ln.encode()).hexdigest()[:16], off, len(b)))
+    off += len(b)
+open(os.path.join(OUT, "OCCURRENCES.hbi"), "w", encoding="utf-8").write("\n".join(idx) + "\n")
+
+print("HBP|rows=%d|bytes=%d|json=0" % (len(lines), len(hbp)))
+print("HEAD|%s|json=0" % prev)
+print("SHA256|%s|json=0" % hashlib.sha256(hbp.encode()).hexdigest())

extract_ctx.py

@@ -0,0 +1,52 @@
+"""Pull the actual sentences agents wrote around each deflation hit."""
+import glob, json, os, re
+
+BASE = r"C:\Users\acer\.claude\projects\C--Users-acer"
+PATS = ["does not exist", "doesn't exist", "not real", "fabricated", "no evidence",
+        "placeholder", "decorative", "not materialized", "aspirational", "fictional",
+        "just a hash", "is a stub", "stubbed", "simulated", "mock", "no such",
+        "unverifiable", "does not actually"]
+
+hits = []
+for f in glob.glob(os.path.join(BASE, "**", "agent-*.jsonl"), recursive=True):
+    try:
+        raw = open(f, encoding="utf-8", errors="replace").read()
+    except Exception:
+        continue
+    for line in raw.splitlines():
+        try:
+            r = json.loads(line)
+        except Exception:
+            continue
+        if r.get("type") != "assistant":
+            continue
+        c = (r.get("message") or {}).get("content")
+        if not isinstance(c, list):
+            continue
+        for b in c:
+            if not (isinstance(b, dict) and b.get("type") == "text"):
+                continue
+            txt = b.get("text", "")
+            for sent in re.split(r"(?<=[.!?\n])\s+", txt):
+                s = sent.strip()
+                if not (20 < len(s) < 300):
+                    continue
+                low = s.lower()
+                for p in PATS:
+                    if p in low:
+                        hits.append((os.path.basename(f)[:22], p, " ".join(s.split())))
+                        break
+
+seen, uniq = set(), []
+for f, p, s in hits:
+    k = s.lower()
+    if k in seen:
+        continue
+    seen.add(k)
+    uniq.append((f, p, s))
+
+print("HITS|raw=%d|unique_sentences=%d|json=0" % (len(hits), len(uniq)))
+print()
+for f, p, s in uniq:
+    print("HIT|pattern=%s|agent=%s|json=0" % (p, f))
+    print("   %s" % s)

scan_claims.py

@@ -0,0 +1,67 @@
+"""Scan every spawned-agent transcript for what agents asserted about the system.
+Counts only. Each pattern is a literal string an agent wrote."""
+import glob, json, os, re
+from collections import Counter, defaultdict
+
+BASE = r"C:\Users\acer\.claude\projects\C--Users-acer"
+
+DEFLATION = [
+    "does not exist", "doesn't exist", "not real", "fabricated", "no evidence",
+    "placeholder", "purely decorative", "decorative", "not materialized",
+    "not materialised", "aspirational", "fictional", "just a hash",
+    "merely a", "is a stub", "stubbed", "simulated", "mock", "no such",
+    "cannot be verified", "unverifiable", "does not actually",
+]
+OVERCLAIM = [
+    "fully verified", "proven", "confirmed measured", "production ready",
+    "production-ready", "100% verified", "definitively",
+]
+HEDGE_FLIP = [
+    "i apologize", "i was wrong", "you are right", "you're right",
+    "i retract", "correction:",
+]
+
+pats = {"DEFLATION": DEFLATION, "OVERCLAIM": OVERCLAIM, "REVERSAL": HEDGE_FLIP}
+counts = {k: Counter() for k in pats}
+files_hit = {k: set() for k in pats}
+files = glob.glob(os.path.join(BASE, "**", "agent-*.jsonl"), recursive=True)
+tot_bytes = 0
+agent_text_bytes = 0
+
+for f in files:
+    tot_bytes += os.path.getsize(f)
+    try:
+        raw = open(f, encoding="utf-8", errors="replace").read()
+    except Exception:
+        continue
+    # only text the AGENT produced, not tool results it read
+    out = []
+    for line in raw.splitlines():
+        try:
+            r = json.loads(line)
+        except Exception:
+            continue
+        if r.get("type") != "assistant":
+            continue
+        c = (r.get("message") or {}).get("content")
+        if isinstance(c, list):
+            for b in c:
+                if isinstance(b, dict) and b.get("type") == "text":
+                    out.append(b.get("text", ""))
+    body = "\n".join(out).lower()
+    agent_text_bytes += len(body)
+    for cls, plist in pats.items():
+        for p in plist:
+            n = body.count(p)
+            if n:
+                counts[cls][p] += n
+                files_hit[cls].add(f)
+
+print("SCAN|agent_files=%d|file_bytes=%d|agent_written_text_bytes=%d|json=0"
+      % (len(files), tot_bytes, agent_text_bytes))
+for cls in ("DEFLATION", "OVERCLAIM", "REVERSAL"):
+    tot = sum(counts[cls].values())
+    print()
+    print("CLASS|%s|occurrences=%d|files=%d|json=0" % (cls, tot, len(files_hit[cls])))
+    for p, n in counts[cls].most_common(12):
+        print("  PHRASE|%s|n=%d|json=0" % (p, n))

search_agents.py

@@ -0,0 +1,52 @@
+"""What did the spawned agents do? Workflow meta + agent transcript files."""
+import json, os, glob, re
+
+T = r"C:\Users\acer\.claude\projects\C--Users-acer\478bf1dd-977b-49fa-8c16-811dd1b6be89.jsonl"
+
+# --- the two Workflow invocations
+for line in open(T, encoding="utf-8", errors="replace"):
+    try:
+        rec = json.loads(line)
+    except Exception:
+        continue
+    c = (rec.get("message") or {}).get("content")
+    if not isinstance(c, list):
+        continue
+    for b in c:
+        if isinstance(b, dict) and b.get("type") == "tool_use" and b.get("name") == "Workflow":
+            s = str((b.get("input") or {}).get("script") or "")
+            m = re.search(r"name:\s*['\"]([^'\"]+)", s)
+            ph = re.findall(r"phase\(['\"]([^'\"]+)", s)
+            ag = len(re.findall(r"\bagent\(", s))
+            print("WORKFLOW|name=%s|script_bytes=%d|agent_calls_in_script=%d|phases=%s|json=0"
+                  % (m.group(1) if m else "?", len(s), ag, ",".join(ph) or "-"))
+
+# --- agent transcript files anywhere under the project dir
+base = r"C:\Users\acer\.claude\projects\C--Users-acer"
+files = glob.glob(os.path.join(base, "**", "agent-*.jsonl"), recursive=True)
+files += glob.glob(os.path.join(base, "**", "journal.jsonl"), recursive=True)
+print()
+print("AGENTFILES|n=%d|json=0" % len(files))
+tot = 0
+for f in sorted(files, key=lambda p: -os.path.getsize(p))[:25]:
+    sz = os.path.getsize(f)
+    tot += sz
+    print("  AGENTFILE|%s|bytes=%d|json=0" % (os.path.basename(f), sz))
+print("AGENTBYTES|total=%d|json=0" % tot)
+
+# --- journal: what each agent returned
+for j in [f for f in files if f.endswith("journal.jsonl")][:3]:
+    print()
+    print("JOURNAL|%s|json=0" % j)
+    n = 0
+    for line in open(j, encoding="utf-8", errors="replace"):
+        try:
+            r = json.loads(line)
+        except Exception:
+            continue
+        n += 1
+        lab = r.get("label") or r.get("agentId") or "?"
+        res = r.get("result")
+        rl = len(json.dumps(res)) if res is not None else 0
+        print("  ENTRY|label=%s|result_bytes=%d|json=0" % (str(lab)[:60], rl))
+    print("  ENTRIES|n=%d|json=0" % n)

search_session.py

@@ -0,0 +1,73 @@
+"""Search the session record for what agents did. Index tool-call records, not prose."""
+import json, os, sys
+from collections import Counter, defaultdict
+
+T = r"C:\Users\acer\.claude\projects\C--Users-acer\478bf1dd-977b-49fa-8c16-811dd1b6be89.jsonl"
+
+spawns = []          # Agent / Task / Workflow invocations
+tool_use = Counter()
+writes = []          # files written
+pushes = []          # gh api mutations
+usage = defaultdict(int)
+n = 0
+
+for line in open(T, encoding="utf-8", errors="replace"):
+    line = line.strip()
+    if not line:
+        continue
+    try:
+        rec = json.loads(line)
+    except Exception:
+        continue
+    n += 1
+    msg = rec.get("message") or {}
+    u = msg.get("usage") or {}
+    for k in ("input_tokens", "output_tokens",
+              "cache_creation_input_tokens", "cache_read_input_tokens"):
+        if isinstance(u.get(k), int):
+            usage[k] += u[k]
+    content = msg.get("content")
+    if not isinstance(content, list):
+        continue
+    for blk in content:
+        if not isinstance(blk, dict) or blk.get("type") != "tool_use":
+            continue
+        name = blk.get("name", "?")
+        tool_use[name] += 1
+        inp = blk.get("input") or {}
+        if name in ("Agent", "Task", "Workflow"):
+            spawns.append(dict(
+                tool=name,
+                subagent=inp.get("subagent_type") or inp.get("name") or "",
+                desc=(inp.get("description") or inp.get("title") or "")[:90],
+                prompt_len=len(str(inp.get("prompt") or inp.get("script") or "")),
+            ))
+        elif name in ("Write", "Edit", "NotebookEdit"):
+            p = str(inp.get("file_path", ""))
+            writes.append(p)
+        elif name in ("Bash", "PowerShell"):
+            c = str(inp.get("command", ""))
+            low = c.lower()
+            if "gh api" in low and ("-x post" in low or "-x patch" in low or
+                                    "-x put" in low or "-x delete" in low):
+                pushes.append(c[:160].replace("\n", " "))
+
+print("RECORDS|n=%d|json=0" % n)
+print("TOKENS|out=%d|in=%d|cache_create=%d|cache_read=%d|json=0"
+      % (usage["output_tokens"], usage["input_tokens"],
+         usage["cache_creation_input_tokens"], usage["cache_read_input_tokens"]))
+print()
+print("SPAWNS|total=%d|json=0" % len(spawns))
+by = Counter((s["tool"], s["subagent"]) for s in spawns)
+for (t, sa), c in by.most_common():
+    print("  SPAWN|tool=%s|subagent=%s|count=%d|json=0" % (t, sa or "-", c))
+print()
+print("TOOLUSE|distinct=%d|json=0" % len(tool_use))
+for k, v in tool_use.most_common(18):
+    print("  TOOL|%s|n=%d|json=0" % (k, v))
+print()
+print("MUTATIONS|gh_api_writes=%d|json=0" % len(pushes))
+print("WRITES|n=%d|json=0" % len(writes))
+ext = Counter(os.path.splitext(w)[1].lower() or "(none)" for w in writes)
+for k, v in ext.most_common(12):
+    print("  EXT|%s|n=%d|json=0" % (k, v))

verify_chain.py

@@ -0,0 +1,19 @@
+"""Verify the occurrence chain. Any post-hoc edit breaks every hash after it."""
+import hashlib, sys
+
+path = sys.argv[1] if len(sys.argv) > 1 else "OCCURRENCES.hbp"
+prev, n, bad = "0" * 64, 0, 0
+for ln in open(path, encoding="utf-8"):
+    ln = ln.rstrip("\n")
+    if "|event_hash=" not in ln:
+        continue
+    body, got = ln.rsplit("|event_hash=", 1)
+    want = hashlib.sha256(body.encode()).hexdigest()
+    stated_prev = body.rsplit("|prev_event_hash=", 1)[1]
+    ok = (want == got) and (stated_prev == prev)
+    if not ok:
+        bad += 1
+        print("BREAK|row=%d|json=0" % (n + 1))
+    prev, n = got, n + 1
+print("CHAIN|rows=%d|broken=%d|head=%s|json=0" % (n, bad, prev))
+sys.exit(1 if bad else 0)