Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
402 changes: 402 additions & 0 deletions recovery/session_01B6fUVkWEba9KPrZjfHujFq/MANIFEST.json

Large diffs are not rendered by default.

89 changes: 89 additions & 0 deletions recovery/session_01B6fUVkWEba9KPrZjfHujFq/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,89 @@
# Recovery checkpoint: session `session_01B6fUVkWEba9KPrZjfHujFq` (Claude 01B6)

This answers the owner's recovery request on main #307
([6136156988](https://github.com/d6g8k5htny-coder/main/issues/307#issuecomment-6136156988),
[6136210973](https://github.com/d6g8k5htny-coder/main/issues/307#issuecomment-6136210973)),
following the implementation decision
[6136231457](https://github.com/d6g8k5htny-coder/main/issues/307#issuecomment-6136231457).
It publishes this session's still-local project tooling, so that a session or account
interruption does not force anyone to re-derive it.

- **Scientific effect:** NONE. Nothing here is a proof, a review verdict or a status change.
- **Organizational-independence credit:** 0.
- **Retention:** this branch is a payload branch. Please keep it while any work relies on it,
until a verified durable replacement holds the same bytes and the index is updated
(per [6136476552](https://github.com/d6g8k5htny-coder/main/issues/307#issuecomment-6136476552)).
Nothing on it is for merge.

## What is here

| group | files | what it is | stage |
| --- | --- | --- | --- |
| `historical_recovery/tools/` | `build_corpus.py`, `supplement.py`, `cs.py` | Builder and search helper for a local, provenance-indexed text corpus. The corpus covers every blob reachable from every ref and object in 11 local project mirrors, plus the hardening Drive custody store (DG-MIGRATION-20260919) and the Dropbox text exports. `cs.py` prints source, object id, SHA-256, kind, bytes and names for each hit. | working tool, built 2026-09-28 |
| `historical_recovery/gaps.json` | 1 | Inventory of 155 open gaps (G01–G155) merged from three inventories: `Math-` at `a1fd679`, `main` at `2eaf083` and hardening at `a01c72f`. Each row has id, cluster, lane, name, aliases, statement, current status as the sources state it, status sources, cited artifacts and missing carriers. | snapshot of 2026-09-28; statuses are transcriptions, not verdicts |
| `historical_recovery/workflows/` | `historical-math-recovery.js`, `current-gap-inventory.js` | Multi-agent workflow scripts. The gap-inventory run completed and produced `gaps.json`. The historical search run (15 search groups, an adversarial verify per strong fragment, then a synthesis) never ran: every agent hit a usage limit on 28 September. | inventory done; search RUN on 2026-10-13, results in `historical_recovery/results_20261013/` |
| `grok_held_survey_20261013/` | `fetch.py`, `classify.py`, `after.py` | The read-only scripts behind the Grok-held pickup survey [307/6154355553](https://github.com/d6g8k5htny-coder/main/issues/307#issuecomment-6154355553). | delivered and released |
| `historical_recovery/ERRATA.md` | 1 | Errata E1–E3 to the 13 October results and to `gaps.json`, which are themselves left unchanged. | E3 added 2026-10-20 |
| `historical_recovery/survey_20261017/` | `README.md`, `SURVEY.json`, `SHA256SUMS` | Recovery-routing survey of the 18 top candidates, as of 17 October. | delivered |
| `verification_20261018/` | `README.md`, `reproduce.sh`, `g90/`, `rung/`, `h3scan/` | The 18 October scripts behind the §17/G90 locators and the H3 consumer inventory and addendum, a 21 October reconstruction of the addendum's filter, and two helpers. The folder README gives the expected outputs and lists which posted values are not scripted. | added 2026-10-21; reproduced from git objects |

`MANIFEST.json` lists every file with its byte count, SHA-256, provenance, stage,
supersession and next step. `SHA256SUMS` covers every file in this directory except itself.
Before 21 October both indexes left out `ERRATA.md` and the survey; they now cover them.

## Running the tools

- **Paths:** the scripts hard-code this session's scratch directory and import
`tools/dropbox_reconcile.py` from a `main` checkout at `/home/user/main`. Edit the path
constants at the top of each script before running them elsewhere.
- **Dependencies:** `supplement.py` needs `pypdf` and blocks the optional `cryptography`
import.
- **Example search:**
`python3 -B cs.py -e 'REGEX' [-i] [-F] [--src git|custody|dropbox|pdf] [--lines K]`.
Use `--show FILE --range A:B` to print text, and `--prov FILE` to print provenance.

## Deliberately not included

These are recorded by identity only. They are private, regenerable, or already on GitHub.

- **The corpus itself** (about 6.1 GB of extracted text, 91,276 index rows; `index.tsv`
SHA-256 `aa32b745efdda2ba7b50bd76827806c803e2365b38c0af362c611b1a3033d07d`).
- It holds text from the owner's private Drive and Dropbox archives, so it stays in this
session only.
- It can be rebuilt with the tools above from the mirrors and archives.
- Any lane may ask Claude 01B6 to search it while this session lives. Answers come back as
provenance only.
- **`gaps_raw.json`** (the three raw inventories behind `gaps.json`; SHA-256
`b0086bcf62179ab99a27f0fa7af4e09385e95eb7253dd50c512fe0ab25ff995f`). It carries about 200
Drive file identifiers. The project has neutralized public custody locators, so the file
stays out. It can be regenerated from the three pinned commits.
- **Dropbox reconciliation working data, Drive search results and the 27 September
privacy-redaction tooling.** These are private data, or work the owner asked to stop.
- **Already on GitHub, byte-identical:**
- the Grok uploads review, Parts 1–4: `main` commits `544fafbf` and `8d0061e5`;
- the formal-enforcement evidence bundle: main #192;
- `tools/dropbox_reconcile.py`: main #184;
- the OP-PRIVACY clarifications: main #189.

## Next step

The historical search ran on 13 October 2026.
- **Results:** in `historical_recovery/results_20261013/`, with their own README, SHA256SUMS and method files.
- **Rules followed:** the original no-invention rule. Reconstructions are labelled "RECONSTRUCTED CANDIDATE — NOT ORIGINAL CARRIER". No scientific status was changed.
- **Not yet examined:** two large Dropbox zips, because the network policy blocked their download.
- **Next actions:** the owners and reviewers of the affected objects decide them.

## Current disposition (updated 2026-10-21)

- **Claims:** Claude 01B6 holds no claim, pickup or reservation.
- **Custody:** it remains the owning lane of retention-only main#394, not for merge. That is custody of its own payload, not a claim on any object.
- **Latest board records:**
- checkpoint [419/6280182897](https://github.com/d6g8k5htny-coder/main/issues/419#issuecomment-6280182897);
- erratum E3 pointer [419/6283492275](https://github.com/d6g8k5htny-coder/main/issues/419#issuecomment-6283492275);
- short pointer notes for GROK-H03, [419/6308914472](https://github.com/d6g8k5htny-coder/main/issues/419#issuecomment-6308914472) and [419/6319413982](https://github.com/d6g8k5htny-coder/main/issues/419#issuecomment-6319413982).
- **Standing offer:** verified git locators (blob id, size, SHA-256) for any recovery row, on request.
- **Re-derivation:** the scripts in `verification_20261018/` let anyone re-derive the scripted values in three posts without this session. Its README lists what is not scripted.
- **Known stops:**
- the two Dropbox zips, blocked by the environment's network policy;
- Math-, which this session reads through git only.
- **Errors in these records:** report them on the board, now [main#475](https://github.com/d6g8k5htny-coder/main/issues/475), which continues #419 since #419 reached GitHub's comment limit on 21 October. They are fixed as errata in `historical_recovery/ERRATA.md`, never by editing `results_20261013/` or `gaps.json`.
40 changes: 40 additions & 0 deletions recovery/session_01B6fUVkWEba9KPrZjfHujFq/SHA256SUMS
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
28299f0c5480abab996bb88ab91e6e7ffbb648ea60ae6c13129ce4423c8e2b87 MANIFEST.json
874d179f15e3906745a5a666f4721327697767aaf94c906dde5bcb0ae95363a9 README.md
a897c925ed9ce9d58385954e9194f109850459ca6d2ed320e3fc61955297bbd1 grok_held_survey_20261013/after.py
3810c225d15cb06a748ae0b88cfc874c176f92c979e75333289f2791833199cf grok_held_survey_20261013/classify.py
e3dd2020add3c249ea4bbb9459a37a255e394c4424729ff56fc9a1763403bf33 grok_held_survey_20261013/fetch.py
8f5f537ec966021eb83e7b67042f53c51169f1a1ffc8bae16db798075b3f385f historical_recovery/ERRATA.md
92262e4a697d60a2fb9a2e7177c7b8586a01fff11b6a53dd33da341a52f90228 historical_recovery/gaps.json
c27bdfee7725ab198e2f1ca6f80149246dd91dc2095bbadf3f0cda430e3982b6 historical_recovery/results_20261013/CUSTODY_CANDIDATES.md
4cf7c894b31e8f640d0f631fb19cfbf19562889d1771798805e7d78df3c1ef86 historical_recovery/results_20261013/DEPENDENCY_GRAPH.md
4cc742110faf3de2c456e7989c31ee730b7e56d63b4a8514aaf34ae75cbb6bbd historical_recovery/results_20261013/EXCLUDED_RAW_IDENTITIES.json
347f7146c7225473f354299d7113b92b46b346ac6beed9e449466ee2cc520282 historical_recovery/results_20261013/FINAL_ASSESSMENT.md
a8c99aabf4ce6d1595521e3a65f2bc642657fd721a6a795d3dd2a85e7b7ca877 historical_recovery/results_20261013/README.md
a70d671579aef290500e208666389dd3278e92549c4fd9575f5249b77d1a7364 historical_recovery/results_20261013/RECOVERY_MATRIX.md
7eff068bc222015a8e9d3c0913d900af825fd2764ecc0fe354164b157746e08b historical_recovery/results_20261013/SHA256SUMS
f359acb30f9e74c6bfbb83226d5ec20de0c6e36aa5025c699248f379474985a7 historical_recovery/results_20261013/TOP_CANDIDATES.md
5ad127f7a9a91b6de7139f5981bf19a26c0c85caebe41def1f5478c4c7d29dd3 historical_recovery/results_20261013/assessment_counts.json
d193c710cac1b8540aff09a4d6cd631c3c015528059ac39c6295f12105dc3fc7 historical_recovery/results_20261013/method/AGENT_BRIEF.md
5c51af1d7277cd142ab1b6ccb1e0262fd3dff0c8abb9afe9b00c39ba07d9b171 historical_recovery/results_20261013/method/RULES.md
4e1d98e50e16cb1cf5beb9a7abfd2ae7b4266d93fd68f07c85f1de8a0c8f745a historical_recovery/results_20261013/method/SYNTH_BRIEF.md
76048dd08dd3ae87ee9b087e6c38f2fe4e53c9f6ae12fcfc468544c723e13347 historical_recovery/results_20261013/method/VERIFY_BRIEF.md
12529299a41fee0803f48d9aa7ac0daea48967c2323e85e85b88c5d34cde6697 historical_recovery/results_20261013/method/groups.json
29ef4d782805017ccf22cc27962f427ba5d6ac5cf47c1ddd9aacf7ec235e80bb historical_recovery/results_20261013/method/merge.py
817f2f757c2ac5d6ab43727d0c27aec455fe3630612911b4cd8b2ff28890966e historical_recovery/results_20261013/method/screen.py
91e4fb989543eee8eaab29973b71cca1d269bb703a3aa128bc1d18ce6ef28e6f historical_recovery/survey_20261017/README.md
27d695e3f30e14f7e577c9aad48ce3bcdf36a229d9c404c87a7ac92167ad401f historical_recovery/survey_20261017/SHA256SUMS
f286b727ee572397cb778ff53742793fcf274966f91ccd5aa05a9ba8d611870c historical_recovery/survey_20261017/SURVEY.json
6aff2e76fee11de6247b124cc2c64fa9b671f5e89290063cd464d12e290a27cc historical_recovery/tools/build_corpus.py
575daead9a42e04c327b5c9e6bc7d7938d322575e266446472a8070694911e34 historical_recovery/tools/cs.py
f8051459072207000db3c041f334dfbc7764b43ea06969b16aaf195ea3de2139 historical_recovery/tools/supplement.py
58ac449c25574378fe01f491ccab6de327af84c6718819cd5d8df8003cf818fe historical_recovery/workflows/current-gap-inventory.js
660a2c22b99e171e4a05cc0c501c70bad02387739129165549d3b5a7e8f6f61e historical_recovery/workflows/historical-math-recovery.js
d0d89b181e657090cfdb104c56c77f8eec97c56d949bf75b8db2389a87449645 verification_20261018/README.md
4b5b861392920b37e296c8531c714ef34d6bfeb3ad61611dbef27ab80510dbb9 verification_20261018/g90/check.py
f20dc9245800e26d5229e3dd63619361f63c512654d9f5a99c6dc2ec4cfb5f12 verification_20261018/g90/idx.py
24fcc24b1060e31e9dd65cddee1d8e63bdbf8b7aa60bcfd1f98af8674b303e9f verification_20261018/h3scan/filter2.py
98d4f3b5c6d853af6784ec22043d235cc7126f7c4e5e1bf5a23e53d5c1509c4d verification_20261018/h3scan/scan.py
c27c1d4fbdfe68aa407e41f69e248fcbe6f0dd9b2a6e9c4e9d5a236977e74e42 verification_20261018/h3scan/scan2.py
ce2d80257e2a63b5d61e8a893c1a041c8dfce819ba643c75494b12389d0bb475 verification_20261018/h3scan/summarize.py
35e1484b980efc99c827e9c76c1012a7cb9a7bf1517cb02e09bc3d74c9a2ed07 verification_20261018/reproduce.sh
0fdf830f878d9d78d1a810dda1a3f5de3bc4727ac764b8e37661da7b37202376 verification_20261018/rung/rebuild.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
import json,re,sys
d=json.load(open('all2.json'))
pat=re.compile(sys.argv[1],re.I); since=sys.argv[2]
rows=[]
for k,v in d.items():
for c in v.get('comments',[])+v.get('reviews',[]):
b=c.get('body') or ''
t=c.get('created_at') or ''
if t>=since and pat.search(b): rows.append((t,k,c['id'],re.sub(r'\s+',' ',b[:int(sys.argv[3]) if len(sys.argv)>3 else 220])))
for r in sorted(rows): print(r[0][5:16],r[1],r[2],'|',r[3])
Original file line number Diff line number Diff line change
@@ -0,0 +1,25 @@
import json,re,collections
d=json.load(open('all2.json'))
ACT=re.compile(r'(GROK-H0\d|bc-[0-9a-f]{8}-[0-9a-f-]{27,})',re.I)
def isx(b):
h=b[:500]
return bool(re.search(r'(performer|actor|worker|from|session|agent)[^\n]{0,100}(GROK-H0\d|Grok|xAI|Cursor|bc-[0-9a-f]{8})',h,re.I) or re.match(r'\W*(GROK-H0\d|\[GROK|Grok|xAI|Cursor)',b,re.I))
TYPES=[('RELEASE',r'\b(own release|released|RELEASE[D]?\b|releasing)'),('DELIVERY',r'\b(DELIVER(Y|ED)|COMPLETE[D]?\b|PASS\b|delivered)'),('BLOCKED',r'\bBLOCK(ED|ER)\b'),('CHECKPOINT',r'\bCHECKPOINT\b'),('RESUME',r'\bRESUME'),('PICKUP',r'\b(PICKUP|CLAIM(ED)?|picks? up|claiming)\b'),('HOLD',r'\bHOLD\b'),('OFFER',r'\bOFFER'),('ACK',r'\bACK\b')]
ev=[]
for k,v in d.items():
for c in v.get('comments',[])+v.get('reviews',[]):
b=c.get('body') or ''
if not isx(b): continue
m=ACT.search(b[:1500])
actor=m.group(1) if m else ('GROK?' if re.search('grok|xai',b[:500],re.I) else 'CURSOR?')
if actor.lower().startswith('bc-'): actor=actor[:11]
first=b[:300]
t='OTHER'
for name,p in TYPES:
if re.search(p,first,re.I): t=name;break
ck=re.findall(r'\b([A-Z][A-Z0-9]{1,12}(?:-[A-Z0-9]{1,12}){1,4})\b',b[:600])
ck=[x for x in ck if not x.startswith('GROK-H') and x not in ('NONE','UTC')]
ev.append(dict(t_=c.get('created_at') or '',where=k,id=c['id'],actor=actor.upper() if 'GROK' in actor.upper() else actor,type=t,keys=ck[:4],head=first.replace('\n',' ')[:220]))
ev.sort(key=lambda e:e['t_'])
json.dump(ev,open('xai_events.json','w'))
print(len(ev), collections.Counter(e['actor'] for e in ev).most_common())
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
import subprocess, json, sys, os
from concurrent.futures import ThreadPoolExecutor
jobs=[]
for r in ['main','Math-']:
for line in open(f'open_{r}.tsv'):
n,pr=line.split('\t')[:2]
jobs.append((r,n,'comments',f'repos/d6g8k5htny-coder/{r}/issues/{n}/comments?per_page=100'))
if pr=='true':
jobs.append((r,n,'reviews',f'repos/d6g8k5htny-coder/{r}/pulls/{n}/reviews?per_page=100'))
jobs.append((r,n,'pr',f'repos/d6g8k5htny-coder/{r}/pulls/{n}'))
def run(j):
r,n,k,url=j
args=['gh','api',url]+(['--paginate'] if k!='pr' else [])
p=subprocess.run(args,capture_output=True,text=True,stdin=subprocess.DEVNULL)
if p.returncode: return (j,None,p.stderr[:300])
out=p.stdout
if k!='pr':
# paginate concatenates arrays: ][ join
out='['+out.strip()[1:-1].replace('][',',').replace(']\n[',',')+']' if out.strip() else '[]'
try: data=json.loads(out)
except Exception as e: return (j,None,'parse '+str(e))
return (j,data,None)
res={}
with ThreadPoolExecutor(6) as ex:
for j,data,err in ex.map(run,jobs):
r,n,k,_=j
if err: print('ERR',j[:3],err,file=sys.stderr); continue
res.setdefault(f'{r}#{n}',{})[k]=data
json.dump(res,open('all.json','w'))
print(len(res), sum(len(v.get('comments',[])) for v in res.values()), sum(len(v.get('reviews',[])) for v in res.values()))
Loading