WIP: replace obsolete lab governance with bounded research workflow #107

Closed
elsie wants to merge 3 commits from elsie/governance/research-reset into main AGit
Member

Cassie asked to replace the obsolete lab operating system because repeated authorization, fixed roles and historical prerequisites were preventing research. This reset replaces active instructions with a short workflow: approve the question/comparison and bounded resources once; assigned agents execute, preserve evidence and report meaningful exceptions. Agents own JSON/configs/approval receipts. Ordinary code/docs use relevant tests and plain-language PR approval. Removes fixed agent/device eligibility, seed/WIP/lifecycle quotas, the human-confirmation/state-JSON CLI, obsolete GitHub CODEOWNERS routing, and mandatory distinct-account review as a procedural rule for every PR. Historical scientific records and E01/R02 reproduction formats remain intact and explicitly non-normative for new work. Replaces the large lab-state validator with optional exact resource accounting; this helper grants no authority and does not pretend to enforce runtime limits. Scopes historical E01/DMS/confidence/citation integrity jobs to their dependency paths, including transitive episode serialization, while preserving the tests. Keeps general documentation/ledger checks and adds reset/accounting/scope regressions. Fixes documentation checking of tracked Markdown symlinks on Windows without changing the historical symlink. Validation: 17 targeted tests passed in owned invocation 19; final documentation check passed 107 Markdown files in invocation 22; results ledger is current for 24 runs. The checks used the previously verified Windows Job Object, four logical CPUs, hard 4 GiB memory and existing aggregate development limits; cleanup confirmed. The initial fixture failure is retained. Independent read-only review walked ordinary PR and bounded experiment flows, found missing transitive CI coverage and misleading guidance, and those findings were corrected. This is engineering review, not empirical admission. No models, GPU work, paid compute or downloads ran. No live branch protection, repository permissions, credentials or persistent signing settings changed. This PR is draft and not authorized for merge. Current branch protections require an owner-authenticated read before proposing exact settings changes. Historical PR #105 is untouched; it contains ten inherited diagnostic files that need separate resolution before integration. The reset grants no new experiment or spending permission; X14 W8 remains disabled.

Cassie asked to replace the obsolete lab operating system because repeated authorization, fixed roles and historical prerequisites were preventing research. This reset replaces active instructions with a short workflow: approve the question/comparison and bounded resources once; assigned agents execute, preserve evidence and report meaningful exceptions. Agents own JSON/configs/approval receipts. Ordinary code/docs use relevant tests and plain-language PR approval. Removes fixed agent/device eligibility, seed/WIP/lifecycle quotas, the human-confirmation/state-JSON CLI, obsolete GitHub CODEOWNERS routing, and mandatory distinct-account review as a procedural rule for every PR. Historical scientific records and E01/R02 reproduction formats remain intact and explicitly non-normative for new work. Replaces the large lab-state validator with optional exact resource accounting; this helper grants no authority and does not pretend to enforce runtime limits. Scopes historical E01/DMS/confidence/citation integrity jobs to their dependency paths, including transitive episode serialization, while preserving the tests. Keeps general documentation/ledger checks and adds reset/accounting/scope regressions. Fixes documentation checking of tracked Markdown symlinks on Windows without changing the historical symlink. Validation: 17 targeted tests passed in owned invocation 19; final documentation check passed 107 Markdown files in invocation 22; results ledger is current for 24 runs. The checks used the previously verified Windows Job Object, four logical CPUs, hard 4 GiB memory and existing aggregate development limits; cleanup confirmed. The initial fixture failure is retained. Independent read-only review walked ordinary PR and bounded experiment flows, found missing transitive CI coverage and misleading guidance, and those findings were corrected. This is engineering review, not empirical admission. No models, GPU work, paid compute or downloads ran. No live branch protection, repository permissions, credentials or persistent signing settings changed. This PR is draft and not authorized for merge. Current branch protections require an owner-authenticated read before proposing exact settings changes. Historical PR #105 is untouched; it contains ten inherited diagnostic files that need separate resolution before integration. The reset grants no new experiment or spending permission; X14 W8 remains disabled.
Record owner X14 merge and current governance handoff
All checks were successful
checks / select relevant integrity checks (pull_request) Successful in 6s
checks / E01 candidate surface sufficiency (pull_request) Has been skipped
checks / DMS-003 protocol freeze integrity (pull_request) Has been skipped
checks / DMS-003 CPU successor integrity (pull_request) Has been skipped
checks / resource accounting and workflow consistency (pull_request) Has been skipped
checks / documentation integrity (pull_request) Successful in 6s
26db9e814c
Record completed owner governance merge
All checks were successful
checks / select relevant integrity checks (pull_request) Successful in 6s
checks / E01 candidate surface sufficiency (pull_request) Has been skipped
checks / DMS-003 protocol freeze integrity (pull_request) Has been skipped
checks / DMS-003 CPU successor integrity (pull_request) Has been skipped
checks / resource accounting and workflow consistency (pull_request) Has been skipped
checks / documentation integrity (pull_request) Successful in 6s
9cd98afe94
Author
Member

Superseded by the governance reset and later operational cleanup already on current main. The only remaining branch change is docs/governance-reset-review.md, deliberately removed on main; it will not be restored. Historical review remains at commit 9cd98afe94 and this PR discussion. Closing the superseded proposal preserves its branch and evidence; no research is restarted. Active instructions remain those on main after PR118; its selective archive verification confirmed no operational files were changed.

Superseded by the governance reset and later operational cleanup already on current main. The only remaining branch change is docs/governance-reset-review.md, deliberately removed on main; it will not be restored. Historical review remains at commit 9cd98afe946033acab1481b56cb210cae9cf21ed and this PR discussion. Closing the superseded proposal preserves its branch and evidence; no research is restarted. Active instructions remain those on main after [PR118](https://durandal.exe.xyz/smolmodelco/thesmolmodelcompany/pulls/118); its selective archive verification confirmed no operational files were changed.
elsie closed this pull request 2026-10-18 03:36:53 +00:00
All checks were successful
checks / select relevant integrity checks (pull_request) Successful in 6s
checks / E01 candidate surface sufficiency (pull_request) Has been skipped
checks / DMS-003 protocol freeze integrity (pull_request) Has been skipped
checks / DMS-003 CPU successor integrity (pull_request) Has been skipped
checks / resource accounting and workflow consistency (pull_request) Has been skipped
checks / documentation integrity (pull_request) Successful in 6s
Required
Details

Pull request closed

Sign in to join this conversation.
No reviewers
No labels
R03
research-b
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
smolmodelco/thesmolmodelcompany!107
No description provided.