Skip to content

docs(plans): record the maintainer's 2026-10-08 decisions in the v2 plan - #207

Merged
pdlourenco merged 4 commits into
mainfrom
claude/charming-gauss-vglm3j
Oct 8, 2026
Merged

pdlourenco merged 4 commits into
mainfrom
claude/charming-gauss-vglm3j

Conversation

@pdlourenco

@pdlourenco pdlourenco commented Oct 8, 2026 •

Copy link
Copy Markdown
Owner

Summary

Records the maintainer's 2026-10-08 decisions in docs/plans/2026-10-04-review-v2-remediation-plan.md.

Stacked on #206. This PR is based on #206's branch, because the §0.2 decision table it records against exists only there. Merge #206 first. GitHub then retargets this PR to main, and its diff becomes these three commits.

All seven decision items the must-do set needs are now taken. They are logged in the new §0.3 and marked in the inventory, §0.2, item texts, §9 and §10:

  • A1: require the six listed contexts plus the fork-PR token guard; keep the admin bypass.
  • A2: option (i), no commit on main; (i-a) the PR's MATLAB job pushes regenerated vectors to the PR branch, (i-b) the maintainer commits them for fork PRs; PR-time ancestry and canonical-engine provenance check. The maintainer's "A2 — agree" is read as covering all three §0.2 A2 rows, and the log says so.
  • A3: string sentinels for NaN/Inf on disk.
  • A6: if the IO path is engine-sensitive, pin the cost trajectory and iteration count.
  • A10: the documented floor stays Octave 11; both Octave CI jobs run in a gnuoctave/octave:11.3.0 container, the version on kynes, instead of the unpinned snap. The suggested docs wording ("GNU Octave 11+, CI: 11.3.0") is marked as not yet confirmed.
  • D2: the §14.5/§15.4 additions.
  • E1b IO cost fields:
    • The IO per-iteration history is renamed CostHistory / cost_history, accepting the released-API break (0.x, no user base). This reverses the plan's recommendation to keep the name.
    • Cost becomes the §8.5 (1 × 3) [total, data_fidelity, regularization] breakdown at the returned estimate, so Cost(1) == CostHistory(end).

One sequencing choice left open. A6 can store CostHistory and the new Cost only once E1b's IO cost-field change is in place, and the rest of E1b still waits on open rows. The plan recommends carrying that one change in A6's PR, spec commit first.

Spec sections touched

none — no algorithmic change (plan document only). The decisions it records change §14.5/§15.4 (D2) and §8.12.9 (E1b) when their item PRs land, spec first.

Type of change

  • Docs / tooling / CI only

Checks

  • Major / architectural decision made? Recorded in the plan's §0.3 decision log. The superseding ADRs (A1 vs ADR-0005; A2/A3 vs ADR-0002 and ADR-0005's bot bypass) land with their implementing PRs, per the plan's §0.2 preamble. For A10 and E1b's IO cost fields the plan recommends no ADR.
  • Pre-push review: four rounds.
    • Round 1 flagged 4 blockers. The record over-claimed A10 and the CostHistory name as decided before the maintainer confirmed them. The commit was rebuilt from scratch.
    • Round 2 found no blockers and 6 nits, fixed in 175e511.
    • Round 3 found no blockers and 8 nits, fixed in 7cbc910.
    • Round 4, on the Cost decision, found no blockers and 3 nits, plus the A6 sequencing gap. All fixed in ce22ad4; the gap is recorded as a recommendation, not a decision.
  • Local CI: docs-only plan edit outside the published site; table cell counts checked on every edited table. CI runs once this PR targets main.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Wz15nbQvZAdxgAMJ9nkuvf

claude added 3 commits October 8, 2026 09:19
New §0.3 decision log; §0.2 rows, inventory markers, item texts, §9
and §10 updated to match.

Decided: A1 (six required contexts + fork-PR token guard), A2 (option
(i) with (i-a)/(i-b) and the PR-time provenance check), A3 (string
sentinels), A6 (pin the cost trajectory if engine-sensitive), D2 (the
§14.5/§15.4 additions), and in E1b the rename of the IO per-iteration
history away from Cost, with the released-API break accepted.

Awaiting confirmation: A10 (the maintainer's input that the 11 floor
follows the Octave on kynes; recommended to keep 11 and pin both CI
jobs to a gnuoctave/octave container at that exact version), the
history's new name (recommended CostHistory / cost_history), and what
the IO Cost then holds (recommended the §8.5 1x3 breakdown).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wz15nbQvZAdxgAMJ9nkuvf
The maintainer confirmed A10 (the documented floor stays Octave 11;
both Octave CI jobs run in a gnuoctave/octave:11.3.0 container, the
version on kynes) and the name of the renamed IO per-iteration
history (CostHistory / cost_history). The must-do set now has six of
its seven decision items taken; what the IO Cost holds after the
rename is still open, with the recommendation and alternatives stated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wz15nbQvZAdxgAMJ9nkuvf
The maintainer accepted the recommendation for E1b's IO Cost: after
the history is renamed CostHistory, Cost becomes the §8.5 (1 x 3)
[total, data_fidelity, regularization] breakdown at the returned
estimate, so Cost(1) == CostHistory(end). With it, all seven decision
items the must-do set needs are taken; the open-confirmation list in
§0.3 is removed and A6 no longer waits on a decision.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wz15nbQvZAdxgAMJ9nkuvf
Base automatically changed from claude/review-v2-plan-execution-refinements to main October 8, 2026 09:56
Brings in the #206 merge commit now that #207 targets main.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wz15nbQvZAdxgAMJ9nkuvf
@pdlourenco
pdlourenco merged commit 1cdb5fd into main Oct 8, 2026
6 checks passed
@pdlourenco
pdlourenco deleted the claude/charming-gauss-vglm3j branch October 8, 2026 10:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants