Skip to content

docs: record the 011.04 closure and how to verify its commitment chain - #50

Merged
drernie merged 1 commit into
mainfrom
docs/011-closure-and-provenance-refs
Sep 13, 2026
Merged

drernie merged 1 commit into
mainfrom
docs/011-closure-and-provenance-refs

Conversation

@drernie

@drernie drernie commented Sep 13, 2026

Copy link
Copy Markdown
Contributor

Why

Two gaps surfaced while merging #49.

The experiments README stopped at 011.02's disposition C. main now carries the turn that closed Issue 011, but its README still described the learner as reaching only 5 of 8 seeds and stated H_discovery_global_finite is NOT banked. Anyone reading main got the superseded half of the story.

Two of the eight fast checks cannot pass from main. fano_optimization_evaluate.py and fano_optimization_analysis.py assert that the recorded pre-evaluation commit is an ancestor of HEAD. #49 was squash-merged, so on main both fail with recorded pre-evaluation commit is not an ancestor of HEAD. No number or artifact differs — but the commit boundary that proves selection preceded evaluation is not reachable from a squashed history, and that boundary is the whole anti-cheating argument of 011.04 §4.

What this adds

  • An 011.04 section: the unchanged-metric table against 011.02 (5/8 → 16/16, every threshold at 1.0000), the four-layer separation (identifiability / representation / optimization / generalization) that makes it a closure rather than a redesign, the closure files, and the closure fence.
  • Three cautions a reader needs and would not get from the metric block alone: Query A is cheaply exact so the evidence is Query B, fano_plane_valid entails certified_plane_equivalent so two rows are one finding, and selection read the already-public development seeds.
  • A "Verifying the commitment chain — not from main" subsection naming the provenance/* tags and the retained experiment branches to run those two checks from.
  • A superseded-note under the 011.02 heading, matching the idiom already used for 009.02 → 009.06.
  • Three internal anchors fixed: they used 009-06 style where GitHub's slugger emits 00906.

Related ref changes (already pushed, not part of this diff)

Nine annotated provenance/* tags now pin every commit 011.04 §11 cites, plus the two commits 009 cites that no ref on origin reached. Both experiment branches are restored, and delete_branch_on_merge is off so future squashes stop orphaning cited commits.

🤖 Generated with Claude Code

The experiments README stopped at 011.02's disposition C, so main described the
011 learner as unreliable while carrying the turn that closed it. Add an 011.04
section: the unchanged-metric table against 011.02, the four-layer separation
that makes it a closure rather than a redesign, the closure files, and the three
cautions a reader needs (Query A is cheaply exact, plane-validity entails
certified equivalence, selection used the already-public development seeds).

Also document that fano_optimization_evaluate and fano_optimization_analysis
cannot pass from main. Both assert the recorded pre-evaluation commit is an
ancestor of HEAD, and the squash merge in #49 leaves no such ancestor. No number
or artifact differs; the commit boundary proving selection preceded evaluation
is simply not reachable from a squashed history. The section names the preserved
provenance/* tags and the retained experiment branches to run them from.

Fix three internal anchors that used 009-06 style where GitHub's slugger emits
00906.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@drernie
drernie merged commit afd3868 into main Sep 13, 2026
7 checks passed
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.

1 participant