Skip to content

OAI26 F011: Lean-prove external totient fiber definition equivalence and infinitude transfer - #549

Merged
fyremael merged 5 commits into
mainfrom
review/oai26-f011-positive-fiber-count-equiv-20261111
Nov 11, 2026
Merged

fyremael merged 5 commits into
mainfrom
review/oai26-f011-positive-fiber-count-equiv-20261111

Conversation

@fyremael

Copy link
Copy Markdown
Collaborator

Bounded source-to-statement mathematical closure; not analytic proof certification. At external cadamcat/totient-fibers-lean4@f75d4f2c2d6685a0547f17fa07240377aa2c2ed4, Erdos821/Challenge.lean (blob ac85c9a6a0878d18de3d923ca1a14e8213089c03) defines g n := {m | Nat.totient m=n}.ncard; pinned GCL openai/math@1581d4cbdd69a0faf34b2089614fe2a814316a6a (blob 217cfe4d7f9c39981eabe9bbeeb8458febb0fc76) defines g n over strictly positive preimages. This PR proves in Lean 4.34.1: (1) the counts coincide for every n>0 (because Nat.totient 0=0); (2) removing the exceptional target {0} preserves infinitude; and (3) the complete external all-ε>0 Erdős 821 statement implies GCL's all-ε>0 positive-target statement. Compiler exit 0, 3/3 #print axioms reports exactly propext, Classical.choice, Quot.sound; source SHA256 30886e87944de53e0f681ccfe8d6d4a71bd5fe6607abe16e44b524bf3153c31c, log SHA256 10f421668aa69af088709ba075d72a3761edf706c593120da9f9b69ea510f165, 9/9 hostile custody tests pass locally. Exact source, proof and CI owner committed. This closes the source-statement discrepancy identified in WP03 after PR #548. Scope: the external theorem proof remains a distinct Colab source-locked worker gcl-f011-xu-replay4-20261111, with the full external Erdos821.Final.supply / Erdos821.erdos_821 kernel replay still pending; this PR does not import or independently validate that external proof. No MATHCERT, full paper proof acceptance, or programme closure. Requires exact-head CI, scoped adversarial review and protected merge.

@fyremael
fyremael requested a review from a team as a code owner November 11, 2026 13:56

Copy link
Copy Markdown
Collaborator Author

GCL-OAI26-F011/EXTERNAL-FIBER-SEMANTICS-CRITICAL-REVIEW/1 — scope exactly grandchallenge/MATHFORGE PR #549, head aac8d3eacc4ccaac71aa0a58d13cbc7f79ac5068, protected base e1ee3957b54d4081bcbbda4bf3866056f41c935a.

Semantics: The published Erdos821/Challenge.lean external count {m:ℕ | Nat.totient m=n}.ncard includes zero as a preimage of zero. GCL's pinned OAI.TotientFibers.g n counts {m | 0<m ∧ Nat.totient m=n}.ncard. The new Lean proof explicitly derives their equality for each n>0, using Nat.totient 0=0; it then removes only the singleton target zero from an infinite set (a finite deletion), and transfers both infinitude for each fixed ε and the full all-ε>0 statement. The resulting theorem is GCL.OAI26.Family011.gcl_claim_of_external_claim, not a proof that the external theorem itself holds. No weakening of the unbounded ε quantifier, positivity, or strict > inequality.

Exact replay: source file SHA-256 30886e87944de53e0f681ccfe8d6d4a71bd5fe6607abe16e44b524bf3153c31c matched branch upload byte-for-byte. Under openai/math@1581d4cbdd69a0faf34b2089614fe2a814316a6a, manifest SHA-256 cf6105a25d9dca2f166b241d9191bd12c7e13305890dc9c4d0952351cccc0794, pinned Lean 4.34.1, compiler exit 0 and all three #print axioms exactly [propext, Classical.choice, Quot.sound], no sorryAx, placeholders or new axioms. Log SHA-256 10f421668aa69af088709ba075d72a3761edf706c593120da9f9b69ea510f165. Nine source/receipt hostile tests pass at updated exact branch head, including source swap, false kernel claim, fabricated external success, missing theorem, removed zero-exclusion step and sorryAx perturbation. The first test run found a fixture bug (receipt key execution was erroneously addressed under proof); it was fixed at head aac8... and freshly replayed 9/9. Do not count the failed preliminary fixture run as acceptance.

Governance: This is a source/statement equivalence proof. The external analytic theorem Erdos821.Final.supply is still independently being rebuilt on the existing source-locked CPU Colab job gcl-f011-xu-replay4-20261111, not proved by this PR. No full paper review, independent external reviewer, or MATHCERT certification. Reviewer is the same authenticated operator in a critical logical role, not a separate non-author identity. As of this review, required PR checks have not all become terminal, so protected admission must wait for exact-head required CI. Disposition: F011_ZERO_EXCEPTION_LEAN_EQUIVALENCE_PASS / PROTECTED_PR_ADMISSION_PENDING_CI.

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