fcesco claims GPT-6 Astra produced a valid proof of the Berge–Fulkerson conjecture for C(20) graphs, which would extend frontier-model-assisted theorem proving to a new bounded graph-theory result.
state: expiredheat: lowuncertainty: highknownscott: lowgpt-6-astra automated-theorem-proving formal-mathematicsfcescoOpenAI
What is this?
The Berge–Fulkerson conjecture states that every 2-connected or bridgeless cubic graph has six perfect matchings such that each edge appears in exactly two of them. The case attributes to fcesco a claim that “GPT-6 Astra” proved the conjecture for a bounded class called C(20), but the supplied results neither identify that model nor define C(20), provide the proof, or independently validate the claim. On this evidence, the reported bounded result remains unverified, and the general conjecture remains open.
Why it matters to Scott
Scott’s Evidence Package, Verification Loops, and Provenance-Coupled Work pages already require inspectable artifacts and independent checks for claims like this. With no supplied proof, model identification, definition of C(20), or validation, this is only another unverified AI-mathematics claim rather than evidence that would update his position or practice; the radar already tracks closely analogous theorem-proof claims.
ip:concept.evidence-packageip:concept.verification-loopsip:framework.provenance-coupled-workip:concept.mechanically-different-verifiersradar:gpt-5-6-cycle-double-cover-proofradar:concept.theorem-provingradar:concept.formal-verificationradar:concept.ai-mathematics
queries asked of Scott's wikis
- LLM-assisted theorem proving and proof verification
- formal proof artifacts versus natural-language proofs
- evaluation of frontier models on novel mathematics
- bounded-case proofs as evidence of mathematical reasoning
- trust and reproducibility for AI-generated proofs
- agents using theorem provers or proof assistants
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (1) — ⭐ canonical anchor
Interpretation history
2026-09-06T23:49:50Z
No new evidence since creation beyond staleness timeout; claim remains a single unverified low-engagement post with no proof artifact, model ID, or validation. Nothing to track further absent new evidence.
2026-09-04T23:31:55Z
The reevaluation adds no substantive evidence: the claim still lacks a proof artifact, a definition of C(20), model identification, or independent validation, while the tiny engagement increase is immaterial.
2026-09-04T23:30:04Z
grounded: known/low — Scott’s Evidence Package, Verification Loops, and Provenance-Coupled Work pages already require inspectable artifacts and independent checks for claims like thi
2026-09-04T23:26:56Z
case created — The specific mathematical claim is resolvable and distinct from the existing case about GPT-6 Astra's safety controls.
Decision trace
- 09-07 09:49expireNo new evidence since creation beyond staleness timeout; claim remains a single unverified low-engagement post with no proof artifact, model ID, or validation. Nothing to track further absent new evid
- 09-07 09:49alert_silentUnverified single-source claim with zero comments and negligible score; no independent corroboration or artifact has appeared, and repeated review cycles have added nothing new.
- 09-07 09:49alert_routeUnverified single-source claim with zero comments and negligible score; no independent corroboration or artifact has appeared, and repeated review cycles have added nothing new.
- 09-05 09:31repriceThe reevaluation adds no substantive evidence: the claim still lacks a proof artifact, a definition of C(20), model identification, or independent validation, while the tiny engagement increase is imm
- 09-05 09:31alert_silentNo consequential delta has occurred since the prior review; a one-point score increase without comments or new evidence does not justify interrupting Scott or change the unverified status of the claim
- 09-05 09:31alert_routeNo consequential delta has occurred since the prior review; a one-point score increase without comments or new evidence does not justify interrupting Scott or change the unverified status of the claim
- 09-05 09:30alert_silentA lone low-engagement headline linking to an uninspected social post provides no proof artifact, model identification, definition of C(20), or validation, and the claimant has no supplied domain stand
- 09-05 09:30alert_routeA lone low-engagement headline linking to an uninspected social post provides no proof artifact, model identification, definition of C(20), or validation, and the claimant has no supplied domain stand
- 09-05 09:30groundScott’s Evidence Package, Verification Loops, and Provenance-Coupled Work pages already require inspectable artifacts and independent checks for claims like this. With no supplied proof, model identif
- 09-05 09:26createThe specific mathematical claim is resolvable and distinct from the existing case about GPT-6 Astra's safety controls.