ProofAtlas.ai and zero0_one1 claim a ProofAtlas harness using GPT-5.6 Pro improved the lower bound for Moser’s convex worm problem from 0.2322 to greater than 0.2374.
state: expiredheat: lowuncertainty: highconvergesscott: mediumai-assisted-mathematics agent-harnessesProofAtlas.aizero0_one1OpenAI
What is this?
ProofAtlas.ai and zero0_one1 report that a ProofAtlas agent harness using OpenAI’s GPT-5.6 Pro produced a Lean-verified lower bound greater than 0.2374 for Moser’s convex worm problem, improving the previously published bound of about 0.232239. The supplied Wikipedia snippet confirms the older bound and says the problem remains open. However, the other search results concern a different zeroth-order convex-optimization result, so they do not independently substantiate the ProofAtlas claim or the assertion that it closes the broader theoretical gap.
Why it matters to Scott
The claimed workflow concretely converges with Scott’s verification-loop architecture: an agent harness generates mathematical work while Lean supplies a mechanically different, deterministic acceptance boundary. A second ProofAtlas result could strengthen the pattern already tracked in “Proof Atlas Collatz formalization,” but the supplied evidence remains announcement-level and lacks independent expert review or reproduction.
ip:concept.verification-loopsip:concept.mechanically-different-verifiersip:concept.evidence-class-ladderip:framework.agent-loopradar:proofatlas-collatz-formalizationradar:concept.formal-verificationradar:concept.mathematical-discoveryradar:concept.ai-mathematics
queries asked of Scott's wikis
- agent harnesses for mathematical discovery
- Lean verification as an AI-agent feedback loop
- formal proof systems for grounding LLM outputs
- long-context prompting versus iterative agent workflows
- evidence standards for claimed AI scientific breakthroughs
- human-model division of labor in theorem proving
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
2026-09-09T19:40:11Z
Repeated reviews have produced no inspectable proof artifact, independent assessment, or evidence of the agent’s contribution, and no concrete follow-up is expected. Retire this announcement from active monitoring without treating the mathematical claim as disproved; a reproducible formalization or substantive expert review would warrant reopening it.
2026-09-07T19:37:41Z
This remains an unverified mathematical-result announcement, not yet evidence of a reproducible agent–verifier discovery workflow; the staleness check adds no substantive support or contradiction. OpenAI’s standing does not validate a third party’s attribution to its model, and the claimed formalization still needs inspection separately from the claimed AI contribution.
2026-09-05T18:31:43Z
No substantive new evidence changes the announcement-level claim; the reported Lean verification still needs inspection of the theorem statement, assumptions, and reproducible artifact, separately from attribution of the work to the agent. Elapsed silence neither disproves the result nor reduces its potential significance if validated.
2026-09-03T17:59:51Z
The added attention supplies no independent validation, expert review, or reproduction of the claimed bound or AI contribution. The case remains a bounded but announcement-level claim, so the initial alert potential has cooled.
2026-09-03T17:39:37Z
grounded: converges/medium — The claimed workflow concretely converges with Scott’s verification-loop architecture: an agent harness generates mathematical work while Lean supplies a mechan
2026-09-03T17:36:13Z
case created — The numerical theorem claim is bounded and falsifiable and attributes the result to a named AI research harness and model.
Decision trace
- 09-10 05:40expireRepeated reviews have produced no inspectable proof artifact, independent assessment, or evidence of the agent’s contribution, and no concrete follow-up is expected. Retire this announcement from acti
- 09-10 05:40alert_silentThe new observation adds only engagement, not a consequential factual delta. There is no scheduled confirmation to wait for and no new information Scott needs today.
- 09-10 05:40alert_routeThe new observation adds only engagement, not a consequential factual delta. There is no scheduled confirmation to wait for and no new information Scott needs today.
- 09-08 05:37repriceThis remains an unverified mathematical-result announcement, not yet evidence of a reproducible agent–verifier discovery workflow; the staleness check adds no substantive support or contradiction. Ope
- 09-08 05:37alert_silentNo new artifact, independent assessment, or consequential workflow detail has arrived since the previous review. Repeating the announcement would add no decision value; its potential significance rema
- 09-08 05:37alert_routeNo new artifact, independent assessment, or consequential workflow detail has arrived since the previous review. Repeating the announcement would add no decision value; its potential significance rema
- 09-06 04:31repriceNo substantive new evidence changes the announcement-level claim; the reported Lean verification still needs inspection of the theorem statement, assumptions, and reproducible artifact, separately fro
- 09-06 04:31alert_silentThe staleness trigger brings no new artifact, expert assessment, or workflow evidence to surface today. The earlier announcement remains potentially relevant, but this look adds no consequential delta
- 09-06 04:31alert_routeThe staleness trigger brings no new artifact, expert assessment, or workflow evidence to surface today. The earlier announcement remains potentially relevant, but this look adds no consequential delta
- 09-04 09:21sensor_dirtyengagement_update
- 09-04 05:21sensor_dirtyengagement_update
- 09-04 03:59repriceThe added attention supplies no independent validation, expert review, or reproduction of the claimed bound or AI contribution. The case remains a bounded but announcement-level claim, so the initial
- 09-04 03:59alert_silentOnly Reddit engagement increased; no consequential factual delta has appeared since the prior routing, so another alert would repeat the same unverified claim.
- 09-04 03:59alert_routeOnly Reddit engagement increased; no consequential factual delta has appeared since the prior routing, so another alert would repeat the same unverified claim.
- 09-04 03:53alert_shadowA concrete formalization is reportedly available now for inspection, making this more than a vague capability claim and directly relevant to Scott’s agent-plus-deterministic-verifier architecture. The
- 09-04 03:53alert_routeA concrete formalization is reportedly available now for inspection, making this more than a vague capability claim and directly relevant to Scott’s agent-plus-deterministic-verifier architecture. The
- 09-04 03:39groundThe claimed workflow concretely converges with Scott’s verification-loop architecture: an agent harness generates mathematical work while Lean supplies a mechanically different, deterministic acceptan
- 09-04 03:36createThe numerical theorem claim is bounded and falsifiable and attributes the result to a named AI research harness and model.