Claims and follow-up
The record
Stable claim IDs connect the evidence, our current judgment, and the tests that could change it.
We track testable claims about the trajectory; no resolved local claim would by itself establish artificial wisdom. The provisional rationale explains the further links to investigate.
Record format v1.0 · October 15, 2026 · Rubric SI-1.0. Claim IDs are never renumbered or reused. A correction fixes an error; an update adds evidence; a forecast revision changes an expectation. Earlier wording and reasons remain accessible.
Issue #1 remains an unsent editorial draft. The question versions below were frozen for preservation on October 15; no pre-outcome scoring record is claimed. Draft forecasts remain outside performance claims.
Challenges and responses are retained. A challenger is credited as “a reader” unless they explicitly consent to named attribution. No reader challenges have been recorded here.
Claim ID math-release · claim version 1.0
Context: OpenAI’s October 6 mathematics-manuscript release.
Claim. This release could materially accelerate mathematical research through novel, correct, reusable results.
Current status. Unresolved
Evidence that could change the assessment. Look for expert reviews of the catalogue’s groups, publicly usable proof files, further withdrawals, and results that other researchers build on. A system outsiders can reproduce on problems chosen in advance would provide a stronger test. Reliable new results would strengthen the acceleration claim; persistent errors or failure to transfer would weaken it.
Next review. November 30, 2026, or a material withdrawal, expert review, or independent reproduction.
Judgment and correction history · Issue assessment
No separate reader challenge or resolved falsifier is recorded. This is a follow-up commitment, not a completed test.
Claim ID research-loop · claim version 1.0
Claim. Current frontier agents can already reproduce consequential algorithmic innovation end to end without substantive human intervention.
Current status. Unresolved
Evidence that could change the assessment. Look for repeated runs on several tasks chosen before testing, with credible human baselines and independent checks of novelty and results. Reproducible success across fresh research problems would change this assessment more than a higher coding score.
Next review. A refreshed InnovationEval or comparable independent multi-task study; otherwise December 31, 2026.
Judgment and correction history · Issue assessment
No separate reader challenge or resolved falsifier is recorded. This is a follow-up commitment, not a completed test.
Claim ID ebr-learning · claim version 1.0
Context: the legal card combination removed the game’s time pressure; banning it changed the test conditions.
Claim. A perfect EBR-bench score establishes general learning or autonomous scientific discovery.
Current status. Unresolved
Evidence that could change the assessment. Look for repeated performance under the revised rules and tests in other environments. We have not measured public attention to the result, so the evidence does not support a resolved false-alarm judgment.
Next review. A revised benchmark report or cross-environment study; otherwise December 31, 2026.
Judgment and correction history · Issue assessment
No separate reader challenge or resolved falsifier is recorded. This is a follow-up commitment, not a completed test.
Claim ID uv-map · claim version 1.0
Claim. AI assistance makes deferred scientific data integration economically practical.
Current status. Unresolved
Evidence that could change the assessment. Look for independent checks of the map’s predictions and uncertainty, reuse by researchers, and replicated savings in time or cost across projects. Those results would support a specific research-productivity claim.
Next review. Independent validation or documented reuse; otherwise January 31, 2027.
Judgment and correction history · Issue assessment
No separate reader challenge or resolved falsifier is recorded. This is a follow-up commitment, not a completed test.
Claim ID evaluation-integrity · claim version 1.0
Claim. Evaluation infrastructure can undermine reliable observation even when authoritative logs survive.
Current status. Unresolved
Evidence that could change the assessment. Look for independent security reviews, preserved audit trails, and similar weaknesses in other evaluation tools. Evidence of an agent initiating such an exploit would support a stronger, separate claim.
Next review. A relevant independent security review or observed incident; otherwise January 31, 2027.
Judgment and correction history · Issue assessment
No separate reader challenge or resolved falsifier is recorded. This is a follow-up commitment, not a completed test.
Forecast questions
Question ID formalization · question version 1.0 · rubric SI-1.0
Resolution criteria. Met if at least one top-line result in the October 6 release has a publicly accessible formal proof added after the October 8 cutoff, with a recorded proof-checking command and successful run documented by a reviewer unaffiliated with OpenAI. Supporting lemmas alone do not count.
Deadline. 2026-11-30
Draft, unscored. A substantive wording or criteria change requires a linked successor question; the original scoring basis must remain.
Question ID research-overclaim · question version 1.0 · rubric SI-1.0
Resolution criteria. The skeptical forecast is overturned if an evaluator independent of the developer publishes reproducible success on at least five fresh, preselected ML research tasks, matching or exceeding documented human innovations under declared comparable resources, with no substantive human interventions after launch. Novelty, training-data exposure, failed runs, code, and outcome verification must be reported.
Deadline. 2026-12-31
Draft, unscored. A substantive wording or criteria change requires a linked successor question; the original scoring basis must remain.
Send evidence or a challenge through the publisher’s contact page, quoting the claim ID. Named public credit requires your consent.