# Claims and evidence

## C01

October 6 initial collection included 722 manuscripts in 372 related research families

Evidence: source:openai-math. Status: source_supported. As of: 2026-10-06.

## C02

October 7 log documents 3 withdrawn manuscripts and 14 revised manuscripts, plus 13 follow-on citation updates

Evidence: source:openai-history-oct7. Status: source_supported. As of: 2026-10-07.

## C03

October 7 formalized top-line coverage reported 300/719 (~42%)

Evidence: source:openai-history-oct7. Status: source_supported. As of: 2026-10-07. Coverage is not independent verification of natural-language correspondence.

## C04

OpenAI reports Navier–Stokes successful group approximately 10,000 concurrent agents, 130 billion output tokens, and about 88 hours plus 17 hours formalization

Evidence: source:openai-navier. Status: lab_report_not_independent. As of: 2026-09-08. 300 billion output tokens is across all attempted problems in September, not the single Navier–Stokes attempt.

## C05

Bastounis, Circelli and Hansen publicly contest semantic faithfulness of OpenAI's Navier–Stokes formalization

Evidence: source:bastounis-oct6. Status: published_preprint_claim. As of: 2026-10-06. Do not treat critique as consensus or theorem-status finality.

## C06

AGMAI cautioned that release begins the process of community understanding and was not its endorsement of impact or process

Evidence: source:agmai-oct6. Status: source_supported. As of: 2026-10-06.

## C07

Epistemic debt can arise when claim generation and dependencies outpace validation

Evidence: C01, C02, C03. Status: interpretive_thesis. As of: 2026-10-08.

## C08

Metacognition for autonomous systems should be evaluated through behavior under contradiction, changed evidence, and corrected dependencies

Evidence: C05, C07. Status: untested_here. As of: 2026-10-08.
