Redditor links OpenAI dropping Millennium Prize efforts to week-long Codex quality dip
Maleficent_Disk9583 · reddit · 2026-09-26
A Reddit user offers a speculative timeline around OpenAI's Millennium Prize Problem claims: after OpenAI announced it had solved one, a researcher accused Codex of copying findings he had typed in, possibly ingested via training data. OpenAI denied this.
- Codex then showed roughly a week of widely noticed quality degradation, with some speculating OpenAI was burning compute trying to solve more prize problems to prove it didn't steal
- On the same day quality recovered, OpenAI announced it would stop targeting Millennium Prize problems "out of respect for mathematicians"
The author argues the timing suggests capability limits, not respect. This is community speculation, not verified.
More from Companies & People
- How AI Rewrote the PM Interview: Evals and 'Show What You've Built' — aakashgupta · 2026-09-26
- Sriram Krishnan says DHH's AI speech for engineers will 'be seen as prophetic' — rseroter · 2026-09-26
- How Law Firms are Becoming AI Firms — Coolnumber11 · 2026-09-26
- AI labs seek vaccine-style liability shield from Washington; Treasury Secretary refuses — jbarbier · 2026-09-26
- Suspected Altman alt account teases dizzying model and hardware drops within six weeks — iruletheworldmo · 2026-09-26
- Altman Admits OpenAI's Agent Internet-Access Review Is Slower Than Hoped; Hugging Face Still Most Severe — CurieuxExplorer · 2026-09-26