What's this about?
NYU mathematician Tristan Buckmaster says OpenAI began a massive parallel effort after rumours of his private work with Levent Alpöge, then proposed that Alpöge — an Anthropic employee collaborating personally — be removed from authorship. OpenAI says it never saw their work, recognises their forced-Euler priority, and produced materially different proofs.
What exactly is an “academic priority” dispute?
Priority determines who first made a discovery and receives scholarly credit. It is not resolved only by who publishes first: private circulation, independent derivation, shared ideas and proper attribution all matter, especially when one party can spend vastly more compute to finish quickly.
What's the community saying?
HN 157 points and 58 comments, plus extensive discussion inside the main 981-comment thread.
- Many considered OpenAI's most sympathetic timeline — hearing a rumour and racing to finish first — unacceptable research conduct even if the proof was independently generated.
- Others said the decisive technical question is how different the proofs and exact results really are.
- Users focused on OpenAI's sentence that it cannot rule out de-identified product data helping train the model, asking whether Codex research sessions entered training.
- A minority argued the capability result remains extraordinary despite poor handling of credit.