Tao's blog blasts OpenAI's 400-theorem dump: 'This is normal bad work'
RexDouglass · x · 2026-10-15
A guest post by Sophie Huiberts on Terence Tao's blog argues OpenAI's mass theorem dumps violate scholarly norms: poor attribution, 372 results sliced across 719 PDFs, retraction-worthy errors, promised-but-missing code, and unreadable proofs—'easy rejects' that set back the literature. RexDouglass adds industry shops don't produce scientific results anyway, a pattern ML already went through.
More from Companies & People
- Supabase-backed NYC hackathon on Nov 4 offers prizes and credits for startup builders — santiviquez · 2026-10-15
- Ari Emanuel on AI: Agent Era Threatens Google and Amazon Ads, Social Networks Undervalued — Stratechery · 2026-10-15
- Which AI and cloud training platforms are worth it? A pragmatic comparison — DavidLinthicum · 2026-10-15
- Cerebras CEO: OpenAI went with Nvidia because Cerebras lacked capacity for GPT-6 — firstadopter · 2026-10-15
- Fireworks AI CEO: 0.1% per-turn numerical drift compounds to ~10% over 100 agent turns — AccBalanced · 2026-10-15
- Cohere to Acquire AI Security Startup AISLE, Which Found 500+ Real CVEs in Its First Year — stanislavfort · 2026-10-15