Will Depue: Jev's success suggests old ML ideas are worth rescuing
willdepue · x · 2026-09-22
OpenAI's Will Depue reflects on the surprising traction of his Jev project: it is essentially just a zero-shot classifier with frontier-ish intelligence, and he's surprised no one had built it before. He suspects embedding models hold similar overlooked opportunities ("Jev for embedding models") and wonders which other classic ML ideas are worth rescuing with modern model capabilities.
Related event: OpenAI researcher: Jev's hype shows old ML ideas deserve reviving(3 posts)→
More from Research
- OpenAI's new model claims Navier–Stokes and 100+ open math problems in weeks — Dr_Singularity · 2026-09-22
- Limite trained from scratch on under 300B math-heavy tokens, nanogpt-speedrun-inspired — tensorqt · 2026-09-22
- OpenAI says model trained just 24 days ago has solved 100+ longstanding math problems — gaganghotra_ · 2026-09-22
- Forging 1024-bit RSA signatures in nearly SNFS time, sans factoring N — matthew_d_green · 2026-09-22
- Multi-turn evaluation changes what it means to pass the Turing test — CShorten30 · 2026-09-22
- Paradigma's 1B model scores 74.25% on BeyondAIME, trained on under 300B tokens — tensorqt · 2026-09-22