Jev founder Diogo Almeida: RLHF is an epic wrong turn on the road to automation
dotey · x · 2026-09-20
A recap of Jev founder Diogo Almeida's talk:
- Jev was built around automation from the start. Almeida argues RLHF is an epic wrong turn: great at pleasing humans, terrible at full automation, since humans must stay in the loop.
- ChatGPT and Claude Code share the same RLHF-rooted paradigm—Claude Code isn't the next era; true automation is.
- RLVR isn't the answer either: RLHF optimizes human preference, RLVR pure correctness; Jev pursues a third path optimizing calibrated decision-making.
- Data matters more than compute, and picking the right tasks matters more than data. Pretrained models are already smart enough—preference optimization is what bent them off course.
Related event: Ex-OpenAI Founder Says RLHF Is a Detour, Human-AI Collaboration Era Ending(3 posts)→
More from AGI Musings
- Newsweek op-ed: US should fund hospitals and universities in federated AI learning — Saul_Loveman · 2026-11-12
- 10-line deterministic polynomial-time primality test relies on OpenAI's recent quasi-GRH result, making AKS obsolete — fortnow · 2026-11-11
- AI2027 Author Kokotajlo Shifts AGI Median Timeline to Around 2030 — JacquesThibs · 2026-11-11
- Ron Bodkin: automated research taste advances faster than expected, but the evidence is weak — ronbodkin · 2026-11-11
- Should devs still hand-write code? Author proposes 1,000-hour baseline before AI agents — bendee983 · 2026-11-11
- DEF CON panel with Bruce Schneier and Google security VP posted: scaling laws and open weights — JasonDClinton · 2026-11-11