Ex-OpenAI RLHF builder launches Jev: a chatless decision model with 150ms responses and $42/B input tokens
SucceededMind · x · 2026-09-16
Diogo Almeida, who helped build RLHF at OpenAI, has launched TypeSafeAI and its first model, Jev. The thesis: generating prose is a major bottleneck for agentic workflows, so structured decisions shouldn't require words.
Jev is a pure API "System One" model — ask a question, get back typed answers with exact confidence scores, no generative text at all. That yields:
- 150ms response times, claimed 200x faster than standard models
- $42 per billion input tokens
- Output tokens free forever
The author frames it as a complement to chat models: route logic and structured decisions to Jev so your if-statements get a decision they can actually trust, while chat models handle the parts that genuinely need generated text.
Related event: TypeSafe exits stealth with decision model Jev and RLCD training method(86 posts)→
More from coding & agent
- Cohere Labs agent team-construction method cuts multi-agent costs by 70% — Cohere_Labs · 2026-11-02
- Long-read: AI's real value comes after the efficiency era—infra glut, agent tooling boom — yangyi · 2026-11-02
- offprompt: MIT-licensed MCP server hands agents API keys without ever exposing them — Vast_Ad_7338 · 2026-11-02
- Reddit idea: expose every OOP public method as AI agent tools to agentify legacy apps — out_of_nowhere__ · 2026-11-02
- DeepSeek design doc: a person can only follow 3-5 agent sessions at once, attention is the real bottleneck — heyshrutimishra · 2026-11-02
- Software fundamentals are key to getting good results from AI coding agents — bendee983 · 2026-11-02