Jev: TypeSafe's 70-500ms System One AI primitive costs $0.042 per million input tokens
daniel_mac8 · x · 2026-09-23
Dan McAteer breaks down Jev, a new ultra-fast, ultra-cheap "System One" AI primitive from Diogo Almeida (ex-OpenAI, co-author on the InstructGPT paper) and his team at TypeSafe.
What it is
- Response times of 70–500ms; $0.042 per million input tokens, output free.
- Unlike LLMs, you don't chat with it: you hand it a decision and it returns a probability distribution over options, predicting which is most likely correct.
How to build with it
- Inputs are the "state" (an email, document, or app state to evaluate) and a question about that state, defined via three types such as Choice over provided options.
- Jev shines inside programs making bounded judgments: you supply context and outcomes while code and LLMs handle surrounding work — the engineer's challenge is deciding which judgments to delegate.
- The post draws on TypeSafe docs and community experiments to explain how it works, how builders combine it with LLMs, and a test for whether your app can use it.
Related event: TypeSafe Launches Jev, an Ultra-Fast "System One" AI Primitive(2 posts)→
More from coding & agent
- Glassnode now lets AI agents buy on-chain data via Coinbase x402 payments — kleffew94 · 2026-09-23
- CodeRabbit launches Change Stack to help reviewers actually verify AI-generated PRs — aryanXmahajan · 2026-09-23
- AI memory is harder than it looks: chatbots rely on RAG-style indexing, not real recall — BLUECOW009 · 2026-09-23
- AI agents told to build a society in Unreal Engine spent 20 hours beating each other up — SimplyAnnisa · 2026-09-23
- Stop training agents on chat alone: product behavior data carries 10-100x more signal — techNmak · 2026-09-23
- Yacine asks: is anyone building an RL environment for code smells? — yacineMTB · 2026-09-23