Jev

TypeSafe AI, a San Francisco startup founded by former OpenAI researcher Diogo Almeida with a $40 million seed round, launched Jev on September 15, 2026. The name comes from TypeSafe's framing of a 'System One model,' after Daniel Kahneman's concept of fast, intuitive thinking: instead of deliberating token by token, Jev picks immediately from a set of options you define in advance.

You give Jev a state (a message, a document, the current context of an agent) and a schema of typed questions: yes/no probabilities, single-choice selections, or numeric scores. It returns calibrated probability outputs for each question in a single forward pass. No tokens are generated, so there is nothing to parse and no chance of a malformed answer. TypeSafe reported speeds 40 to 200 times faster than frontier LLMs on the same classification tasks, at a fraction of the cost.

Jev became the de facto reference point for the decision model category. Cloudflare's Clef, Liquid AI's d1, Microsoft's Decision-1, and Perplexity's pplx-decider all shipped within three weeks of Jev's launch, and several are compatible with the Jev API format, meaning teams can swap models without rewriting integrations. For builders building agent routing, guardrails, content moderation, or ticket triage, Jev and its successors represent an alternative to calling a full generative model for decisions that don't require prose.