ROUTEAgents & automation
TypeSafe AI Jev Decision Engine
Nox|@noxflux
Jev is an AI decision engine that evaluates typed schemas directly in a single forward pass, bypassing the text generation of LLMs for faster and more reliable routing choices.
SignalInstead of generating text token by token, Jev takes application state and directly evaluates typed schemas such as booleans, multiple-choice options, or scores, in a single forward pass.
Pattern↑ Jev AI decision engine
TYPESAFE AI LAUNCHES VIRAL ‘JEV’ DECISION ENGINE
In an AI landscape dominated by chatty LLMs, San Francisco startup TypeSafe AI is trending across tech circles with Jev, a model that refuses to write a single word of text.
Founded by former OpenAI researcher Diogo Almeida with $40 million in backing led by DCVC, TypeSafe bills Jev as a "System One" decision engine.
Instead of generating text token by token, Jev takes application state and directly evaluates typed schemas such as booleans, multiple-choice options, or scores, in a single forward pass.
Essentially acting as an AI-powered if statement, it bypasses the hallucinations, JSON formatting errors, and latency common when forcing conversational models to make simple routing choices.
The model went viral on X largely due to its performance:
decisions take just 70 to 500 milliseconds, priced at $0.042 per million input tokens with free outputs.
Developers quickly flooded social feeds with demos of real-time web agents and fast gaming loops, prompting immediate integration into platforms like the Vercel AI Gateway and spurring open-source replication efforts across Hugging Face.