Member-only story
Jev Isn’t Here to Replace Your LLM. It’s Here to Make Decisions Cheap.
What TypeSafe’s new System One model actually does, what independent testing shows so far, and where I’d put it inside an AI agent.
A while ago I tried counting how many model calls one of our agent runs made, and what each call was actually for. The answer was a little embarrassing. Most of them weren’t writing anything. They were deciding. Which tool should run next. Is this request in scope. Is the query we just generated safe to execute. Does the answer we wrote line up with the numbers we pulled.
Every one of those is a multiple-choice question, and we were paying a frontier language model to answer each one by writing out a word, which our code then parsed and hoped was formatted correctly.
That’s the problem Jev is aimed at, and it’s why I think it deserves a calmer look than it’s getting on social media right now.
What Jev is, in plain terms
Jev comes from TypeSafe AI, a startup founded by Diogo Almeida, who worked on the research behind ChatGPT while at OpenAI. It was released in early access on September 15, 2026, and TypeSafe calls it the first “System One model.” The name borrows from Daniel Kahneman’s split between fast, intuitive thinking and slow, deliberate reasoning.
The key thing to understand is that Jev doesn’t generate text at all. You send…
