In mid-September 2026 a company called TypeSafe announced Jev, which it calls a "System One" model. It does not write text. It takes some input and a list of allowed answers and returns a typed decision with a confidence figure (TypeSafe's announcement). The company says the name was inspired by Daniel Kahneman's Thinking, Fast and Slow, and DataCamp's write-up repeats that.
Within about three weeks, others shipped models of the same kind: OpenAI's Decisions API in public beta, and Amazon's Strands Labs with Strands Decider 2B. TechCrunch covered the Amazon release on 1 October 2026.
What the Short had no room for: the mapping is not settled. A 2025 survey, From System 1 to System 2, likens ordinary language models to System 1 and reasoning models to System 2. DataCamp describes today's chatbots as System 2. Both are analogies, not facts about machines. Kahneman himself called the two systems "fictitious characters", not parts of the brain, and the book's own page is the place to start if you want his version: Macmillan's page for the book.
The durable idea is the confidence number. It lets software act on its own when the model is sure and hand the case to a person when it is not. The speed and accuracy claims in the launch posts are the companies' own, so the links above are there for you to read them yourself.