TypeSafe AI launched publicly on Sept. 15, 2026, after working in secret, with $40 million in seed funding led by DCVC and early access to Jev, its first System One model. That is TypeSafe's name for AI built to make fast, structured decisions that other software can use directly, instead of chatting with a person in a browser. TypeSafe is pitching a model that programs can call for answers in a fixed, ready-to-use form, so developers would not have to pick apart chat replies. Early access has a waitlist at typesafe.ai.
Jev is not a chat model. Developers give it the current state of their program and a question in one of TypeSafe's set formats, such as Choice (picking one of several options), Score (rating something on a scale) or Noul (true or false). It returns answers in fixed types, with probabilities and confidence scores, that software can use directly. TypeSafe trained it with a method it calls Reinforcement Learning for Calibrated Decisions, or RLCD, a form of learning by trial and error with a reward for getting it right, and says Jev cannot write free-form text the way ChatGPT-style models do.
TypeSafe was founded in 2024 and is based in San Francisco. Its co-founders are Diogo Almeida, a former OpenAI researcher and co-inventor of reinforcement learning from human feedback (RLHF) methods tied to ChatGPT, Erik Gafni and Sasha Sheng. In company materials, Almeida said most intelligence should eventually live inside software rather than only serve people directly.
TypeSafe's launch announcement said Jev responds in less than 100 milliseconds and is up to 100 times faster and less expensive than other frontier models, the most capable AI models available today. The company blog lists a price of $0.042 per million input tokens, the small chunks of text, often parts of words, that AI models read and write. It says output is free and gives response times of 70 to 500 milliseconds from start to finish. Its homepage figures of about 194 times faster and about 445 times cheaper come from TypeSafe's own tests on its workflows, and the company says it expects them to be at the higher end of real-world gains.
A news report described a TypeSafe demo using the video game Doom. The same report said TypeSafe's pitch that Jev is "hallucination-free," meaning it never states something false as if it were true, is not a fair comparison with chat models, because Jev does not answer in ordinary language. It added that Jev's structured answers can still be wrong. A startup-funding database puts the $40 million round among the largest seed rounds in AI in its records. It is not yet known whether outside groups will see the same speed, cost and reliability on their own work, or how far Jev's typed decisions will reach beyond early-access demos.