TypeSafe Debuts Jev, an AI System That Answers Preset Questions With Confidence Scores

TypeSafe, the company founded by former OpenAI researcher Diogo Almeida, emerged from stealth on Sept. 16, 2026, with Jev, a new AI system designed to answer preset questions inside software while returning confidence scores. The company says Jev cannot hallucinate because it only chooses from options set in advance, describing the system as a "frontier-intelligence function call."

According to TypeSafe, Jev costs $42 per billion input tokens with free output, a price the company estimates is about 238x below Claude Fable 5.1's pricing. Responses are delivered in 70 to 500 milliseconds, making it 40 to 200 times faster than current LLMs. Use cases include quick judgment calls inside apps, such as sorting requests, scoring records, or screening another AI's outputs for jailbreaks.

Almeida said the system is not meant to be stacked up against LLMs, calling it "more like a database than a coworker." The report also cited the Jevons paradox, noting that if Jev is fast, cheap, and reliable, it could become a standard component inside software.