Jev, an AI for making quick decisions, has been a viral hit. But OpenAI is hot on its heels

TypeSafe’s Jev has barely had time to settle in, and OpenAI is already after it. On Oct. 6, just three weeks after Jev launched, OpenAI introduced Decisions API, a direct shot at the same kind of routine AI judgment work that TypeSafe is chasing.

This is not about chatbots writing essays. It is about whether software should be able to make thousands of small yes-or-no calls every day without burning time and money on a full language model response. Think customer-service routing, spam filtering, screening changes, or deciding whether a message needs a human. The economics matter here. At scale, tiny gains in speed, cost and error rates decide whether the whole thing is worth automating.

Diogo Almeida says he started thinking about Jev after working on ChatGPT technology at OpenAI and noticing a gap: models could answer questions, but they were not built as clean tools for routine work. He says he wanted to fix AI’s “massive over-promise under-deliver issue” and avoid an “AI winter.” The company took two years to make Jev reliable, far longer than the week Almeida first imagined.

What TypeSafe built is a faster, cheaper classifier that still takes natural-language instructions. Developers tell Jev what to judge, and it returns choices, scores or probabilities in a single shot rather than generating a long answer step by step. Users are already poking at it in the real world. One developer liked its speed in a spelling-and-grammar prototype. Another found that more specific questions reduced missed email alerts. Paolo Rosson at Dext said the appeal is simple: with LLMs, “you pay for a full written answer when all you need is a label.”

OpenAI is pushing the same idea from a very different place. Its Decisions API uses GPT-6 Luna without extra training, can handle images, and OpenAI says it can be up to ten times faster than using GPT-6 Luna through the general Responses API. It also has the usual OpenAI advantages: deep pockets, a huge user base, and a shortcut for companies already using its other tools. Still, TypeSafe has the edge of time spent on reliability, plus a $40 million seed round from DCVC to keep going. The race now is simple enough. TypeSafe spotted the need first. OpenAI showed up with more muscle.