Last week, Deutschlandfunk Kultur’s pop culture show Kompressor interviewed me about Jev, the new “System 1” model from the San Francisco start-up Typesafe AI.
Want to hear how a lightning-fast AI model could moderate thousands of Twitch chat messages per minute, play Doom, or flag emergencies at a festival in a split second? In the radio interview, I explain what sets Jev apart from LLMs like ChatGPT or Claude: rather than writing essays, it ticks boxes in a multiple-choice test, making fast, reflexive decisions instead of slow, deliberate ones. We also discuss whether Jev is the next big hype, and why I think its future lies alongside LLMs rather than in competition with them.
You can listen to the full interview (in German), “Blitzschnelle KI – Was bringt JEV der Popkultur?”, on the Deutschlandfunk Kultur website.
We have also put Jev to the test ourselves: in our paper “Evaluating and Benchmarking the System One Model Jev”, we benchmark it zero-shot on 37 datasets covering classification, routing, moderation, and rubric scoring, as well as on sequential decision-making tasks.