Allie Laabs
TypeSafe AI · Jev, the System 1 model
Allie Laabs joined Alex Volkov on ThursdAI the week TypeSafe AI came out of stealth with Jev, a non-LLM System 1 decision model from RLHF co-creator Diogo Almeida. The founding DevRel walked through the choice, score, and null primitives that expose calibrated probabilities directly to code, why LLMs are the wrong tool for machine-to-machine decisions, and the pricing: $42 per billion input tokens, with RLCD training making output tokens too cheap to meter.
Allie on ThursdAI
Why LLMs are the wrong tool for machine-to-machine work
Machine-to-machine decisions don't need prose — they need calibrated probabilities code can act on.
$42 per billion tokens, and output is free
Jev is priced per billion input tokens — RLCD training makes output tokens too cheap to meter.
Jev is not an LLM, it is a System 1 model
Choice, score, and null primitives returning calibrated decisions in 70–500ms — a different animal from an LLM round-trip.
Suggested posts
ThursdAI — The weekly AI podcast, hosted by Alex Volkov. Every Thursday, live.
Subscribe Free →