CONTROLAgents & automation
Jev: A Decision-Making AI Model
QFS17@riabcevv
Jev is an AI model that provides typed decisions, probabilities, and confidence scores for software to act upon, differing from traditional LLMs by not generating answers token by token.
Pattern↑ Jev AI decision models
a chatgpt co-creator just unveiled an ai model that doesn’t talk.
it decides.
former openai researcher diogo almeida has launched typesafe ai and its first “system one model” called jev.
unlike llms, jev doesn’t generate an answer token by token. you give it a structured question and it returns typed decisions, probabilities and confidence scores that software can act on immediately.
the company claims jev is:
• 20x to 200x faster than llms
• 40x to 400x cheaper
• priced at $0.042 per million input tokens
• free for output tokens
• capable of evaluating hundreds of options in parallel
in one demo, jev played doom by making around 10 decisions every second.
but the real use case isn’t gaming.
it’s the millions of tiny decisions hidden inside software: fraud detection, routing, ranking, moderation, pricing and agent workflows.
llms were built to talk to humans.
jev was built to talk directly to software.
typesafe has already raised $40 million and opened an early-access waitlist:
typesafe.ai