You Can Learn Jev VS LLM In 22 Min | System 1 vs System 2, AGI, Eval
Sean Chen explains TypeSafe AI's Jev decision model, contrasting its System 1 fast JSON architecture with conversational LLMs. He walks through Noul, choice, and score questions in the playground, then evaluates accuracy, latency, and cost across custom agent benchmarks.
Original by Sean‘s AI StoriesEvaluationIntermediate22 min 29 secPublished
Jev operates as a System 1 decision engine taking JSON state and questions, returning probability distributions without conversational text.
Jev supports three distinct decision types: Noul (boolean binary classification), choice (categorical selection from mutually exclusive options), and score (progressive ordinal expected value calculation).
In the creator's dashboard benchmark across 15 questions per category, Jev showed lower latency and cost than Opus or Haiku, but exhibited lower accuracy on complex scoring tasks.
Worth knowing
The benchmark was conducted on a small sample of 15 custom, human-labeled questions across five narrow categories rather than an extensive, controlled industry benchmark.