Tags: inference
All the articles with the tag "inference".
The Decision Plane: Why Jev Changes Enterprise Agent Architecture
Published: at 02:00 PMMost enterprise agent traffic is decisions, not prose — yet every branch is produced by an autoregressive model. Here is what changes when the decision layer comes out of the LLM and back into code, and the cascade architecture the accuracy numbers actually justify.
Jev: The Frontier Model That Refuses to Write
Published: at 09:00 AMJev is a frontier model that never generates text. It scores the options you supply in a single parallel pass — 8.5 seconds of LLM output replaced by 0.1 seconds of arithmetic. Here's the trade, and why it matters for the cost architecture of every agent you run.