Observed arrival · 2026-09-21
Jev Engineering’s decision layer for AI agents
A technical course and browser demo for routing, scoring, approving, and escalating agent decisions without spending a frontier-model call on every fork.
- For
- Engineers building agentic software gates
- Worth noticing
- The browser demo exposes verdict, blast radius, and human-escalation probabilities without making an API call.
Field notes
The homepage models a merge gate with three signals: a categorical verdict, a blast-radius score, and a probability that human review is needed. Its example escalates when the radius reaches 1.5 or uncertainty reaches 0.70, while a confident ship decision requires 0.88 confidence. The page also shows a 32,000-token context claim, 70–500 ms stated latency, and a cost comparison based on 1,000 input tokens per decision.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue