SCOREAgents & automation
Jev Decision Engine
Ujjal Bhattacharjee@ujjalcal
Jev is a lightweight decision engine designed to score, classify, route, and decide on options based on provided criteria, offering structured scores, consistent signals, and fast, cost-effective processing.
TakeInstead of asking an LLM to reason, explain, and generate prose, give Jev a focused decision with clear criteria and let it score the options.
Pattern✦ Decision layer primitives
Jev has been rocking the internet, so yesterday I gave it a try.
Setup → OpenRouter → API key → Muse
Action → Use Jev for decisions that need scoring.
Instead of asking an LLM to reason, explain, and generate prose, give Jev a focused decision with clear criteria and let it score the options.
Output →
Structured scores / probabilities
Consistent decision signals
Fast enough to run across hundreds of items
Extremely cheap
I tested it against my pipeline.
194 opportunities → 194 Jev calls → 6 questions each
Result:
4 → Queue
94 → Second look
75 → Human review
21 → Archive
Total cost: $0.0122
Median latency: ~0.5 sec
The interesting part wasn’t just speed or cost.
Adding substantially more context didn’t improve the triage result. A curated evidence digest was enough.
That changed how I think about Jev:
Frontier model → understand, research, generate, reason
Jev → score, classify, route, decide
Not a replacement for the agent.
A tiny decision engine inside the agent.
Related builds
