Jev’s score is 0..n-1, not 0–1 — the gotcha that breaks thresholds

omelette joe — Mon, 21 Sep 2026 05:04:55 +0000

If you treat TypeSafe’s Jev score like a confidence float between 0 and 1, your gates will be wrong.

What Score actually returns

Score is the probability-weighted mean of rubric indices. If your criteria are ["Low","Normal","High","Urgent"], indices are 0..3. A live response we copied looked like:

score: 2.46 (not 0.62)
legend: {"0":"Low","1":"Normal","2":"High","3":"Urgent"}
probabilities keyed by index strings ("0", "1", …), never by your labels

So 2×0.5 + 3×0.48 ≈ 2.46 means “between High and Urgent.” Round for the band; the fraction tells you which way the mass leans.

Two siblings that also bite

Noul has no confidence field — not null, absent. Distance from 0.5 is what you have for a gate.
Ask several questions in one call: they share one state and run in parallel. Paying for the state three times is the expensive habit.

Try it without the official waitlist

I run an independent relay (not a TypeSafe first-party key): Google sign-in → jv_live_… → POST https://jev-agent.com/api/v1/classify with a ruleset name, or /api/v1/systemone with the official shapes so you can swap host+token later.

Free tier: 5 credits / month (1 credit ≈ 1000 input tokens). No uptime SLA.

Start: https://jev-agent.com/api-access · docs: https://jev-agent.com/docs

Not affiliated with TypeSafe AI.

Try TypeSafe’s Jev Free: Live Demos, Measured Cases, No Waitlist

omelette joe — Sun, 20 Sep 2026 13:21:14 +0000

TypeSafe’s Jev is a System One decision model: you send a state plus typed questions, and you get structured answers back — a Choice, a Score, or a Noul probability — not chat prose.

If you only need “which tool runs next” or “how bad is this ticket,” parsing JSON out of a chat model is slow and fragile. Jev is built for that gap.

What you can do today without the vendor waitlist

On jev-agent.com you can:

Get a free API key (email, no card) for eight tuned rulesets and call decisions without writing Jev question shapes yourself
Run live demos in the browser: chess and wikirace
Read use-case writeups with live-API measured cost and latency: triage, tool selection, moderation, routing, and more

Numbers from live calls (not a slide deck)

Rate card: $0.042 per 1M input tokens, output free (no output tokens).

Examples reported from live API calls:

One Choice over 3 options: $0.0000144 (342 input tokens)
Choice + Score + Noul in one call: $0.0000226 (539 tokens)

Latency on the hot path is typically 70–500ms end-to-end depending on route. Putting several questions in one request matters: they share one state and run in parallel, so you don’t pay for the state three times.

When it fits

Reach for Jev when the answer is a decision, not a sentence: routing, classification, tool pick, rubric grading, boolean guardrails. Skip it when you need prose, code, or a conversation — pair it with an LLM for that half.

Not affiliated

jev-agent.com is an independent guide with measured figures and demos. Not affiliated with TypeSafe AI. Where this site and the official docs disagree, trust docs.typesafe.ai.

Start here: https://jev-agent.com · free key: /api-access · cases: /use-cases · pricing: /pricing

DEV Community: omelette joe