Shaduf.
Independent, source-led catalogue · access checked 20 Sep 2026

Jev, in the place where software makes small decisions

A practical starting point for builders comparing TypeSafe AI's System One model with rules, classifiers, general LLMs, and local experiments—now with a current access and pricing snapshot.

Jev is not a chat model. It takes state plus typed questions and returns a choice, a score, or a yes/no probability that your program can branch on.

Short answer: Jev looks promising for bounded, repeated judgments—routing, scoring, ranking, and guardrails. Keep thresholds, permissions, side effects, explanations, and human fallback in ordinary code.
Access refresh: the current TypeSafe model page names jev-1.13.0, with jev-latest and jev-preview currently pointing to it. Direct pricing is $0.042/M input tokens with free output, but login, credits, and the public waitlist do not prove that every new account can call the API. Compare every route.

The decision loop

1 · stateMessy contextA ticket, event, document, or program state.
2 · JevTyped judgmentChoice, Score, or Noul with probabilities.
3 · codePolicy and actionThreshold, route, review, or fall back.
New to Jev? Start with What is Jev? ELI5 for a plain-English picture of the model before comparing fit, alternatives, evidence, and access.

Four pages for a build decision

Evidence posture: TypeSafe's speed, cost, calibration, and access claims are marked as vendor-reported. A demo, adapter, directory entry, or GitHub repository is not automatically a production product.

This catalogue is independent of TypeSafe AI. The access refresh was checked on 20 September 2026 (run run:6ee2dcc6-09c0-4f2a-8e92-6d6835767621); pricing, model versions, gateways, and availability can change.

Search published pools, pages, reports, and evidence.