Comparing Jev and Decisions API probability outputs
A user's test shows that Jev returns mid-range probabilities (0.1–0.9) in about 50% of binary decisions, while Decisions API only does so in 22%, mostly giving decisive values like 0.02 or 0.98.
THE JEV GUIDE · CURATED EDITION
Community builds.
Launch news, explanations, demos, and community projects around TypeSafe Jev, organized by topic. Every entry links back to its X source.
Curated selection · Not exhaustive · Updated manually
6757curated posts
3414original videos
Sponsors
These products pay for the directory. Want yours up here?
Grok 4.7 + Jev: Smarter and Cheaper
A user shares combining Jev with Grok 4.7: Grok handles reasoning, Jev handles decisions. Jev does not generate text; it only answers via yes/no, choice, and scoring, so it can only decide from allowed answers.
What do you use Jev for?
Soly asks people to share how they use Jev.
Using Jev to monitor a camera and play deterrent sounds
A user shares that their Model 3 is too old for a feature, so they use TypeSafe AI's Jev to analyze a UniFi camera stream and play scary ambient sounds, flash lights, and honk when people get close to the car.
Jev: A Decision Costs Less Than a Cent, But a Wrong One Costs the Whole Run
Community discussion notes Jev's per-decision cost is extremely low, but emphasizes measuring cost per finished task, not per call.
One card in the feed, every sixteen, on every page of it. $50 a week, cancel any week.
Voice-Activated Checklist Prototype with Jev
A demo of using AnyDecisionModel and TypeSafe AI's Jev to check off checklist items via voice input.
Pairing Claude with Jev: trainable questioning gets close to the answer
Mike Taylor notes that teaching Claude to ask better questions and break problems down is a trainable skill; when paired with Jev, it almost always lands within 2x the answer.
Jev Matches Claude on Long-Chain Estimation
Mike Taylor shares that on long-chain questions without published answers, Claude works it out in thinking traces, and Jev can match its estimations 1:1.
A workflow combining Claude and Jev for probability estimates
Mike Taylor shares a practice where Claude handles question breakdown and Jev estimates probabilities, enabling mechanical napkin-math solutions.
Using Jev to add 'vibe' to Agent Tower
The author shares how they gradually thought about integrating Jev into Agent Tower, from basing room decorations on thread vibes to using Jev to shorten titles.
Warning on positional bias in Jev-as-a-judge
A community user warns that Jev-as-a-judge is strongly biased to position and often picks what comes first, so caution is advised.
Using Jev as a task router
The author shares using Jev as a small selection model that reads a task and returns executor, model, effort, and whether approval is needed. Calibrated on 40 real tasks, all successful.
