Start here

You found a directory of things people build with Jev, the decision model from TypeSafe AI. This page is the map. Sixty seconds of orientation, then a path picked by what you came for.

The 60-second version

Jev answers small, closed questions (labels, yes/no, scores, choices) at very high speed for fractions of a cent. It doesn't chat and it doesn't write; that's the design, not a gap. Because a huge share of software's work is secretly small judgments (route this ticket, flag this post, pick the next click), a model that makes judgments nearly free turned out to unlock a lot, fast: 400+ cataloged builds within weeks of the September 2026 launch.

This site catalogs those builds with their receipts: every entry carries the cost and speed numbers its author reported. Reported is the operative word; we keep the sources linked and the hedging attached, and there are no official benchmarks yet.

How this directory works

Builds live in eight categories, from triage and routing to games and real-time. Each entry links the author's own write-up and numbers. Listings are free and human-curated; sponsor slots exist and are labeled; and if you ship something, submit it with your numbers, because receipts are the house currency. We are not affiliated with TypeSafe AI.

Pick your path

I want to build something. Read What is Jev Engineering first; question design is the actual skill and it's learnable in an afternoon. Then the getting-started tutorial for your first calibrated pipeline, and the use-cases tour to steal a proven starting point. Official mechanics (access, API, limits) live at TypeSafe's docs, which always outrank anything written here.

I'm evaluating whether this is real. The review is our honest verdict, the limitations page is the unhyped list, what builds actually cost is the receipts ledger, and the alternatives landscape tells you when the answer is "not this."

I want to understand the shift. What is Jev for the object, why decisions became free for the thesis, and the glossary when the vocabulary gets ahead of you.

I run a team and smell a use case. Start where the volume hurts: support triage, email, lead scoring, or evals and guardrails if you already ship AI. The production ops guide is what to read before anything touches customers.

The reading order, if you just want one

What is Jev → What is Jev Engineering → getting started → limitations → build something → submit it. That loop, run once, is worth more than any amount of further reading, which is why the sidebar you came from is five links long.

Frequently asked questions

Where are the official Jev docs?

At docs.typesafe.ai, maintained by TypeSafe AI: access, API mechanics, pricing, and limits all live there. This site deliberately never reproduces API details, because stale copies are worse than links.

Is this site affiliated with TypeSafe AI?

No. Independent directory, human-curated, receipts required; sponsor slots are labeled as such and listings are free.

I have zero AI background. Can I still build with this?

If you can write clear criteria, yes: the barrier is question design, not infrastructure. The tutorial path above assumes no ML background at all.

What should I absolutely read before shipping to real users?

The limitations and the production guide: confidence gates, escalation, versioned questions, and the rule that irreversible actions never ride a lone verdict.