Intro: Jev launched Sep 15 and drew a 140k waitlist in days. On Sep 20 TypeSafe removed the waitlist entirely — open registration with $5 in free credit (~120M tokens). This post covers what Jev actually is, what it costs, and the fastest way to start.
Section 1 — What Jev is: Not a text generator. Returns typed decisions — Choice (up to 255 options), Score, Noul (yes/no probability) — each with calibrated confidence. 64K context, read-only. "Zero hallucination" = schema guarantee, not correctness guarantee — understand this distinction before building.
Section 2 — The cost model: $0.042/MTok input, output free, ~$0.000081 per decision. $5 credit ≈ 120M tokens ≈ enough to prototype seriously. Worked example: 1,018 papers classified for $0.08. Comparison table vs LLM-based classification. Caveat: vendor says it can't prove the price isn't subsidized — treat it as today's price.
Section 3 — Access paths (with the correction): Direct console (open now, $5 credit) / Vercel AI Gateway (typesafe-ai/jev) / Cloudflare Workers AI (typesafe/jev, 32K context). Note: launch-week posts claimed OpenRouter listed Jev — it doesn't. Verify against a provider's own model list.
Section 4 — What to build first: three starter workflows — ticket triage, lead scoring, RAG filtering — each with the decision boundary explained (classification not prediction, atomic questions not compound ones).
Section 5 — When NOT to use it: anything generative, prediction, compound reasoning. Hybrid architecture: LLM generates → Jev decides.
Ending: "I maintain an independent tracker at JEV snapshots, daily news, a cost calculator, and a ranked directory of the community projects built on Jev."