QUICK START PATH From zero to a bot that runs your morning.
The shortest beginner path we can recommend without pretending Jev is simpler than it is. Take the steps in order — each one assumes the previous.
- Five steps, in order
- Mental model → setup → workflow → shared bots → fixes
- Built from patterns used by every curated use case
01
What is Jev?
TypeSafe’s first System One model: state + typed questions in → decisions + probabilities out — no chat text.
- Built for software automation, not human chat (TypeSafe manifesto / launch post)
- Question types: Choice, Score, Boolean (Vercel Gateway docs)
- Early access / waitlist as of Sep 2026 — re-check typesafe.ai
02
Get access (TypeSafe or Vercel AI Gateway)
Two public paths — TypeSafe console key, or call typesafe-ai/jev via Vercel AI Gateway.
- Official API cited publicly: POST https://api.typesafe.ai/v1/systemone, model jev-latest — re-check live docs
- Gateway model id: typesafe-ai/jev; AI SDK experimental_evaluate (Vercel changelog 2026-09-16)
- Pricing claim on TypeSafe site: $0.042 / M input tokens, output free — operator-published; re-check live pricing
03
Run one evaluate call
Send shared state + a map of named questions; read typed answers and confidence in code.
- Prefer official SDKs / AI SDK evaluate API over hand-rolled parsers
- Keep questions narrow; put branching logic in your code
- Never paste API keys into content pages
04
Threshold confidence, then act or escalate
Every answer includes calibrated confidence — automate high confidence, escalate the rest.
- TypeSafe markets RLCD for calibrated decisions (launch post)
- Probabilities ≠ proof; calibrate on your labeled workflow
- Low confidence → human review or LLM fallback (pattern, not a TypeSafe SLA)
05
What Jev is not
Jev does not generate prose, code, or free-form tool strings — surrounding code owns loops and safety.
- Not an LLM chat model; not a screenshot-native VLM by default
- OpenJev / SemIf reproduce an interface pattern, not Jev weights (repo disclaimers)
- Official jaggedness / limitations docs — link when URL confirmed in TypeSafe Docs