Jev (TypeSafe AI) — unofficial developer notes

Practical notes and examples for Jev, TypeSafe AI's System One model — the AI that returns typed, calibrated decisions for software instead of text.

🧪 Try Jev free in your browser (no waitlist): jevtypesafeai.com Live playground that runs real /v1/systemone calls.

This is a community reference, not affiliated with TypeSafe AI. For official access and docs see typesafe.ai and docs.typesafe.ai.

---

What is Jev?

Large language models are built to talk to people. Jev is built to be called by machines. You give it some context (your program state) and the exact questions you need answered, and it replies with structured values instead of prose. Because the shape of the answer is fixed by your request, Jev can't hallucinate a format or emit an invalid type.

• Speed: 70–500 ms (≈ 40–200× faster than a frontier LLM on comparable tasks)
• Cost: $0.042 / million input tokens, output free (≈ $0.0004 per decision)
• Safety: no hallucinated formats, no type errors, by construction
• Made by: TypeSafe AI (founded by ex-OpenAI researcher Diogo Almeida), launched Sept 2026

More background: What is Jev?

The three question types

Every question you send Jev is one of three primitives:

type	returns	use for
choice	one of up to 255 labelled options + probabilities	routing, classification
score	a position on a 2–10 level ordered scale	risk, urgency, quality, fit
noul	a calibrated yes/no as a probability (0–1)	gates, filters, guardrails

Quick example

curl -X POST https://api.typesafe.ai/v1/systemone \
  -H "Authorization: Bearer $TYPESAFE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jev-latest",
    "state": "Customer: I was charged twice and I am furious.",
    "questions": {
      "topic":  { "type": "choice", "instructions": "What is the issue about?",
                  "criteria": { "billing": "money problems", "bug": "broken product" } },
      "urgent": { "type": "noul", "instructions": "Escalate to a human now?" }
    }
  }'

Response:

{
  "model": "jev-1.13.0",
  "answers": {
    "topic":  { "type": "choice", "choice": "billing", "confidence": 1.0,
                "probabilities": { "billing": 1.0, "bug": 0.0 } },
    "urgent": { "type": "noul", "noul": 0.82 }
  },
  "usage": { "input_tokens": 382, "output_tokens": 55 }
}

Because the types are fixed, you branch on results with plain code — if (answers.urgent.noul > 0.7) — no parsing, no regex, no malformed-response risk.

Common use cases

• Route & classify — send a ticket, email, event or document to the right place
• Score & prioritize — turn fuzzy judgments into a threshold you can act on
• Guardrail & gate — auto-approve the easy cases, escalate the uncertain ones
• Extract — turn free text into typed, structured fields

Links

• ▶️ Live playground (free, no waitlist): https://jevtypesafeai.com
• 📖 What is Jev: https://jevtypesafeai.com/what-is-jev
• 🔌 How to use the API: https://jevtypesafeai.com/how-to-use
• 💵 Pricing: https://jevtypesafeai.com/pricing
• 🏢 Official site: https://typesafe.ai

License

MIT — these notes are free to reuse. "Jev", "System One" and "TypeSafe AI" belong to their respective owners.