jevTYPESAFE AI · PRIMARY SOURCE
Introduction & quick start ↗
Start here to understand state, questions and answers before sending a request.
docs.typesafe.ai
FIELDNOTES / PRIMARY SOURCES
Official Jev reading list
A guided path through TypeSafe’s documentation, demonstrations and published evaluations.
Start with the basics, choose a workflow, then check limitations and evaluation examples.
16 SELECTED RESOURCES · REVIEWED 28 SEP 2026
01 / OFFICIAL RESOURCES
get started
TYPESAFE AI · PRIMARY SOURCE
Choice, Score & Noul ↗
Choose a category, an ordered rating or a yes/no probability to match the result your code needs.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
How to build with System One ↗
Work backward from product behavior, keep rules and execution in code, and use Jev for the semantic judgment.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
TypeSafe agent skill ↗
Install the official skill so Codex, Claude Code or another coding agent can find current Jev patterns and API guidance.
docs.typesafe.ai
02 / OFFICIAL RESOURCES
build a workflow
TYPESAFE AI · PRIMARY SOURCE
Confidence-gated routing ↗
Use a decision and its confidence to choose when to act and when to ask for review.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Speculative fan-out ↗
Ask candidate questions together, then let code use the relevant answers. Useful when calls do not truly depend on one another.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Smart home assistant demo ↗
An example of device-control questions, result selection and a fallback to a text-generating model.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Official Jev use-case map ↗
Scan decision patterns by industry, then translate a promising idea into explicit state, questions, code policy and review paths.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
LLM guardrails cookbook ↗
Screen LLM inputs and outputs with separate hazard judgments, then let explicit code thresholds pass, review, block or route.
docs.typesafe.ai
03 / OFFICIAL RESOURCES
check the limits
TYPESAFE AI · PRIMARY SOURCE
Confidence and probability ↗
Distinguish confidence from answer probability. Noul has no separate confidence field.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Jev 1.13 model jaggedness ↗
Read the documented weak spots and turn relevant ones into test cases for your application.
docs.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Models, pricing & input limits ↗
Check model identifiers, pricing and input limits before configuring a deployment.
docs.typesafe.ai
04 / OFFICIAL RESOURCES
assess the evidence
TYPESAFE AI · PRIMARY SOURCE
Introducing System One Models & Jev ↗
Find the side-by-side, Doom and Wikiracing demonstrations. Read the conditions beside each demo.
typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Lies, Damned Lies, and Benchmarks ↗
Read the publisher’s case against optimizing for public leaderboards and its stated preference for dated, caveated evaluation snapshots.
typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Official workflow evaluations ↗
Explore four publisher-evaluated workflows. Their reference answers are based on model consensus.
evals.typesafe.ai
TYPESAFE AI · PRIMARY SOURCE
Customer service workflow ↗
Inspect a customer-service workflow, including inputs, code policies and prediction disagreements.
evals.typesafe.ai
LIVE PRODUCT / PUBLIC BETA
Move from the official request contract to your own fixtures.
The Workbench gives the reading path a practical next step: encode one Choice, Score, or Noul policy, keep ambiguous cases in review, and test it without handing the product control of downstream actions.
20 free live Decision Runs for eligible new accounts∞ mock runs while you shape the policy50 sanitized fixtures in one versioned Pack
Google sign-in · provider key stays server-side · no automatic tool actions · no payment required during beta