Package I · Add-on T7

Learning paths for TypeSafe Jev (System One)

Curated external starters and indexes so you can learn Jev without waiting on every chapter card. Named links only — we do not mirror awesome lists in full, invent livestream official cases, or promote clone tutorials as first-class paths.

How this relates to Use cases. Use cases holds structured case cards (T6). This page is the Tutorials learning-path index (T7): start docs, demos, and directory hubs. High-priority monitoring short-table cases are filed into the chapter pages after T7.

Quick answers

What is Jev?

TypeSafe Jev is a System One evaluation model: it returns typed Choice, Score, and Boolean answers with probabilities — not free-form prose.

Choice / Score / Noul

Choice picks one named option; Score places state on an ordered rubric; Noul (boolean) estimates P(true). Confidence comes from the option distribution.

When to use a human

Low confidence → human queue with no default action. Money, refunds, and irreversible side effects stay human-owned — suggestions only.

1 · Start here

Official map, docs, hello world, then optional community explainers and demos.

Use-case map

What you learn: See where System One / Jev decisions fit before you pick a pattern.

Link: docs.typesafe.ai/concepts/use-case-map

Notes: Community-sourced · not affiliated with TypeSafe

hello-jev

What you learn: Minimal first repo to call Jev and see a typed decision return.

Link: github.com/nishimotz/hello-jev

Notes: Community-sourced · not affiliated with TypeSafe

Jev is GPU for LLMs (Zenn) Optional

What you learn: Community essay framing when a decision model beats text generation.

Link: zenn.dev/mizchi/articles/jev-is-gpu-for-llms

Notes: Community-sourced · not affiliated with TypeSafe

typesafe-demo (Val Town) Optional

What you learn: Try a live Val Town demo without standing up your own stack.

Link: typesafe-demo.val.run

Notes: Community-sourced · not affiliated with TypeSafe

2 · Killer demos

High-signal public repos that show Jev under load or in agent harnesses.

jev-ultrafast

What you learn: Browser-use launch demo focused on speed and low-latency decisions.

Link: github.com/browser-use/jev-ultrafast

Notes: Community-sourced · not affiliated with TypeSafe

jev-harness

What you learn: Harness scaffolding for evaluating and wiring Jev decisions.

Link: github.com/AntonioCoppe/jev-harness

Notes: Community-sourced · not affiliated with TypeSafe

pi-heed

What you learn: Agent / tooling experiment that gates actions with Jev-style checks.

Link: github.com/Nyarlathoteppppp/pi-heed

Notes: Community-sourced · not affiliated with TypeSafe

3 · Directories

Index links only — open the list on GitHub or the directory site; we do not expand each repo as a fake case here.

awesome-jev (awesomejev.cc)

What you learn: Community index of Jev links and projects — browse the list, do not copy it here.

Link: awesomejev.cc · github.com/daftAI2026/awesome-jev

Notes: Community-sourced · not affiliated with TypeSafe

awesome-jev (AnotiaWang)

What you learn: Additional community awesome list — use as an index, not a mirror.

Link: github.com/AnotiaWang/awesome-jev

Notes: Community-sourced · not affiliated with TypeSafe

awesome-jev-usecases

What you learn: Use-case-oriented awesome index for discovery (not chapter cards).

Link: github.com/aliaihub/awesome-jev-usecases

Notes: Community-sourced · not affiliated with TypeSafe

Out of scope here. Open-jev / Laya / Hemant-style clones are not first-class tutorials on this page. Do not treat directory mirrors as official TypeSafe cases. Paid pack templates stay under Templates / Livestream.

FAQ

How should I start learning TypeSafe Jev?

Use the Start here paths: official map and docs, then a minimal hello repo. Killer demos and directories are optional indexes — not full mirrors.

Are directory lists official TypeSafe cases?

No. Awesome-style directories are community indexes. This site does not expand every repo as a fake case or treat clones as first-class tutorials.

Where are paid pack templates?

Paid starters stay under Templates and Livestream. Learning paths link out; they are not a substitute for the Gumroad pack.