Skip to content

Jev By Example

Ten runnable JavaScript experiments on agent decisions, each pairing a small Jev judgment with visible application policy code, plus authored fixtures so the whole set runs offline.

Stars
2
Language
JavaScript
License
MIT
Published by
Community
Author
ReallyArtificial
Last verified
Oct 2026

About this project

Should a new memory replace an old one? Did a successful tool call actually finish the task? Is a retry about to create the same invoice twice? Ten runnable experiments in building agents that know what to do next—and when to ask. Each combines a small Jev judgment with application logic you can read, change, and test. By Really Artificial · JavaScript · Node.js 20.17+ · Zero dependencies · MIT.

Quoted from the project's README. Read the full README.

Highlights

  • Every example ships a readable policy, contrasting inputs, intended outcomes, a deliberately simple baseline and a known failure mode.
  • Fixture mode needs no key, account or network, so you can read the decision logic before any of it reaches the model.
  • Reports record exact requests and answers, source file hashes, intended outcomes, baseline comparisons and an option-order check.
  • The check flag returns a failing exit code on disagreement with the intended outcome, and reports refuse to overwrite a file.

Watch out

The headline result is not a Jev result. The README is explicit that a 34/34 fixture pass is a local software check, because the responses were authored to illustrate the policies rather than obtained from the model, and that live model behavior has not yet been verified. Fixture runs exercise request shapes, parsing, policy branches and reporting but establish nothing about accuracy, latency, calibration or cost; live runs add elapsed time and token usage and still establish no general accuracy or production-ready threshold. The thresholds in each policy are teaching choices rather than validated operating points, and a high confidence value does not certify correctness. A score value can be fractional because it is the probability-weighted position across levels, so the same mean can hide different uncertainty. The shipped cases contain fictional data, and substituting your own sends that state to TypeSafe on a live run. The code and docs were substantially authored with an LLM. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

2GitHub stars
Three dependency-free files that put Jev in a browser agent's hot path: code builds the choices from a table of visible controls, Jev picks an operation and target in one forward pass, and the code executes the pick.
Jev Demos & Experiments#community#jevbest#catalog-import
21GitHub stars
Pi guard plugin that checks inputs and tool calls with local rules plus Jev semantic judgements, with redaction and verification reminders.
Jev Demos & Experiments#community#jevbest#catalog-import
24GitHub stars
Native macOS computer-use agent on Apple silicon that finds controls with local CoreML detection, OCR, and accessibility, then asks Jev to pick the next finite-choice action.
Jev Demos & Experiments#community#jevbest#catalog-import
114GitHub stars
Browser agent where Jev picks each action from indexed page candidates while an LLM plans and reads, returning cited answers or schema-checked data with credential controls.
Jev Demos & Experiments#community#jevbest#catalog-import
3GitHub stars
Browser agent where Jev makes every navigation decision and agent-browser performs clicks and typing, stopping when the named goal page is reached or blocked.
Jev Demos & Experiments#community#jevbest#catalog-import
5GitHub stars
An MCP server that lets Claude or Codex drive a real Chrome window: Claude plans the task and Jev picks each click, keystroke and scroll from a numbered element list, with irreversible actions held for confirmation.
Jev Demos & Experiments#community#awesomejev#catalog-import

From the Jev guides

Original write-ups on what this entry is for.

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.
Jev Demos & Experiments#gaming#agents#control
Read article
Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai