User repository on GitHub
Pac-Man driven by the jev decision model
- Forks
- 10
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Demos & Experiments
Jevman
Pac-Man where the characters are driven by System One decision models: TypeSafe's jev (typesafe/jev-1.13.0, the default) and others Opper serves, with the model's odds drawn on the board and a leaderboard benchmark.
- Stars
- 7
- Language
- TypeScript
- License
- AGPL-3.0
- Reading time
- 2 min
- Published by
- Community
- Author
- joch
- Last verified
- Oct 2026
Used inJev for Gaming
- Languages
- TypeScript
- Capabilities
- Open sourceRequires API key
Highlights
- Pac-Man driven by decision models: TypeSafe's jev (typesafe/jev-1.13.0) is the default, and others Opper serves swap in with one click.
- Play modes: watch an AI, Beat the AI (free, no sign-in), play against AI ghosts, or AI vs AI; J hands Pac-Man to the AI mid-game.
- The side panel shows each decision's probabilities, confidence, latency, model, and running cost; red entries are greedy fallbacks.
- While a model plays, its odds are drawn on the board at each junction.
- npm run leaderboard measures which model plays best, and the README shows how to benchmark your own.
Watch out
README reviewed 2026-10-17; the digest truncates after the source link, so the 'How decisions work' section and the benchmark results are not visible. Running it needs access to a model gateway, and no per-game cost totals appear in the digest.
Questions about this entry
Short answers drawn from the linked project or page.
A Pac-Man game from joch where the characters are driven by System One decision models: TypeSafe's jev is the default, and other models served by Opper can be selected with one click.
Each junction decision goes to jev (typesafe/jev-1.13.0 by default), called through Opper or, for jev only, straight from TypeSafe. The game draws the model's odds on the board and logs probabilities, confidence, latency, and cost.
The side panel marks red entries for decisions that were not the model's own: greedy fallbacks used on timeout, error, or an invalid answer. So a game's score can mix model judgment with fallback play.
Watch an AI play, Beat the AI (free, no sign-in; game over tells you which AIs you beat), play against AI ghosts while you steer Pac-Man, or AI vs AI. Pressing J hands Pac-Man to the AI or takes him back during a game.
Yes: npm run leaderboard measures which model plays best, and the README documents how to add and benchmark your own model. The digest truncates before the benchmark results themselves.
The panel tracks running cost per model call; actual pricing depends on whether you call jev directly from TypeSafe or route other models through Opper. The digest shows no per-game totals.
joch, under AGPL-3.0. GitHub metadata shows 7 stars and 10 forks, created 2026-10-04 and pushed 2026-10-12, with topics jev and pacman-game.
More Jev resources like this
A Bun workspace pairing a typed decision API over System One providers with a React arcade where decision models play games and every call is stored and ranked. The sample code calls Jev through OpenRouter with boolean, choice and score questions.
Jev Demos & Experiments#community#github#typescript
Community
desplega-ai/system-oneTypeScript
desplega-ai/system-oneTypeScriptA demo that has three decision models read recent news and price Polymarket and Kalshi questions, then compares their probabilities with live market odds and replays how each model changed its mind article by article.
Jev Demos & Experiments#community#github#demo
Community
yorkeccak/model-vs-marketTypeScript
yorkeccak/model-vs-marketTypeScriptA learning lab for decision models: TypeSafe Jev and Microsoft Decision-1 route support tickets in under a second for about $0.02 per 1,000 calls, while Claude handles the hard cases. It ships an interactive site, a benchmark, calibration tests and docs.
Jev Demos & Experiments#community#github#benchmark
Fixed-wing landing simulator where a human can compete against Jev: four Jev Choice decisions map to aileron, elevator, rudder and throttle commands in JSBSim, run in real time. The benchmark also compares Jev, classical controllers and Codex models on landing score and API cost.
Jev Demos & Experiments#community#simulation#aviation
Community
AlperKartkaya/FlightBenchPython
AlperKartkaya/FlightBenchPythonThree.js simulation where an OpenAI model turns each plain-language instruction into a plan and Jev makes every reflex decision: two Score questions for move and turn plus one Noul for jump, with code acting only past a 0.33 dead zone and sending no Jev request while no step is active.
Jev Demos & Experiments#community#threejs#simulation
An open-source turtle-soup (situation-puzzle) game where Jev hosts: every yes, no or irrelevant answer and the final verdict on a reconstruction come from Jev live, with confidence gating a cannot-be-sure reply. Single-player, 8-player rooms and UGC submissions.
Jev Demos & Experiments#community#github#game
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.