User repository on GitHub
Snake controlled by parallel Jev assessments, with one API call per game tick.
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Cookbooks
Snake Jev
A desktop Snake experiment where Jev assesses the board each tick and Python combines its answers into a move; one recorded run reached 29 foods in 461 ticks before trapping itself.
- Stars
- 2
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- siroccomask
- Last verified
- Sep 2026
About this project
A desktop Snake experiment powered by Jev / System One. Jev assesses the board; Python combines its answers into a move. The original p5 Snake game handles movement, food, growth, and collisions. One recorded run reached 29 foods in 461 ticks, then trapped itself in its body. That is one run, not a reliability guarantee. The agent detects immediate hazards; it does not plan a route through the full body layout.
Quoted from the project's README. Read the full README.
Works forJev for Routing
Highlights
- Each tick makes one real API request with nine questions: wall collision, body collision, and food progress for turn left, straight, and turn right.
- Code composes only Jev's answers: it prefers food progress among collision-free actions, then lowest assessed collision risk, then straight.
- Inputs per direction are moves_before_wall, body_distance_moves, and food_offset_moves; behind is observed but not an available action.
- Redacted traces in .jev_traces/ hold the request, the API answers, the code composition, and before/after board snapshots.
- The agent detects immediate hazards but does not plan a route through the full body layout, so an incorrect assessment can still kill the snake.
Quickstart
bash
git clone https://github.com/siroccomask/snake-jev.git
cd snake-jev
uv sync
cp .env.example .env
uv run python sketch.pyWatch out
MIT-licensed. Needs uv, a desktop with OpenGL support, and your own Jev API key (JEV_API_KEY); tested on macOS with Python 3.12, other platforms are unverified, and each tick may incur API charges.
Questions about this entry
Short answers drawn from the linked project or page.
Hard to separate. One API call per tick asks nine questions about wall collision, body collision, and food progress for left, straight, and right, and Python composes the move from those answers alone. The record is one run on one food seed, and the author calls it not a reliability guarantee.
Because the agent judges local hazards rather than routes. It detects immediate collisions but does not plan through the full body layout, so nothing stops it closing off space it will need later, and the code never inspects the board to overrule a wrong assessment.
One real billed request per tick, so that 461-tick run was 461 calls. The default has no request limit, though a --max-calls 100 flag bounds a session, and the game starts paused so you can step through single ticks with the keyboard.
Not the game itself, which needs a key in .env, uv, and a desktop with OpenGL, tested only on macOS with Python 3.12. The test suite does run offline using fake inference and two recorded fixtures, and redacted traces land in a local directory for debugging a bad run.
Wall collision, body collision, and food progress for turn left, straight, and turn right, nine independent questions in one API call. Python composes the move from those answers using distances supplied in whole-move offsets.
compose_assessments combines only Jev's answers: it prefers food progress among actions Jev assesses as collision-free, and if none are clear it chooses the lowest assessed collision risk, with ties preferring lower risk then straight. There is no second API call and code does not inspect the board to override Jev.
Pass --max-calls 100 to limit requests, --seed 2022 to set the food seed, and --until-death to start immediately. Redacted traces go to .jev_traces/, each with the request, API answers, separately labeled code composition and before/after board snapshots. Keys stay local and are redacted.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker News: Software Bombe and live Enigma demo where Jev ranks cribs and judges plaintext.
ag_212 pointsShow HN: Jev as the Bletchley Park Analyst: A Software Bombe on EnigmaRead the thread on Hacker NewsHacker News: Show HN for public Minecraft PvP where a diffusion LM picks moves from structured state.
2 pointsShow HN: Minecraft PvP against DiffusionGemma-as-Jev, 24 ms per decisionRead the thread on Hacker NewsHacker News: Show HN: Jev finished a Pokémon Red run, with the Hall of Fame stats on the landing page.
pancomplex272 pointsShow HN: Jev Plays Pokémon RedRead the thread on Hacker NewsHacker News: Show HN: a mouse colony where each next move is Jev or a fixed rule.
1 pointShow HN: Jev-mice – mouse colony simulation using Jev and deterministic engineRead the thread on Hacker NewsMore Jev resources like this
A side-by-side T-Rex game where local Laya runs on Apple Silicon through MLX and hosted Jev uses the TypeSafe API, showing choices, latency, survival streaks, and crash replays on one course.
Jev Cookbooks#community#python#games
Community
virajbhartiya/laya-vs-jevPython
virajbhartiya/laya-vs-jevPythonInteractive browser experiments that expose Jev's judgments: a support-message Decision Lab and a driving simulation where structured sensor state becomes typed steer, brake, and overtake decisions.
Jev Cookbooks#community#javascript#demo
Pokémon Red under PyBoy where ordinary Python owns the route and arithmetic and Jev is asked only at branches, with each battle turn's faint prediction scored by Brier.
Jev Cookbooks#community#python#games
An original Civilization II browser harness for Jev: named Choice vectors cover empire policy, cities, research, diplomacy, units, exploration, and warfare, with live decision probabilities in a Roman-themed spectator display.
Jev Cookbooks#community#python#games
Community
phyous/tsai-civ2Python
phyous/tsai-civ2PythonAn experimental controller that has Jev choose NES inputs to play Super Mario Bros. from structured emulator telemetry instead of screenshots, logging every decision with probabilities and latency.
Jev Cookbooks#community#python#games
Community
fhshaik/typesafe-marioPython
fhshaik/typesafe-marioPythonA System One harness that plays the original StarCraft shareware mission Strongarm by reading structured game state and issuing mouse and keyboard commands from Jev choices.
Jev Cookbooks#community#python#games
Communityphyous/tsai-scPython
phyous/tsai-scPythonFrom 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.