User repository on GitHub
Jev, an AI decision model, plays Pokémon Red. It beat the game in 37h 40m.
- Forks
- 10
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Demos & Experiments
Jev Plays Pokémon Red
A completed Pokémon Red run: Jev picked menus, catches, and moves from RAM, and the harness handled pathfinding. The public page records the Hall of Fame stats.
- Stars
- 127
- Language
- TypeScript
- License
- GPL-2.0
- Published by
- Community
- Author
- Christian Mathiesen
- Last verified
- Sep 2026
About this project
Jev), TypeSafe AI's decision model, plays Pokémon Red. There are no scripts or cheats: the harness reads the game's memory, lists the legal options with some facts about each, and Jev picks one. Jev beat the game. The live stream ran on YouTube from Sep 25 to Sep 26, 2026 and has ended. The highlights are on the landing page. every menu answer (names, starter, YES/NO, shop, heal, learn/forget moves) what to focus on (progress, heal, train, catch, shop, explore, team)
Quoted from the project's README. Read the full README.
Works forJev for Realtime control
Used inJev for Gaming
- Languages
- JavaScript
- Capabilities
- Open sourceRequires API key
Highlights
- The run finished in 37h 40m: 16,150 decisions, about 39.2 million input tokens, and about $1.65.
- A typical decision took about 0.4 seconds. The team wiped 16 times, and the Elite Four took 15 attempts.
- Jev chooses menus, what to focus on, which Pokémon to catch, and moves. Pathfinding stays in the harness.
- The final team was Charizard 83, Graveler 62, Nidoqueen 45, Beedrill 44, Haunter 39, and Primeape 29.
Watch out
GPL-2.0-or-later because of the emulator core. No ROM is included. The landing page carries a Frigade ad. Distinct from valentynkit/jev-plays-pokemon-red. The YouTube stream ended on 26 September 2026. Not affiliated with Nintendo, Game Freak, or The Pokémon Company.
Questions about this entry
Short answers drawn from the linked project or page.
It reads what a player could see. The harness reads RAM and presses buttons and never writes to game memory, hidden items are withheld because a human would not know where they are, and the only game knowledge passed in is the list of story milestones, with progress checked against the game's real event flags.
About the same order as the recorded run: roughly $1.65 for some 39.2 million input tokens across 37 hours 40 minutes. The README puts real-time speed at about $1 to $1.70 per day, and the throttle's worst case of 90 calls a minute nonstop at about $7 a day. Output is not billed and a typical call is about 1,200 tokens.
Yes. JEV_MODE defaults to mock, described as a free and deliberately dumb stand-in, so the emulator, harness, and viewer work with no key. Setting JEV_MODE=gateway with AI_GATEWAY_API_KEY switches in the real model; JEV_MIN_INTERVAL_MS and JEV_MAX_PER_MIN throttle it.
macOS or Linux, Node 20+, and git, then npm run setup, which installs rgbds through Homebrew and clones and builds pret/pokered to generate symbol data. You must supply your own legally obtained US/EU ROM: the harness reads RAM addresses from that exact disassembly, so other versions will not work, and no ROM ships with the repo.
Jev beat the game in 37h 40m across 16,150 decisions and about 39.2M input tokens, at roughly $1.65 total. It suffered 16 team wipes, 14 of them at the Elite Four, over 15 Elite Four attempts, and finished with a team led by a level-83 Charizard.
Options that were already tried without anything changing get tagged, and Jev is told to try something new. If the same failing choice keeps coming back, the harness samples an alternative; as a last resort it reloads the latest milestone checkpoint.
It only reads memory and presses buttons; it never writes to game memory. Game knowledge is limited to the story milestones in src/knowledge/milestones.ts, progress is checked against the game's real event flags, and hidden items are not shown because a human would not know where they are.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker 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: 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: 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
185GitHub stars
A local site, games, a benchmark, and an agent skill for Laya. The public pages replay recorded runs; the clone runs the model on your machine.
Jev Demos & Experiments#community#javascript#open-models
Community
wdobry/laya-playgroundJavaScript
wdobry/laya-playgroundJavaScript584GitHub stars
A Mineflayer agent for Minecraft Java 1.16.5 where a planner sets the objective and Jev picks the next bounded action from game observations.
Jev Demos & Experiments#community#javascript#gaming
Community
rmalde/minecraft-agentJavaScript
rmalde/minecraft-agentJavaScript4GitHub stars
A Craftax (Crafter) survival harness where deterministic code lists every feasible action with its facts and Jev picks one Choice per step, optionally guided by an LLM-written objective and standing rules; three-seed ablations are included.
Jev Demos & Experiments#community#craftax#games
Community
mansicer/jev-playsPython
mansicer/jev-playsPython0GitHub stars
Browser Pac-Man race where a deterministic three-junction simulation filters fatal routes, then Jev makes one typed Choice among the surviving route IDs while the server rejects any answer outside the supplied set. Self-hosted Laya runs the same contract for comparison.
Jev Demos & Experiments#community#gaming#simulation
Community
MaryNfs/pacman-ai-raceJavaScript
MaryNfs/pacman-ai-raceJavaScript36GitHub stars
A Clash Royale bot that turns a phone screenshot into JSON once a second and asks Jev for a strategy, a card, and a square.
Jev Demos & Experiments#community#python#gaming
Community
bytelabs-oss/clash-jevPython
bytelabs-oss/clash-jevPython2GitHub stars
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/FlightBenchPythonFrom 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
From the Jev community
Posts from builders shipping with Jev right now.