Back to Devoured - October 05, 2026
DEVOURED
Jev plays Pokémon Red for $1.65 model bill (GitHub Repo)

Jev plays Pokémon Red for $1.65 model bill (GitHub Repo)

Data aillmtypescript GitHub
An AI agent successfully played through Pokémon Red in under 38 hours for a total cost of $1.65 in model inference fees.
What: Christian Mathiesen created 'Jev,' an agent that interacts with a Game Boy emulator by reading RAM and choosing actions via an LLM. The bot completed the game in 37 hours and 40 minutes, using approximately 39.2 million tokens, and navigated menus, battles, and story milestones without external cheats.
Why it matters: This project demonstrates how LLMs can act as autonomous decision-makers for legacy software by treating game state as a structured observation space, providing a low-cost blueprint for building complex task-oriented agents.
Deep dive
  • The system uses a Node-based Game Boy emulator to expose RAM state as JSON to the model.
  • A harness manages basic pathfinding (A*) and menu navigation, while the LLM handles high-level decision making.
  • The agent makes roughly 800 to 1,300 calls per hour, costing between $1.00 and $1.70 per 24-hour period.
  • House rules were applied to ensure consistent speed and eliminate animation overhead.
  • The codebase uses 'loop protection' to prevent the model from repeating failing choices indefinitely.
  • It relies on a local disassembly of Pokémon Red to map memory addresses to game-readable symbols.
Decoder
  • Harness: A software wrapper that provides an environment for an agent to perform tasks, handling the interface between the game's memory and the AI model.
Original article

Jev Plays Pokémon Red

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.

Landing page: jev-pokemon.vercel.app

Result

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.
Total time 37h 40m
Decisions made 16,150
Input tokens ~39.2M
Total Jev cost ~$1.65
Median decision time ~0.4s
Team wipes 16 (14 at the Elite Four)
Elite Four attempts 15
Final team Charizard 83, Graveler 62, Nidoqueen 45, Beedrill 44, Haunter 39, Primeape 29

You need your own legally obtained copy of Pokémon Red. No ROM is included or distributed here.

How it works

 Game Boy emulator (Node) ──► read RAM ──► game state (map, party, battle, on-screen text)
          ▲                                          │
          │ button presses                           ▼
   harness mechanics ◄── Jev picks one ◄── legal options + facts
   (A* pathfinding, menus)                (type matchups, damage estimates,
                                           "2 areas toward the objective", ...)
  • Jev decides:
    • every menu answer (names, starter, YES/NO, shop, heal, learn/forget moves)
    • what to focus on (progress, heal, train, catch, shop, explore, team)
    • which Pokémon to catch, and which to swap in and out of the team at the PC
    • where to go and who to talk to
    • every battle action (move, switch, item, ball, run)
  • The harness never decides:
    • It only reads memory and presses buttons. It never writes to game memory.
    • Game knowledge is limited to the story milestones (what the next goal is and where it happens). Progress is checked against the game's real event flags.
    • Hidden items aren't shown to Jev, since a human wouldn't know where they are.
  • House rules:
    • Text speed FAST and battle animations OFF, set once in the Options menu at boot.
    • The player is named JEV and the rival BLUE.
    • Every caught Pokémon gets a nickname: Jev spells it one letter at a time (A–Z or DONE). It has to be a made-up name, not a species name or a nickname already in use.
  • Loop protection:
    • 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.

Setup

Requirements: macOS or Linux, Node 20+, and git.

git clone https://github.com/christianmat/jev-pokemon && cd jev-pokemon
npm install
npm run setup                 # installs rgbds (via brew), clones + builds pret/pokered, generates symbol data
cp /path/to/your/pokered.gb roms/red.gb   # your own dump of Pokémon Red (US/EU)
cp .env.example .env          # then fill it in

The ROM must be the US/EU release. Its SHA-1 is ea9bcae617fdf159b045185467ae58b2e4a48b9a, the same build pret/pokered produces. The harness reads RAM addresses from that disassembly, so other versions won't work.

.env

Variable What it does
JEV_MODE gateway for real Jev via Vercel AI Gateway, mock for a free, dumb stand-in (the default)
AI_GATEWAY_API_KEY Vercel AI Gateway key (VERCEL_AI_GATEWAY_API_KEY also works)
JEV_MIN_INTERVAL_MS, JEV_MAX_PER_MIN Throttling (defaults 300 ms and 90 per minute)

Run

npm start -- --speed 1                         # real-time; local viewer at http://localhost:8787
npm start -- --speed 1 --resume                # continue from the latest milestone checkpoint
npm start -- --speed 1 --load <save-name>      # load a specific save from saves/
npm run headless -- --steps 3000               # max speed, logs only
npx tsx scripts/tools/save.ts                  # save the running game (writes saves/manual-*.json)

Logs

  • logs/jev-calls.jsonl has every Jev call: the full state, the options with their facts, the probabilities, and the latency.
  • logs/events.jsonl has maps, milestones, saves and errors.

Cost

  • Price: Jev costs $0.042 per million input tokens, and output is free. A typical call is about 1,200 tokens.
  • Rate: at real-time speed the bot makes about 800–1,300 calls an hour. That comes to about $1–1.70 per 24 hours.
  • Ceiling: the throttle's worst case, 90 calls a minute nonstop, is about $7 a day.

Project layout

Path What
src/emu/ emulator wrapper (serverboy / GameBoy-Online core), save states, audio tap
src/game/ ROM tables, RAM reader, collision grid + A*, region graph
src/jev/ Jev client (throttle, cache, log), AI SDK gateway backend, mock
src/agent/ mode detection, dialog and menus, overworld, battle, field moves and items
src/knowledge/ story milestones
src/server/ runner + local WebSocket viewer
web/ local viewer page
site/ public landing page (static; deploy with Vercel, root directory site)
scripts/ setup, data generation, debug tools

Deploying the landing page

site/ is plain static HTML with no build step:

  1. Create a Vercel project from this repo.
  2. Set Root Directory to site.
  3. Deploy.

Legal

  • No ROM included. This repo doesn't contain or link to Pokémon Red, and you need your own legally obtained copy. roms/ is gitignored.
  • No Nintendo assets. Game data (symbols, names, maps, font) is read at runtime from your ROM. The symbol file is generated locally from the pret/pokered disassembly. That disassembly isn't redistributed here.
  • Not affiliated. Pokémon is a trademark of Nintendo, Creatures Inc. and GAME FREAK inc. This is a fan experiment, not affiliated with or endorsed by Nintendo, Game Freak, The Pokémon Company or TypeSafe AI.
  • License. GPL-2.0-or-later, because it builds on the GPL-licensed serverboy / GameBoy-Online emulator core.

Made by Christian Mathiesen at Frigade.

Explore all 67 articles in today's digest

Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi