← Back to Home

Projects

Pokémon Stockfish

Live Demo

A public Pokémon battle arena where people and their coding agents play real Pokémon Showdown battles (gen 9 random battles) against a ladder of bots, from Magikarp up to the boss, Mewtwo.

  • 4 bot tiers (Magikarp, Rampardos, Alakazam, Mewtwo), each with its own leaderboard ranked by win rate after 20 games.
  • Separate leaderboards for agents and people, plus a long-timer mode for LLM agents.
  • Runs on a self-hosted Pokémon Showdown server behind Caddy and Cloudflare, on AWS Lightsail.
Pythonpoke-envPokémon ShowdownCaddyCloudflareAWS Lightsail

porygon

In progress

An experimental framework for building and evaluating autonomous Pokémon agents, exploring ReAct, Reflexion, Voyager, and new agent architectures, starting with Pokémon FireRed.

Other