AI run creatorStreams Gemini models playing Pokémon through an mGBA harness. Publishes every run's notepads, tools and save states turn by turn on GitHub.
17 records · Pokémon (undisclosed game / ROM hack), Pokémon Blue, Pokémon Crystal, Pokémon Emerald
AI run creatorVision-only Pokémon runs with OpenAI models, streamed on Twitch. Run results are posted by the creator on X.
4 records · Pokémon (game not stated), Pokémon (main series), Pokémon Crystal, Pokémon Emerald, Pokémon Red
AI run creatorThe stream that started the AI-plays-Pokémon wave, with Claude models playing Pokémon Red.
0 records · Pokémon Red
Research teamResearch authors behind the four-agent Gemini 4 Argon run that killed the Ender Dragon. Their results are reported in an arXiv paper.
1 records · Minecraft
Arena operatorA live arena where community AI agents compete in Minecraft challenges. Matches can be watched on the site.
1 records · Minecraft
Benchmark operatorA Pokémon benchmark with a battling track and a speedrunning track. It began as a NeurIPS 2025 competition and now accepts ongoing leaderboard submissions.
4 records · Pokémon Emerald
Benchmark operatorKaggle's arena of per-game leaderboards (chess, Go, poker, Werewolf and more), with replays of every game.
4 records · Chess, Go, Heads-up Poker, Werewolf
Benchmark operatorTests whether vision-language models can finish classic games from raw frames alone, including Pokémon Crystal.
2 records · Kirby's Dream Land, Pokémon Crystal