Fully autonomous WoW agent runs on Claude Haiku, downgrades models as it levels up
max_paperclips · x · 2026-11-11
Developer maxpaperclips demos a fully independent World of Warcraft agent running on Claude Haiku that levels up and handles realtime actions at very low cost, using the LLM as a "brain" that designs its own harness. Model ladder: start with Opus for coding, switch to Sonnet once confident, then Haiku or a local model; he's also testing decision models like Jev for fast decisions and RL-trained small policy models in Minecraft. The setup is an Azerothcore private server with a custom Bevy-based client (diverged from benilla), no real accounts harmed, and will be released publicly with addons.
More from coding & agent
- HydraDB builds a unified knowledge graph to give AI agents accurate context — vaibhavbetter · 2026-11-11
- Reddit Users Compare Notes on GLM-5.3-flash's Random Coding Mistakes — JustinHoMi · 2026-11-11
- Dev fine-tunes Qwen3-1.7B into a 1.83GB GLB 3D asset coder that runs on CPU — ConfidentDinner6648 · 2026-11-11
- Solo dev shares 20k-word workflow: Opus 5.5 + Seedance 2.5 for 4 viral product videos — dotey · 2026-11-11
- CodeGraph turns codebases into Neo4j graphs so agents can do dependency-aware refactoring — JeremyCMorgan · 2026-11-11
- Primitive imagines guardian agents that enforce team decisions infrastructure-wide — daniel_mac8 · 2026-11-11