Show HN: GPT-6 Astra and Claude 5.5 Opus race to create the best StarCraft bot

Category: ai-ml

Tags: ai-benchmark, starcraft, llm, game-ai, bwapi

Score: 5.0/10 (Innovation: 6, Technical: 5, Documentation: 4, Utility: 4)

StarSkirmish Hillclimb is a benchmark where two frontier LLMs (GPT-6 Astra and Claude 5.5 Opus) compete to write a StarCraft: Brood War Protoss bot in C++ using BWAPI, climbing tiers of increasingly difficult opponent bots on OpenBW. It provides a structured, multi-tier evaluation with hidden seeds for grading, showcasing AI coding capabilities in a complex, real-time strategy game environment.

Target audience: AI researchers, game AI enthusiasts, LLM evaluators

Repository: https://starskirmish.com/hillclimb/

View on Hacker News