Show HN: Gemini 4 Argon plays StarCraft

	Show HN: Gemini 4 Argon plays StarCraft(starskirmish.com)
	242 points by __cayenne__ 9 days ago \| 108 comments
	tl;dr: StarSkirmish Bench v0.2 pits flagship LLMs against each other by giving them one hour to write a Protoss StarCraft: Brood War bot in C++, then scoring win rates against established human-written bots. GPT-6 Astra leads, followed by GPT-6.1 Sol (which offers the best cost/performance), with Gemini 4 Argon, Claude Opus 5.5, and Claude Fable 5.1 tied behind; however, all models remain far below top human-written bots, winning only 7% of games against the strongest ones. Scores correlate with how much each model practiced against strong opponents, and track Vals IOI/WeirdML better than mainstream benchmarks.
	HN Discussion: ↑Rankings correlate with other independent LLM benchmarks like 9x9 Go, validating the results •Curious what can be learned from the bot code itself and how much is memorized vs novel ↓Unclear what the benchmark actually measures given many confounding factors ↓Personal experience contradicts the ranking - Opus 5.5 outperformed GPT-6 models on their coding task ~Suggestions to extend the benchmark (compare harnesses, test unseen maps, benchmark against Pluto)