LLM 'benchmark' as a 1v1 RTS game where models write code controlling the units
Summary
The article documents a 1v1 RTS-style benchmark where LLMs generate JavaScript code to control units in a simulated arena. It highlights the iterative write–play–review loop and reports results with Gemini 3.1 Pro leading, followed by Claude Sonnet 4.6 and GPT-5.3 Codex, underscoring insights into AI model capabilities and testing methods.