Benchmark Environment Design

A 1MB script that never looks at the screen beats frontier models on computer-use benchmarks

A preprint proves a blind replay agent's expected score equals the source model's pass@k, which means static computer-use leaderboards have been grading environment determinism.