PreAct: Computer-Using Agents that Get Faster on Repeated Tasks

In three linesPreAct compiles successful runs of computer-using agents into small state-machine programs, replayed 8.5-13x faster with no per-step LLM calls. An independent evaluator validates each program before storage. Across three benchmarks (mobile, desktop, web), this verification prevents faulty program accumulation (+1.75-2.6 tasks).

Summary generated by Claude — human-verified