Scaling AI Agents for Reliable Computer Automation

The Unreasonable Effectiveness of Scaling Agents for Computer Use