Factory Droid in Practice: What the Public Record Shows
Builders praise Droid for running many vendors' models in one agent and complain most about reliability. Team accounts are rare, so a pilot has to measure team results itself.
Research
What is changing across AI coding agents?
Reading Factory's Benchmark Claims
Factory submitted every Droid score on the public Terminal-Bench leaderboards itself. What each Terminal-Bench, Legacy-Bench, code review, and Factory Router figure measures, what outside testers found, and how to test the claims on your own repositories.