factory-ai-autonomous-coding-agents¶
Insight: Factory.ai builds "droids" — autonomous software engineering agents that handle code generation to incident response. Key architectural decisions: browser-based (not IDE-based) for enterprise accessibility, test-driven development as verification mechanism, and retrieval over large context windows for cost efficiency. SWE-bench as an evaluation benchmark is considered dead by practitioners building production agents.
Detail: Factory.ai raised a $15M Series A from Sequoia. Their "delegation vs collaboration" framing distinguishes their approach from IDE copilots: rather than assisting developers, droids take full ownership of tasks. Enterprise metrics focus on code churn and ROI. For large refactors and migrations, they use parallel execution. This represents the "teammate agent" category in the AI coding taxonomy (alongside Devin, Cosine), distinct from CLI agents (Claude Code) and AI IDEs (Cursor).