Factory 2.0 picks up where the original Factory left off in 2023, shifting the focus from speeding up individual engineers to creating a full-scale “software factory.” Matan Grinberg and Eno Reyes outline a system made of AI agents that handles everything from incoming bug reports and customer feedback to code reviews, security scans, deployments and monitoring. Each step feeds back into the next, creating an end-to-end loop that the factory refines over time. They claim this loop remains mostly uninstrumented outside a few early adopters—but expect rapid growth as organizations chase broader productivity gains. Three core requirements anchor their vision. First, model independence: enterprises pick or auto-route among AI models based on cost, speed and accuracy instead of committing to one. Second, sovereign intelligence: every session, code review and incident stays within an organization’s own data plane—cloud-hosted, key-managed or fully air-gapped. Third, continual learning: a shared agent core and model router lets a security finding trigger a documentation update, or a deployment prompt a QA run. Each automation or integration automatically benefits all teams because they draw on the same platform and context. Factory says it has already deployed these software factories at companies including NVIDIA, EY, Adobe, Palo Alto Networks, Adyen, Blackstone, Wipro and Comarch. They stress that autonomy rolls out gradually—starting with simple Droid agents on well-defined tasks, then ramping up to multi-agent “Missions” that run for hours or days. The goal: engineers shift from writing every line of code to building and governing the factories themselves. If your team wants in, you can reach out via their contact page or use the Factory desktop app to manage your own factory.