AI software factories are back: Prototypes become mass-produced apps

The software factory concept, popular around 2008, is being revived by AI. CloudBees CEO Moritz Plassnig told ZDNET that coding previously created bottlenecks—you had to hire expensive engineers. Now, with foundation AI models and agentic coding, "we're less constrained on the coding side." Software factories automate the development lifecycle into repeatable chunks, taking a vibe-coded prototype, validating it, testing it, and mass-producing it for wide distribution, much like a car factory produces vehicles from a prototype.

According to Jaymin West, a forward-deployed engineer, leading AI companies—Anthropic, Cognition, Cursor, Factory, Google, GitHub, OpenAI, and Ramp—have independently converged on the same software factory architecture over the past 18 months. This pattern structures their entire engineering framework. Plassnig explains that anyone with an idea—say, a customer support rep spotting a pattern—can submit it to the factory. The first iteration gets created, product and engineering teams review it, and the factory refines and scales it. Developer craftsmanship isn't disappearing; instead, engineers shift from writing code to exercising judgment about what should be built and shipped.