How do you build a software factory around coding agents? It starts with harness engineering: giving agents the context, tools and feedback they need to produce verified changes. Drawing on practical examples, this talk shows how technical checks and outcome grading support longer stretches of autonomous work, with human judgment where it matters.
A software factory is an operating model for how a team builds software and learns from the results. The factory itself needs engineering, too: you improve it alongside the products it builds, turning recurring failures into better context and stronger checks.
We'll close with a glimpse of Loop Intelligence and how evidence from agent work helps teams decide what to improve next.
