Insights
Field notes from the practice.
What we're learning in the work: private AI deployments, AI-era engineering discipline, cloud economics, and getting off legacy systems without a rewrite.
Why Private, On-Premise AI Is Having a Moment
Cloud AI APIs are the fast path, but for many organizations, the right path runs through their own hardware. Here's when private AI deployments make sense, and what they actually involve.
AI Writes the Code. You Are the Quality Control.
AI-assisted development multiplies output, including the output of mistakes. The teams winning with tools like Claude Code and Copilot are the ones who redesigned their review discipline, not just their tooling.
A Pragmatic Path Off Legacy Systems
Big-bang rewrites fail for predictable reasons. A staged modernization, stabilizing, carving, then replacing, gets you to a maintainable system without betting the business on a rewrite.