I write about software engineering practice, the expert witness work I do, and the technical disciplines that turn out to matter once a software project ends up in dispute.
Below is a selection of recent pieces, chosen rather than listed by date. There is more on the blog: the pieces here are the ones I would point a first-time reader to.
How to Prepare a Legal Case With AI, Without a LawyerThe argument that AI puts the pre-action groundwork of a legal case within reach of people who cannot afford a lawyer. Used with a disciplined verification process, it readies the evidence and the correspondence; the trust sits in that process, not the model.
How I Built an AI Platform for ServiceNow Incident AnalysisWhat looked like a requirements problem turned out to be a quality problem hiding in 2,000+ ServiceNow incidents. A nine-month Solutions Architect engagement built an AI platform that reduced developer triage time by 80%.
Agentic Coding at Enterprise Quality and Startup SpeedSenior-quality code from an AI agent at speed depends on the discipline around the agent, not the agent itself. The quality gate is the specification; agentic coding is recast as a requirements problem, and as offshore delivery at far greater speed.
The Business Analyst Role Is CollapsingWhy the five specialist software roles, from product owner to developer to operations, are collapsing into one as AI removes the barriers that kept them apart. The surprise in the argument is that it raises the bar for expertise, rather than lowering it.
Lessons from Building a Safe AI Mental Health CoachInsights from a coach made for personal use, the piece turns on an uncomfortable line: making such a tool work for yourself is easy; making it safe for others is the unsolved part. The hard problem is never the model, but everything around it.
The full archive, in the order it was published, is on the blog.