Insights on product thinking and engineering excellence • 4 articles published
Every team with a legacy codebase hits the same wall: you cannot add tests without refactoring, and you cannot refactor without tests. Characterisation tests are the specific tool — invented by Michael Feathers, amplified by AI — that breaks the loop.
Every new model release prompts the same question — is this the one that finally makes AI coding agents reliable? It's the wrong question. What keeps single-agent workflows from scaling to production is architectural, not about the model. Here's the pattern the teams shipping real code keep converging on.
AI coding agents don't understand your architecture — they pattern-match from your codebase. One boundary violation becomes a template that AI replicates hundreds of times. Clean Architecture went from best practice to survival requirement.
Your team adopted GitHub Copilot, spent $180K on licenses, and expected faster delivery. Instead, lead times got longer, defect rates climbed, and your senior engineers are spending more time reviewing AI-generated code than writing their own. Here's why — and what the fix actually is.