A Real Problem Is Not Automatically an Urgent One
The problem was real at every stage. It still was not urgent enough to buy. What 50+ conversations, three pivots, and Maat taught me about validation.
Writing on architecture, AI delivery, developer experience, and technical communication.
The problem was real at every stage. It still was not urgent enough to buy. What 50+ conversations, three pivots, and Maat taught me about validation.
I checked Cal.com's own documented architecture rules against their code, then pointed the same tool at a problem no linter can see: code that means two different things.
PR review is a code quality tool. Treating it as architecture enforcement isn't a discipline problem — it's a mismatch between what the tool can see and what enforcement actually requires.
You can write a perfect ADR and in 18 months nobody follows it. The question is not how to document better. It is what you do with the documentation after it is written.
GitClear's data shows code cloning is up while refactoring is down. That matters because duplicated code creates coupling, and refactoring is the work that forces teams to reconcile it.
I ran three experiments with Claude Code (Opus 4.6) and Codex (GPT-5.4 xhigh) — greenfield creation, extension, and brownfield detection. The conclusion: implicit coupling is a maintenance problem, not a generation problem.