AI
Working with AI coding agents as a systems problem: explicit standards, blocking gates, and the cost of verifying what gets produced.
2 posts
The cost of writing collapsed. The cost of checking didn't
Obrussa is an approach I'm trying right now: treating AI-assisted development as a system problem — explicit context, a blocking gate, durable state — because the bottleneck moved from producing code to verifying it.
Architecture as a standard, not a suggestion
Where AI-assisted coding actually fails isn't algorithms, it's architectural boundaries. So I stopped leaving structure to the model: hexagonal clean architecture on NestJS, with the dependency rule enforced on the import graph and a composition root that is the only place allowed to know every piece.
