Building durable knowledge bases and agentic workflows for software work — Claude Code configuration, design specs, and the AI-as-CTO operating model.
Design specs aren't documentation overhead. When your CTO is an AI, the spec is your management layer — across six projects in different domains.
The 'SaaS is dead' narrative gets the diagnosis right and the prognosis wrong. AI makes code cheap, but it barely touches the parts that defend SaaS.
A hands-on guide to configuring Claude Code with subagents, skills, role definitions, safety protocols, and feedback loops.
AI coding assistants work in one repo but fall apart across projects. I built a 30-document knowledge base that gives AI full context in 30 seconds.