My entire AI workforce lives in local working folders: fourteen employees, from accounting to content, each with their own job, personnel file and knowledge. That is an architecture decision. Because the question that changes your day-to-day is not which model leads the benchmarks. It is: can I turn this AI into an employee that knows my business and delivers every week?
That takes three things. First, a place for permanent knowledge: files and folders on my own computer, not a knowledge store owned by one provider. Second, definable ways of working: with Skills I write work instructions once, and after that they are infrastructure. Third, a tool that can work directly with those files. Today I often use Claude Code for that. Because the role lives in the folder, I can switch to Codex or an open model for the same employee.
I’m a developer and haven’t written a single line of code myself in 1.5 years. What I do instead: hire, onboard and manage employees, just digital ones. On this page you’ll find all my Claude content in one place, from the price comparison to Projects and Skills to making sense of Claude Cowork. If you don’t know the principle behind it yet, start with the fundamentals page Hire an AI employee.