Built by one person, on purpose.

Hi — I'm Alexander Mihalkevich. I run mihalkevich.com and FolderAI is my flagship product. This page exists because in 2026 it shouldn't be a mystery who built the thing you're paying for.

FolderAI started as a tool I needed myself: I kept hiring AI agents and watching them produce confident-looking work that didn't survive a real test. The fix wasn't a better model. It was a tighter loop — short verifiable sprints, written eval cases, forbidden-scope clauses. The same discipline a senior engineer already uses on themselves, just made explicit for the agent.

The simulators teach that discipline. Each chapter is one opinion, not a survey.

• Specs over vibes.
• Context is the model.
• The agent will lie if you let it.

You walk through a real-feeling situation — founder vanished, board demo in 72 hours, patient transcripts hallucinating a dose — and you make calls. The engine grades the calls that are gradeable; the theory shows up after you submit, not before.

Everything you see on this site was written and shipped by me, usually inside the same Ralph-loop discipline Chapter 1 teaches. The content was authored in collaboration with the same kind of agent you'll be supervising after the course. That's deliberate: the only credible authority on prompting an agent is someone who keeps doing it.

If you want to see what I shipped this week, the changelog is public. If you want to start free, Chapter 1 of both courses is open. If you want to email me, a.hypesquad@gmail.com still reaches me directly.

— Alexander