Currently

People reason in something richer than language and only translate to words when they need to communicate. Most machine intelligence does the opposite. ATOM asks whether this difference matters, whether the medium shapes not just how efficiently something reasons but what it can reason about at all. So far: it does, and there is a hard wall that shows up everywhere I look.

Lumin is my attempt to build a sanctuary for genuine mastery. Most tools track what you have consumed. Lumin tracks what you can demonstrate you understand. You capture what interests you, articulate it, and the system tests whether you actually know it. Nothing brightens until you have earned it.

Also exploring what happens when you replace textbooks with structured questioning from first principles, and whether machines genuinely replicate human social behaviour or just approximate it.