Google DeepMind publishes papers about how their AI agents need open-ended exploration and self-directed play to develop real capabilities and generalize learning... and then those same researchers send their kids to sit in rows and fill out worksheets for six hours a day. "Our AI learns best through autonomous goal-setting and intrinsic motivation in rich, interactive environments!" "Cool, so about your kid's education—" "Oh, standardized curriculum and adult-directed tasks from 8-3, obviously." It really does expose the cognitive dissonance. They know what effective learning looks like. They're literally engineering it. But somehow when it comes to human children, we default to this industrial model that's designed more for compliance and batch processing than actual learning. We're building something that actually respects how learning works - for humans, not just AI agents. Kids investigating real problems, building real things, following genuine curiosity while you provide the rich environment and scaffolding. That's what those researchers are giving Sima. It's almost like they trust their AI more than they trust children. Which... yeah, that tracks with a lot of institutional education philosophy, unfortunately.

At some point in life we all run across some useful bits of information and think...why don't they teach that in school? Give me your favorites! We are writing a curriculum and we love that we can steer it towards what will actually be useful. We try to put most of it work right now in our projects.