Would you actually let a real-time AI tutor teach your 5-year-old every day? I’m more worried about the behavior than the answers — if it gets too sticky and engaging, that feels like a problem fast, and if the safety layer interrupts too much, the whole thing will feel dead.
For kids that age, I almost think “teaching” is more about trust and motivation than correct responses. What guardrails would you want before you’d even try it?
source: Teaching a child in <1000 ms: the architecture behind a real-time tutor | Ello
