A believable mistake
“Seven is bigger than one, so 27 must be greater.”
Learning works differently on Planet Zorp
The student is an alien.
The teacher is you.
Children aged 5–8 teach Zorp through real maths and spelling misconceptions. A right answer is not enough—they have to fix the idea underneath.
Watch Zorp learn ↓Live game runs locally with Codex
ZORP THINKS27 > 31
YOUR MISSIONTeach me why!
The crown jewel
Zorp is wrong for a reason.
Every mistake comes from a specific flawed rule found in real childhood learning research. Zorp holds it, uses it consistently, and says the reasoning aloud.
The child teaches why
A separate judge looks for the idea—not a copied answer.
Learning that lasts
Two problems later, Zorp silently tries the repaired skill again.
A real teaching mission
Watch the idea change.
Open up the machine
Magical for children.
Inspectable for grown-ups.
Two minds, one learning loop
The brain cannot mark its own homework.
Zorp’s brain generates the consistent misconception. A fresh Codex call judges the child’s explanation against explicit criteria. TypeScript—not the model—owns every state change.
Child teaches
→Zorp brainwrong rule
→Correction judgecriteria evidence
→Durabilityfresh problem
Try the real thing
Bring your own
Codex login.
The live game stays on your Windows computer. No child account, stored audio or paid inference service.
codex loginnpm installnpm run demo
Tested in current Edge and Chrome. Talking is optional; typing always works.