We built an AI self-reflection bot for teacher trainees in India — deployed on Telegram, tested across two iterations with real users. The results challenged our assumptions about what AI should do in education.
Five years of pilots across 6,000+ students revealed that the upstream problem isn't skill deficit but agency deficit. Here's the chain that actually moves students from information to action.
Before I started working with education orgs on AI adoption, I assumed guardrails would be the hardest problem to solve. I was wrong. The real blocker is predictable pricing.
A chance conversation at a Bangalore mixer cracked open a mental model I hadn't fully articulated before. Why most successful Indian startups are built on cheap labor, not product, and what that means if you're building an AI company.
We set out to build an affordable AI tutor for UPSC test prep. After 9 months, thousands of users, and a 99% drop-off, here's what we learned about trust, motivation, and the real role of AI in education.