We are running our first pilot at Pragya Kunj School, and rather than wait for a tidy success story, we want to share what it actually looks like from inside. Building software in a real classroom is humbling: the things you were proud of go unnoticed, and the small frictions you overlooked turn out to matter most.
What held up
The Socratic approach worked better than we feared. We had worried that students would simply demand the answer and resent the questions. Most did not. Once a few of them experienced getting unstuck by their own reasoning, the hint-by-hint approach stopped feeling like an obstacle. Teachers also took quickly to seeing where a class collectively stumbled, and used it to plan the next lesson.
What we got wrong
- We underestimated how often the connection drops, and how much had to keep working offline
- Our first नेपाली phrasing read stiffly, and students noticed before we did
- Teachers wanted fewer numbers on the dashboard, not more — clarity over completeness
- Onboarding the curriculum took longer than we had assumed for a busy teacher
The teachers redesigned things for us
Some of our best changes came straight from the staff room. A teacher pointed out that the analytics page answered questions she never asked while burying the one she asked daily — which students fell behind this week. We had built for completeness; she needed focus. We reorganised around her question, and the page became genuinely useful.
What surprised us about the students
The practice games drew students back on their own, outside class, more than we expected — the streak and the badges did real work. But we also watched students who would happily play for an hour, which is exactly why a feature like Digital Sunset earns its place. Engagement is a tool, and like any tool it needs a sense of when to stop.
Building in the open
We are not going to pretend the pilot is finished or flawless. It is a conversation with one school that is teaching us how to be useful to many. We will keep writing these notes honestly, including the parts that did not work, because a tool used by children should be built carefully and in plain view.