01 The problem: answers are cheap, thinking isn't
A Class 7 student in a Hindi-medium school can find the answer to almost anything in seconds. What they can't get is someone to notice how they reasoned to it — and stop them when the reasoning is wrong but the answer looks right.
Rote learning rewards the final answer. Every free tool — search, chatbots, YouTube — does the same: it hands over the conclusion and skips the thinking. So the one skill that actually compounds, being able to reason through a problem and catch your own mistake, gets no reps. Socho is built to force those reps, in the language the student actually thinks in.
Problem statement: "a learning platform for the marginalised." Instead of building a whole school, I cut to one sharp wedge: Class 6-8, Science + Social Science + Maths, bilingual Hindi/English, one question a day. Small enough to ship solo and get real feedback; deep enough to prove the core mechanic works.
02 The core mechanic: make them defend the answer
Every session is one question built around a common misconception — not a fact to recall. "The iron gate feels colder than the wooden door, so iron is at a lower temperature. Is she right?" The student picks the intuitive (wrong) answer, and that's where the product actually begins.
Instead of marking it wrong, Socho asks "Why do you think so?" — by voice or text. Then it reflects the student's own words back, names the precise gap ("feeling colder is about how fast heat moves, not the actual temperature"), and asks a follow-up question instead of giving the answer. The student reasons their way out. The home screen tracks the metric that matters: times you corrected your own thinking.
Misconception first
One daily question engineered around a belief students actually hold, not a definition to memorise.
Pick a side
Student chooses an answer — usually the intuitive, wrong one. Commitment is what makes the correction land.
"Why do you think so?"
Voice-first ("Tap to speak") or type. The student has to put reasoning into words, in Hindi or English.
Socratic reply
Reflects their words back, names the exact flaw, asks the next question — never just states the answer.
Self-correction
Student reasons to the right idea themselves. Logged as a "corrected your own thinking" event.
03 Design decisions that carried the product
- Bilingual by toggle, not translation. Hi / Hi-En / Eng switch on every screen. A Hindi-medium student thinks in Hindi — forcing English at the reasoning step would measure their English, not their thinking.
- Voice-first input. The reasoning step defaults to "Tap to speak." Typing a full explanation is a barrier for this user; speaking isn't. "Type instead" stays as the fallback, not the default.
- One question a day. "5 minutes · one question." The scarcity builds a habit and keeps the AI cost per user near-zero — a real constraint for a free product aimed at low-income users.
- A "Listen" button on every prompt. Text-to-speech so a weak reader isn't blocked from a thinking exercise by a reading exercise.
04 What I built it on
Built solo, GUI-only deployment — same shipping stack behind the rest of my products.
05 The buildathon, and what came after
Socho was my solo entry to the HelloPM Cohort 49 Buildathon — pitch deck plus a working product with real user feedback, which I ran through ex-students now in college across India. It was picked in the Top 6, presented at the final, and won 3rd place — listed on HelloPM's public Build Index alongside 47 submissions and 59 builders.
06 What building this taught me
- The mechanic is the moat, not the content. Anyone can generate questions. The defensible thing is a loop that reliably gets a 12-year-old to catch their own mistake — and the judgment to design the question around a real misconception.
- Language and modality are product decisions. Bilingual-by-toggle and voice-first weren't features bolted on; they were the difference between a tool this user can actually reason in and one that just tests their English.
- Ship the wedge, not the vision. The long-term idea is a "complete school." What won 3rd place was one question, three subjects, done well — with real students using it.
See it live
Socho is an ongoing personal build; the daily question, subjects, and reasoning loop are expanded over time.