What If You Could Carry Your Classroom With You?
Every student of economics who sticks with the subject long enough meets a book called Microeconomic Analysis by Hal Varian. Well, either Varian or Mas‑Colell. But one of the two, almost always.
Here’s a controversial take: there isn’t much to be learnt by reading those books—at least not the way they’re written. They belong to the khoda pahad, nikla chooha school: there’s towering mountains of notation and big words, and there’s teensy bits of intuition hidden deep within . Dave Barry once parodied this style for sociology; this is the microeconomics version. And because it’s econ, it comes with equations, symbols and diagrams.
Of course.
Learning Can Be Hard, And Sometimes Unnecessarily So
Learning can be hard. Sometimes productively hard; sometimes needlessly so. We’ve tried to help students learn by using in-person lectures, radio, television, the internet—and now AI.
And yet, as Arnold Kling observes:
We have been promised dramatic improvements in education ever since the appearance of television. So far, spending on education continues on a rising curve, and results continue to disappoint.
Why? Because while inputs (content) and outputs (applications) have exploded, the process of learning hasn’t kept pace.
Inputs
When I was in college, we used to queue up at 7:45 pm in the DR Gadgil library for “overnight borrowing.” We only had a few precious copies of papers or textbook chapters that could be issued, and that only until the a.m. Scarcity made us efficient learners, especially on the eve of exams.
Today, there is no scarcity. JSTOR links, PDFs, YouTube, podcasts, and now AI remixes means that you are never going to run out of material. And I mean that in a very literal sense!
Outputs
Our institutions have produced ideas, tools, and products at scale. Those outputs are what we end up studying in colleges and universities. Take an example from statistics. We produced some pretty awesome statistical methods in our educational institutions over the past 100 years, and students in colleges today spend time learning about these methods. In this sense, it is possible to think of education as closing the loop, and becoming a virtuous cycle. We use inputs in education to produce better learning, and therefore we make better things, and then we use and study these better things, to produce better learning.
But ah, that brings us to the point.
It’s the Process of Learning, Stupid (the bottleneck)
What still lags is how we learn. How to keep quality, engagement, and effort high—at scale, and for each learner? Baumol’s cost disease, all-too-stubborn inertia, and well, a shortage of imagination ensure that Arnold Kling’s line stays stubbornly true.
Why? Because we are a social species
To “solve” education you need four things: quality, scale, customization, and a social setting. AI is getting us close on the first three. The fourth is where online learning keeps failing.
Why? Because when you move learning online, it becomes a lonely experience. And that’s not just bad from a psychological perspective – it also impacts learning outcomes.
Learning happens when a professor explains, sure. But it also happens when a batch‑mate asks a question you hadn’t thought to ask. It happens when you peek at a neighbour’s attempt to solve a problem to get a sense of how you might start. It happens when you DM someone at 7 pm about a homework problem. When friends argue over a definition. When you teach someone else and realise what you didn’t know in the process.
Put one person alone in front of a screen and some of that is inevitably lost.
Can software come close?
Here’s an idea: what if you could carry a classroom in your pocket?
Picture a virtual facilitator who walks you through new material, flanked by virtual classmates who listen, poke, and, every now and then, make a mistake on purpose (and we hope you’ll not just pick up on the mistake, but also correct it!)
Spaced repetition will be built-in. These students might ping you a few hours later to chat about how they’re trying to solve the assigned homework. You, in turn, can ask them about your homework. They won’t solve it for you, but they will give you some pointers. They might nudge you to teach an idea “in your style”. You deliver a quick little masterclass, and the facilitator weighs in immediately with feedback.
These aren’t faceless bots. You get to choose who sits beside you! Maybe you want a budding lawyer when you learn about the Coase theorem, and maybe you want a marketing MBA aspirant when the prof is talking about Richard Thaler’s work. A class on research methodology would be best studied with a would‑be academic, and you want to talk about the Solow model with a budding development professional. You need the maths whiz next to you when you’re tackling Hal Varian and maybe that guy who cracks a dozen jokes a minute, just for the laughs.
But it gets better: these personas stay with you across your whole journey, remembering last year’s PDFs, the podcast you listened to this January, and last night’s spreadsheet. They don’t just remember it, they bring it up when it matters across different things that you’ve studied.
Think NotebookLM’s artefacts—briefing docs, FAQs, mind‑maps, podcasts—plus a social layer that learns with you. And this is worth repeating: the avatars adapt across subjects. Ideas from microeconomics resurface when you study industrial organisation, then information economics. The way questions are asked, the homework you see, the reminders you get—all reflect your particular learning path.
Will it perfectly replace the best offline classroom? No.
Will it beat studying alone online? We believe so.
And we’d like to build it.
We’re just going to do things
This isn’t just a thought experiment. We believe this gap between lonely online learning and effective social learning is one of the most important problems to solve. And we think we have a way to do it.
I care deeply about helping people learn better. Ansh Masand cares deeply about building better software. Together, we’re building a virtual, RAG‑based classroom you can carry everywhere.
Formal courses, hobbies, workplace PDFs—learn all of it with the same facilitator and classmates, over time.
We’ll start text‑only and voice is next. Video if and when it helps (and when the tech gets good enough). We’ll add projects too: real‑world assignments that you can complete solo or with your virtual batch‑mates—because the world increasingly hires on the basis of what you can ship in an AI‑assisted workflow.
We need early testers
To make this useful, we need feedback from an early cohort.
Early‑access fee: ₹500 (to cover development costs)
You get: ongoing access to updates, a say in the roadmap, and a private Discord
We ask: honest feedback—what worked, what didn’t, and why
Interested? Leave your email here and we’ll send you a short form to understand your interests and use‑cases, and then your invite.
If you’d rather not sign up but do want to help, please, tell us what you think. What do you think? What are we missing? What would you love to see? Do leave a comment below, or drop me an email.
Thank you, as always, for reading – and we hope you’ll join us on this journey!