Learning & Teaching

My whole story is a growth-mindset one, and I'd rather tell you the embarrassing version. I was the "smart kid" who avoided studying to protect the label, then the "dumb kid" at Carnegie Mellon who optimized for barely-B grades — and both quietly capped how much I actually grew. What I learn things for now is the opposite of that: not to look capable, but to find out what I'm actually capable of.

Both labels hurt — the smart kid and the dumb kid

I grew up in Bangalore being told I was smart, and that turned out to be a trap. If I studied hard and still didn't ace something, I'd lose the only title I had — so the safer move was to not really try. I've written the ugly logic of it down plainly:

“Being called smart all my childhood had made it part of my identity and I was afraid to lose that title more than I was afraid to fail a class.”— About Me

At CMU the label flipped. Surrounded by people who were genuinely sharp, I became "the dumb kid," and I optimized for that just as hard — except now the optimization target was minimum effort that still passes. I got eerily consistent at it.

“countless 81.5 percentages, proof of my optimization for minimal effort in every class… Unfortunately, this strategy also optimized for minimal learning.”— About Me

Two opposite labels, same outcome: I protected an identity instead of growing. The wound underneath every course I'm building now is that one — and it's why I'm allergic to teaching that's really about grades.

The 2.5-to-4.0 turn — and the honest relapse

The turning point wasn't a motivational montage; it was sleep. A bad health-and-sleep spiral bottomed me out around a 2.5 GPA, near an expulsion warning. The next semester I did one boring thing on purpose — I actually slept, and I tried to learn the material instead of game the grade — and I came out with a 4.0, Dean's List, in my hardest term. The honest part is that I genuinely didn't see it coming about myself. I'd spent years believing the ceiling was the ceiling. It wasn't; the mindset was.

I want to be careful not to dress this up as a clean arc, because it wasn't. I relapsed. A later semester with a single course ended in a 52.5%, a strongly-curved D — and the line I keep about it is the truest thing I can say:

I didn't even know the contents of the course even after having completed it.

That's the counterweight I'd rather hand you myself than have you find. The growth mindset isn't a personality I unlocked once. It's a thing I have to choose, lose, and choose again. Carol Dweck's Mindset is the book that gave me the language for it, and the version I tell people is blunt:

“Wherever you believe in growth, you improve. Wherever you don't, you stay stuck.”— Day 9: Fixed vs Growth Mindset

The method: 30-day sprints, and the feedback loop underneath them

People hear "30-day sprint" and assume the magic is willpower. It isn't. The sprint is just a container. The actual engine is a four-step loop I run on every skill, and it's the same loop the Maker runs on every app:

• Decompose the skill into components.
• Find the real bottleneck — not the part that feels hard, the part that unlocks the most if solved.
• Build a feedback loop so I can actually see whether what I'm doing is working.
• Iterate fast, discard without ego.

“That's what I'm good at. Not being smart — being fast at finding the right thing to work on.”— How I became superhuman in 2 years of being unemployed

The single most me example of this is also the least glamorous. I'd plateaued losing weight, so I treated it like any other bug — I used my computer science degree to debug it. The insight was a signal-to-noise one: daily fat loss is around 0.3 lb, buried under ±2 lb of water-and-food swing, so the raw weigh-in is almost pure static — like trying to hear someone whisper in a thunderstorm. So I stopped reading the noisy number. I drew a target line dropping 0.3 lb a day and tracked the delta between it and reality, where the errors accumulate and the signal climbs out fast — turning every morning into a clear instruction instead of a coin flip, and training your intuition over time. That became Weave — my mom lost 17 pounds with it, I lost 16. The feedback loop is the obsession; the app was just where it landed.

There's a version of this I teach as "three things that make hard work actually pay off," and the order matters: first research the best way to learn before you start (I'll spend days on YouTube, articles, and ChatGPT figuring out where the common advice is wrong); then set up a feedback system, because without it you can't tell whether you're improving or quietly reinforcing a bad habit; then spread the practice across days, because sleep is what solidifies muscle memory and knowledge. Only after all three does the cliché become true.

“consistency is the only differentiator between the best and the mediocre.”— Day 14: Why Hard Work is not enough

The part I want to be loud about, though, is that I'm not a grind guy. I think 16-hour days are counterproductive — I've done them, and they cost me more output across the week than they bought me on the day. My actual bottleneck has never been stopping; it's starting. So most of what I optimize is the on-ramp, not the willpower.

And I'm honest about where the method breaks, because it breaks the same way every time. I ran a "post a video every day for 30 days" challenge, rode a real high through week two — "the longest habit streak I'd ever been on" — and then a couple of bad nights of sleep collapsed the whole shaky foundation. I failed the literal goal. I still got the growth I started it for. The fluke and the win are the same story; I just refuse to only tell the win.

The learning science I actually built

Somewhere along the way the method turned into a small homemade theory of learning, and it's the part I genuinely love — distinct from academic research, which I don't. (At CMU I did research at Norilla, building smart toys to help kids learn faster; the science of how people learn has been a long-running soft spot.) A few load-bearing pieces:

• Machine learning, run in reverse. ML models were designed after human brains, so you can hack your own learning with the same techniques — clean inputs train fast, noisy inputs (sloppy, inconsistent practice) train slow, and once a model's well-trained it shrugs off noise. "More data overcomes noise" is just "practice volume beats fancy gear."
• Rubber-duck self-explanation. Explaining a problem out loud forces you to linearize a messy thought into a sequence — and the gaps light up the moment you do. (I'll cheerfully tell you this is also my biggest social flaw: the thing that makes me learn fast is the same thing that makes me monologue at people. Same mechanism, two reviews.)
• The struggle is the learning; the answer barely matters. Lectures are mildly interesting YouTube videos. The actual learning happened in the problem sets I spent days stuck on — sometimes a single problem, sometimes without even solving it.
• Information isn't skill. Asking an AI how to build an app and feeling like you learned something is a trap.

“It's like trying to learn soccer by reading about soccer and watching soccer — if you don't spend any time fumbling around in the field… you WILL NOT learn those skills.”— How to become a cracked software engineer

There's one belief these all hang off of, and it's the axiom that justifies my whole scattered-looking life. Most people pick one thing and grind it for a decade. I think that's often the worse bet:

“The Pareto principle applies to learning too. In just two years, you can learn 80% of a skill. Instead of spending 10 years perfecting one thing, you could be 80% good at FIVE. … If you want to be truly unstoppable, learn everything.”— Day 11: The Real Cheat Code to Growth

Everything interconnects. Singing made me a better speaker; games trained me to perform under pressure; debugging weight loss taught me product design. None of it is a detour. It all feeds the next thing.

Languages, ranked

If you want to watch the method work on a real skill, watch it on a language. I taught myself Japanese to A1 in about twenty focused hours of a frequency-ordered Anki deck — roughly 1,000 words and most of basic grammar, spread over about forty non-consecutive days. (The headline number people repeat is "three months in Japan." That's wrong. The three months were the immersion that cemented it; the learning was the twenty hours.) The why is pure me: I'm lazy in the good way — I'll spend ten hours finding the path that saves a hundred — and I refused to do this the slow way after a humbling discovery.

“I watched anime for ten years and learned five words. Konnichiwa. Arigatou. Nani. And honestly, ‘nani’ only because of the meme.”— Flua Hero Video Script

The emotional peak of the whole thing came about two weeks in, watching an episode I wasn't even studying:

“I paused the episode and walked a lap around my room. Ten years of watching anime and the language had been a locked door the entire time — and it just, opened.”— Flua Hero Video Script

The thesis that fell out of that is the seed of Flua: the first thousand words — if they're the right thousand, in the right order — cover most of everyday speech, and immersion-alone is a myth because the flashcards are the bootstrap that gets you to the threshold where anime stops being noise and becomes a free, unlimited tutor you were going to watch anyway. (My one genuinely original wrinkle: real conversation is a terrible way to drill recall, so you can just flip the comprehension cards backward and practice production efficiently — simple enough that it's easy to overlook how much it does.) I deliberately, philosophically skipped reading and writing Japanese entirely, which I still think is the single biggest shortcut everyone's too scared to take.

The rest of the roster is honest about what each language is for:

• English — native; it's the language I create and joke in. Hindi sits invisibly underneath it as a baseline (it's how I argue, only half-joking, that English is objectively harder to hear than most languages — loose phonetics, spelling that doesn't match sound, homophones everywhere).
• French — the active, ongoing one, and frankly an immigration lever: it's the rate-limiting step for moving to Canada (TEF/TCF, around B2). I lived in Lyon for a couple of months specifically to immerse, rebuilt my own i+1 French deck more times than I'd like to admit, and went to conversation meetups. Same playbook as Japan.
• Marathi — which I speak fully, not fluently — "I understand all the words and grammar; that doesn't mean they come immediately to mind." The origin is my favorite: I learned it by accident, speaking deliberately broken Marathi to crack my mom up on video calls from New York, until the joke turned into the skill.
• Spanish — genuine travel scraps I actually reach for with strangers, and the next real target.

The courses — built for real learning, not exams

The teaching grew straight out of the wound. I'm building courses from the ground up — Mathematical, Statistical, and Computational Thinking — for people who've already graduated and want to revisit the fundamentals and actually understand them this time. The mission, not the price tag, is the point: a true learning environment, unlike the way the top schools (including my own) taught me.

That sounds like a cheap shot at school, so let me make it precise, because it's autobiographical, not posturing. My critique is specific: education optimizes for grades over learning, treats lectures as the product when the problem sets are where learning actually happens, and forces one pace on everyone. My fix is a different shape entirely:

“It's like going to a gym, where each individual gets the most out of it from doing weights suited to them rather than the average weight in a class by giving everyone the same problems and pace.”— Is Japan Overrated? My experience

The pedagogy under it is the same ordered-difficulty idea that makes a good video game teach you without a tutorial: pick a handful of genuinely fundamental skills per course, build them up one rep past your edge, and wrap the math in something that doesn't feel like math (one session is literally a "Probability Dungeons" game that's secretly an expected-value lesson). I want students to track a "toolbelt" as they go — thinking frameworks, and the sub-tools inside each — so a vague, paralyzing problem becomes "which tool do I reach for," and so progress is visible instead of vibes. And the real-life examples I want to teach with are very on-brand: the math of weight loss, why a heavier cricketer is slower (intuition for estimation), how much money you can actually make from YouTube as a lower bound, and — only half a joke — getting deported in a week: should you take or leave all your belongings? The honest framing is that I didn't build these to flex; I built them because I went looking for a resource that taught these fundamentals concisely and there genuinely wasn't one.

There's also a free-resource gap I keep pointing at: MIT's OpenCourseWare is great, but it's just the lectures, and that's not where MIT grads learn 95% of what they learn — that's the problem sets. So the courses put the core idea up front and the build-up — the problem set, the place the learning actually happens — behind the work. And there's an origin story I don't hide, because it's the growth-mindset thesis applied to me:

“I focused on what I could control… I was never truly upset for even a day… In the process, my passion for teaching has only grown stronger.”— the cancelled-launch message, Core Life Courses

The first paid class was scheduled for the literal day a visa miscommunication forced me out of the US. I left behind most of what I owned. Instead of letting that kill the project, the setback became the curriculum — it's now also about embracing challenges and building a community of people committed to growth. That, more than any module, is what I'm actually teaching.

• Three "Thinking" courses scoped — Mathematical, Statistical, Computational — each broken into 3–10 fundamental skills.
• Real, named consulting and teaching, not vaporware — including a ~10,000-word personalized growth-mindset essay I built for one real person, diagnosed from a questionnaire.
• A fully drafted, hour-long TEDx talk on the three pillars I keep coming back to: Sleep, Health, and Knowledge — with a running mission of fighting health-and-learning misinformation.

I'll admit the part that surprises people: I like learning a little more than I like teaching. Teaching live genuinely wires me up — my heart rate's still elevated an hour after a good call. But the reason I want to teach at all is the same reason any of this is public: connection. The dream isn't a course funnel, it's a community of people who like learning weird things and want to do them together.

What I want to build next

The honest bottleneck, the same one across everything I make, is that insight outruns my follow-through — I'm far better at starting than finishing, and the fix isn't more discipline, it's the right loop and the right people. If you like taking a skill apart to find where the common advice is wrong, or you want to learn something strange alongside me, say hi. The method's more fun with company — and a lot of the time, the worlds I build start the same way a lesson does.