Failure as Teacher

Let’s talk about failure. No one likes to fail. When you’re a kid and fail a class, the translation is immediate: you are a failure. Parents yell, teachers frown, classmates whisper. The carryover effect is brutal, once branded, it bleeds into everything. You start equating low grades with low worth. And because it’s uncomfortable, everyone learns to fear failure instead of learning to use it.

At thirteen I entered Westminster School, one of the UK’s most elite high schools, where half the students march off to Oxford or Cambridge every year. The culture was suffocating: more kids had gold medals from math olympiads than could run a six-minute mile. From the start, I was the odd one. I failed classes, fell behind in rankings, stumbled again and again. For the next 5 years, failure walked with me.

Around sixteen, something broke free. I stopped caring about grades. The suffocation turned into oxygen. Instead of chasing A’s, I chased what fascinated me: ferrofluids, acoustic levitation, building speakers, fencing and so many other little things. I still remember timing how long ferrofluid spikes took to relax after a field was cut. None of that showed up on transcripts, but all of it mattered to me.

i thought “holy sht that worked?” when the first particle was suspended magically in air through. No string. Just physics of sound waves.

School Is Not Life

Paul Graham’s essays gave language to what I had already lived: “School is not life.” At Westminster I always felt behind, yet when I arrived at college in the States I realized I was ahead—already fluent in organic chemistry, physics, advanced math. School had force-fed me knowledge, but never gave me permission to fail. Every day was about metrics and rankings. Failure wasn’t allowed. But failure is the real curriculum.

Failing taught me the one skill school works hard to erase: how to fail productively. My classmates optimized for quartiles; I learned to iterate. My peers feared falling behind; I built my tolerance for rejection. with start ups you fail a lot. So having 5 years of failing really helped me out. I was not scared to hear no or scared to fail. In fact I learnt to love it because it meant that I could improve.

From Exams to First Principles

College made the divide sharper. One hundred multiple-choice questions, memorize the textbook, curve the grade. I never cared about GPA. I passed exams by deduction. In bio, I derived pathways from electron theory instead of rote memorization. That was hacking or solving from first principles, not memorizing scripts. Graham would call it “elegance as escape.”

Meanwhile, startups were the extreme version. School taught the four Ps of marketing; the real world asked me to get 100 users by next Friday. That’s not on a multiple-choice exam. That’s scar tissue.

Over four years I stacked 80-hour weeks on average: 80 hours per week x 52 weeks in a year x 4 years =

16,640 hours of building, pitching, scraping, breaking, failing forward. That number is the gap between me and first-time founders. Not GPA, not test scores, just raw experience.

Notebook of heresy

I keep a notebook of unconventional startup questions:

“what if X were possible, or if Y happened, could we apply that to Z?”

I try to falsify one of those hypotheses every month. That habit forces fast prototyping and quick kills: build the heresy, then let the data decide.

A torrent breaks a large file into many small pieces and distributes those pieces across peers so everyone both downloads and uploads at the same time. Clients request rare pieces first, neighbors re-share what they have, and the network reduces dependence on any single seeder; the result is resilient, bandwidth-efficient distribution of huge files.

You can apply the same idea to large language models. Projects like Petals.dev shard a model’s transformer blocks across volunteer GPUs, assemble a pipeline of those blocks at query time, and stream activations through the chain so a user can run or fine-tune very large models without a single monster GPU. That makes LLMs more accessible at the cost of higher latency and handling peer churn, but it is practical enough today to experiment with and to power interactive apps. (petals.dev)

“Can i torrent an llm?” - that question appeared 2 months before i discovered petals.dev .

So what is a hacker?

Levy’s hackers would’ve recognized the sixteen‑year‑old messing with levitation rigs and ferrofluids for joy, not credit. Graham’s essays describe the next phase: break rules when they fossilize, keep a ledger of unsayable ideas, treat ugly hacks and brilliant hacks as cousins, build systems that ship. Together they map a life.

So what is a hacker, to me?

• Someone who metabolizes ambiguity into artifacts.
• Who learns by doing—and by deleting.
• Who breaks rules to escape local maxima, not to posture.
• Who distrusts authority yet serves users.
• Who can derive the pathway when the textbook is missing—or wrong.
• Who treats failure as a data channel, not a moral verdict.

I want both: Levy’s soul and Graham’s operating system.

The Real Curriculum

The truth is I’ve failed more classes than most of my peers. But I’ve also built more things in the real world. School won’t teach you to cold-start a product to 1000 users; I’ve done it. School won’t teach you to survive repeated rejection; I’ve endured thousands of hours of it. School won’t teach you to improvise when the schema is missing; I’ve lived that as default.

So yes, I failed school many times. But I learned what school couldn’t teach: resilience, curiosity, first-principles problem solving, scar tissue that compounds into wisdom. If Levy’s hackers fought for access and Graham’s hackers fight for meaning, I’m both. I failed in school, but I found my life in failure.

I’m a Levy-esque tinkerer who fell in love with making real things; I’m a Graham-style builder who refuses vibes in favor of clarity. The final lesson from both: ship truth.

In code, in markets, in how you live. That’s the only scoreboard that survives contact with the world.