The Delusional Mind
The one sentence I actually live by
Overconfidence is always better than underconfidence, as long as you strive to make sure that overconfidence becomes the new baseline.
Read that second half again, because it is the whole thing. Not "be overconfident." Anyone can walk around loud and empty. The trick is that you start overconfident on purpose, and then you go and do the work that makes the overconfidence true, until the level you were faking quietly becomes the floor you stand on. Then you fake the next level. Repeat.
Underconfidence does the opposite. It waits for permission. It grinds the same thing everyone else grinds so that, at best, it ends up exactly as good as everyone else. Underconfidence is a very polite way of guaranteeing you stay average.
My sister and brother-in-law spent years telling me to be more like everyone else. "Do LeetCode. Apply like a normal person. Stop being a fool." They loved me and they were describing the safe path. I kept choosing the delusional one.
This is the story of the delusional one working 8 times.
The kid who was buying the whole company
There is a story I read once that I have never been able to shake.
A millionaire asks a sharp, broke kid straight out of high school how long it will take him to buy a BMW. The kid thinks and says, maybe 20 or 30 years. The millionaire is stunned. 30 years? You will have a job in 3 or 4, you could buy one within another 5. Why so long?
And the kid says: you never said "a BMW car." So I assumed you meant the company.
It sounds like a joke. It gives me goosebumps every time. Because that gap right there, between "buy a car" and "buy the company," is the entire difference between the 2 mindsets. Same words, same facts. One brain defaults to the small reading, the other defaults to the enormous one and then spends its life trying to close the distance.
Overconfidence is just choosing the enormous reading by default.
I default to the enormous reading. It has cost me, and I would not trade it. Here is the receipt.
Where the delusion started (India, flukes 1 and 2)
The delusion started early. At 17 I decided I was going to start a company, and at 19 I actually did, with a few friends and a completely straight face, genuinely believing we would one day hire Zuckerberg to come work for us. We called it SayItSocial Pvt. Ltd. But that is a story for another day.
Fluke #1 was FairShare. In my 2nd year of my undergrad degree, I cold reached to the CEO of an IT startup Aniruddha Pimpalkhare, with a pitch: to help him build an end-to-end e-commerce app for the street vendors selling vegetables and fruits. He agreed in the first meeting. I interned there for 4 months and shipped it. Nobody asked me to invert a binary tree. Perhaps I got this internship because I got lucky?
Fluke #2 was StoneX, and it is the one worth telling in full.
On-campus hiring in India works differently than in the US. Companies come to campus for a day or more, run interview rounds after rounds back to back, and hire on the spot. StoneX was one of them. The screening round was LeetCode, plus MySQL, plus an aptitude and IQ section. I left the entire LeetCode section blank on principle - I only did the 2 parts I actually believed measured something - and I passed anyway.
It was 7pm and I was already halfway home when a friend called and told me I had cleared into the next rounds. I was not interested. She made me come back to campus ASAP.
Of 500+ students, 30 were left. 8 would be hired. I was number 30 in the line, about to do the first in-person interview of my entire life.
So I did the only rational thing: I interrogated all 29 people ahead of me as they walked out. One got a brutal graph problem. One got dynamic programming. One got asked networking and hardware questions for a software role. One got grilled on operating-system internals. And there I was, completely unprepared, because I had never actually wanted this. The only thing I had was this: I was insanely confident in React and blockchain. I was the president of the Blockchain Club at Symbiosis, and at that moment I believed I was a better React dev than anyone in that hallway.
My turn. The staff engineer tells me to sit. I stay standing, and I say something like:
Sir, with respect, let me save us both the hour. I am bad at LeetCode and mugging up those patterns. I know how to build stuff instead. I am an amazing React and blockchain dev. If that's what you're hiring, I'd like to sit and continue. If it isn't, I would not wanna waste your time.
He laughs and says, please, sit down.
For the next hour - not the scheduled 35 minutes, the full hour - it is React design problems, live coding, and blockchain concepts. And I am killing it.
His last question was a trap. He offers me a choice: 1 final hard question, or 10 easy ones where a single wrong answer ends the interview. I was exhausted and I wanted to go home quickly, so I said hit me with the hard one. He leans back and asks: "Chicken or egg. Which came first?"
I say "Chicken." Instantly. No thinking.
He says, "Can you explain?"
I laugh hard and say, "You said 1 hard question, right?"
He laughed so hard he sent me straight to the final round.
The final round was with Prashant Sangoi, who runs StoneX's entire Pune operation. He said he'd heard my interview went nothing like the others. Then he asked me exactly one question: walk me through your whole life and how you made your decisions. And so I did. He hired me on the spot. His words were roughly, "I still have to decide on the other 7 hires. But I want you." I accepted the offer right there.
StoneX is the only big company I have ever properly worked at. I interned there for 7 months. I was supposed to join full-time, but instead I left for my master's in the US. Working at StoneX showed me how companies operate on a huge scale and helped me realise: big companies move slow, and I will choose working at a startup any day of the week.
Then the doubt showed up (USA, flukes 3 through 8)
I landed in the US and everyone around me was grinding LeetCode for interviews. Every single person. And I was not.
So the doubt crept in, the exact doubt my sister and brother-in-law had been warning me about. What if India was a fluke? What if FairShare and StoneX only worked because of some local magic that does not exist here? What if they were right and I was the idiot?
There was only one way to find out. Run the experiment again. And again. And again.
Fluke #3 - Spashta AI, June 2024. My first US internship. I cold-outreached Prashant Serai, ex-Meta, ex-Google, a PhD in voice tech, who was building SpashtaAI on noise reduction and accent and language conversion. 2-3 calls with him. Zero LeetCode. I got in. 2 months later he wound the startup down. My first real swing in a brand new country, and it connected. This internship I received must surely be Luck, obviously. I was grinning like an idiot.
Fluke #4 - Dextego, September 2024. I cold-emailed Ioanna Mantzouridou Onasi, the CEO, a full slide deck on what Dextego could be doing better. She was impressed enough to hop on a call a couple times and finally offered me an internship. But we both couldn't agree on the logistics of the offer so I couldn't work there. Didn't matter, the swing connected again. Probably still just luck, right? Sure. Let's keep testing that theory.
Fluke #5 - SpatiumLab, April 2025. This one is my favorite. I had started learning Turkish, purely because I was obsessed with a Turkish drama called Bambaşka Biri. Then I came across Cem Ozen, a Turkish CEO in Bellevue running SpatiumLab. I cold-reached him (in English, obviously haha) and got an interview. I spent the first 2 minutes introducing myself entirely in Turkish. He cracked up. Then came an in-person React and FastAPI hands on coding round, which I aced. 3-month internship. I got a job by opening in a language I was learning off a movie. Please tell me that is luck. I could not stop laughing.
Fluke #6 - Abby, July to November 2025. I had just built an AI Therapist at the UWB Hacks AI hackathon (my 2nd time competing) and lost. Then I found Abby AI, an AI therapy app, used it, and immediately saw a dozen things I would change. So I cold-outbounded the CEO, Julian Sarokin, 3 times with a deck. He said yes. I started with a 1-month internship that converted into a full-time offer lined up exactly with my graduation. I was there 5 months. And now I was enjoying of these "flukes" in Hollywood LA. PS: I helped 5 friends get internship/jobs at Abby intelligence Inc. LMAO what a Fluke!!!
Fluke #7 - Alex AI. Around this time, the founding-engineer role I had lost a year earlier to a visa wall came back around, because I never stopped following up. That whole saga - the lost offer, the 17 emails, the day John finally replied - is its own story, and I already told it in The Wrong Competition. Short version: it worked, and I moved to SF. My so-called flukes were paying rent in San Francisco now and paying me more than my friends at Amazon were getting paid!
Fluke #8
I left Alex in May 2026.
My sister and brother-in-law, right on cue: "Okay. Now will you just do LeetCode and apply like everyone else?"
I laughed and bet on my stupid delusional mind an 8th time.
I reached out to a UW alum named Jessica who connected me to Mek, then the CTO of Stuut AI. 5 in-person rounds from system design to decision making to whatnot. Zero LeetCode. Hired in 5 days. I am there right now, owning the voice-agents work, and happier than I have been anywhere.
8 for 8.
What the tally actually says
From FairShare to Stuut, that is 8 - 1 big company and a long string of startups. And I have still never passed a LeetCode-style interview, because I almost never had to sit one.
I am not smug about it and I am not telling you to skip the grind. If LeetCode is the door to where you want to go, go grind it, and I will cheer for you. The point was never that the door is fake.
Here is what I actually mean by "fluke." Every time one of these worked, a voice - sometimes my family's, mostly my own - said the same word: lucky. A one-off. Don't get cocky. And I understand why. A fluke is a single event, so calling it that let me enjoy the win without having to defend the belief underneath it. It kept me humble and it kept me small in the exact same breath. But a fluke that repeats stops being a fluke. By the 3rd or 4th time, "fluke" was just the story my doubt told itself so it would never have to say the scarier thing out loud: this is working, and I am the reason.
And the delusional mind is just the part of me that keeps betting before the proof shows up. I call it delusional because that is exactly how it feels from the inside - deciding you will pull it off while every reasonable person, half my own family included, is telling you to be realistic. I keep it because the realistic version of me would have grinded LeetCode, blended into the crowd, and disappeared. I would have never been proud of myself If I had done that :)
Being delusional is only embarrassing right up until it works. After that, everyone just calls it vision.
Underconfidence would have had me in the same room as everyone else, grinding the same problems, coming out indistinguishable. Overconfidence put me in rooms I had no business walking into - and then, every single time, I did the work to make it so I did belong there. That is the sentence from the top, running live: start above your level, then drag your level up to meet it, until it's just the baseline. Then do it again.
And the honest, slightly petty truth?
There has been no greater joy on this earth than proving my sister and brother-in-law wrong, 8 times and counting. I love them for doubting me. They gave me something to disprove.
I am going to keep betting on the delusional mind. Every single day. Any day.
And if you take one single thing from all of this, take this:
Always believe in yourself. Nobody on this earth will ever know you better than you know yourself, so back that person before the world gives you permission to.