I dropped out of my PhD.

When I decided to take leave from my PhD, my dad was obviously not pleased. My friends, however, were not surprised. They would have put up a betting pool on it, except no one wanted to take the other side. As far as choices go, however, dropping out my PhD to pursue a start-up is hardly the maverick move it once was. At least, that’s what I tried to tell my dad during Thanksgiving dinner. You don’t need a PhD to go where I want to go. No, people don’t see it as a failure. Yes, in theory, I could go back. No, I’m not planning on it. Yes, I know you gained a lot from yours, but my circumstances are different…

In some ways, he’s different from other Asian immigrant parents. Rather than focusing purely on career and stability, he’s always instilled in my brother and me a deep sense of curiosity: There’s value in learning for learning’s sake and an honor in advancing the frontier of human understanding. There’s a purity here that to this day I admire. I suppose that’s how you end up with one son studying art in college and another philosophy despite being in all other respects, a “Tiger Parent”.

In a lot of ways, however, not dropping out would have been the real betrayal of those values.

Anyone who has spent any amount of time in academia will sense a familiar story: you excitedly show up to campus on the first day of grad school, wide eyed, full of optimism, and ready to make a contribution that will change science forever. Then you realize making advancements is harder than you thought. No duh. You’re trying to discover something the smartest people in the world have not been able to yet. No matter. You roll up your sleeves. You get to work. You can be patient, you tell yourself. Darwin took over a decade to discover evolution after all. But then you hear horror stories of your friend’s advisor (or if unlucky enough, live through your own). And you realize how fragile the path you have chosen is. Someone who is not incentivized or trained to be a good boss can make or break not just 5 years of work, but your entire career. The phrase: people don’t quit jobs, they quit bosses comes to mind. And the experiments are still not working out as you’ve hoped. At this phase, you’ve toned down your ambition. Hopefully, you can just push out this one paper and then you can get out of there. And as you look to the future, you see the politics involved in academia, the horse-trading for positions, and the small chance you’ll even make it to the oh so coveted tenure position. Is this really worth it? You ask yourself. If I’m going to be playing politics anyway, I might as well go to industry and play it to make money. And if you find yourself in a PhD topic like computer science, then it can really be a lot of money.

Of course, my PhD was nothing like that. I had a wonderfully supportive advisor who humored my wide range of interests, related to my core research agenda or not. The PhD was a place where I was given space to explore and grow, not just as a researcher but as a person.

The trouble is, there was something missing. On a personal level, there were the twin issues of tribe and growth. A doctorate degree is often a very lonely endeavor. Sure there are collaborators, but ultimately, your dissertation is your own. I discovered that I work best and find most enjoyment in a tribe, a team that all collectively strove toward a common purpose. Numerous 2AMs wondering why my code didn’t work while sitting in a cold, empty office showed me what I was missing. Late nights aren’t bad if you have someone to share it with. The second is growth. Research is a very deliberate, methodical process. And by all respects, it should be. If the goal is truth and understanding, then it cannot be rushed. History has also shown us how easily we can be deceived, so rigor is a must. But this creates institutions and ecosystems that at times feel mind numbingly slow. On the other hand, there was what every single of my friends who had gone down the start-up journey had told me: It will be hard. You will never feel like you’re moving fast enough. It will suck a lot of the time. But you will learn more than you ever thought you could: about yourself, about making decisions, about your topic. Call it a flaw, but rather than repulsed, I found myself deeply attracted to start-ups as a result.

Even if all of that didn’t matter to me, it also became clear that the best place to advance human understanding in my field of Artificial Intelligence was not at a university. In AI research, there’s a common belief that the real work is done at the frontier labs of OpenAI, Anthropic, DeepMind and the like rather than at universities. This is partially compute resource related. Even Stanford, one of the best universities in the world for AI, has only a fraction of the compute of even modest size research-oriented start-ups. But it goes further than this. The pace of innovation in AI reasoning and agentic capabilities is driven primarily by industry, especially by a small group of start-ups, indie hackers, and frontier labs. The reason for this is structural and almost ironic. Rather than requiring hundreds of millions of dollars in compute to train a new model, innovations in AI agents require only a clever idea. Moreover, vibe coding means that idea can be implemented, tested, and distributed to the community via Github or Twitter in mere days (see here for an example we curated). This is a far cry from the quarterly cadence of AI conferences, which in itself is extremely fast for academia. The scientific rigor isn’t fully there, but you don’t need double blind RCTs when the rate of improvement is this fast.

Thus at both ends of the spectrum, the action isn’t happening at universities but out in industry where opportunity, capital, and momentum have created a huge wave of research and discovery to surf on. And like surfing, timing is everything. That’s when I knew I had to start now, not in 6 months or 1 year. And if AI has not discovered all there is to know when the wave crashes, there will be time for the patient methodological approach.

Then there’s the question of money. I’m not going to sit here and tell you that the thought of potentially being enormously wealthy doesn’t appeal to me a little. But honestly, doing a start-up for money is a fool’s errand. The pay at a start-up is certainly nothing to call home about. It’s not a PhD stipend, but it’s definitely less than other friends who have “sold out”. There is no guaranteed chance of success, no matter how confident and talented I believe my team and myself to be. I’m reminded of the infamous job ad from Ernest Shackleton, recruiting men for a sea voyage that ultimately became stuck in the Antarctic ice for 3 years:

In a stunning example of leadership, he managed to bring every single crew member safely home. I can only hope in our own journey that I am half as good a leader as he was. If any of this piques your interest, please reach out to sam@ferganalabs.com. We are hiring!