As of May 25, 2023, my time at Harvard is officially over. In this post, I’ll recount my experiences within the Institute for Applied Computational Science (IACS), which is embedded in the School of Engineering and Applied Sciences (SEAS) at Harvard.

I’ll warn you in advance that this story is long, but I learned more than I’ve even verbalized at this point. I won’t sanitize my experience or tell you anything less than what I perceived to be true.

Jon Huml: Origins

I came to Harvard as a dumb 22 year old. I’m not being self-deprecating, I just think 22 year olds are dumb. It’s impossible not to be (somewhere, there’s an older person snickering that 24 is an awfully rich age to be making that comment, but I digress…). I went to undergrad 15 minutes away from where I grew up, and UNC-CH, as a public school, mandates 82% of its class be from North Carolina, so a lot of my peers were also my neighbors with the same perspectives and experiences that I myself had. I hadn’t seen much of the world, or if I did, it was the clean and touristy parts of it. The sorts of places where it’s okay to take kids. And that’s fine: I’m not sure why my parents would show me Skid Row as a kindergartner anyway. I’m glad to be from North Carolina, but my world felt artificially small.

When I first arrived in Boston, the city seemed like a bustling metropolis compared to home back in Chapel Hill. Any place with a metro line must be the height of cosmopolitan luxury, of course. As many of us were at the time, too, I was itching to get outside after being in quarantine for over a year. Any place seemed bigger than my living room. I got here in August 2021 but graduated from UNC in 2020 thinking I was headed to UC Berkeley. My dad and I had plane tickets for San Fran on March 9, 2020, just as the shit hit the fan with COVID. We called an audible, and I was unsure of my immediate plans. With the whole world shut down, it only seemed natural to bum around while I still could.

In that gap year, I spent some time taking random math classes I would never “need” but wanted under my belt, as well as doing some research before I got into Harvard in February 2021. I was and am beyond grateful for that downtime. I became a much stronger mathematician being able to linger on problems, focusing on my process instead of just getting the work done. Math, for me, should be relaxing and about creating a small moment of revelation for yourself. Oh, it’s defined that way for this reason! I loved my math and statistics classes in undergrad, but I often felt like I was drowning with 5 of them at once. When you get busy, you look at StackExchange, and when you look at StackExchange, you get robbed of that discovery. A lack of focus is poison for good science.

In addition, I spent a lot of time on my writing at a mechanical, structural, and emotional level. Strunk and White’s book as well as King’s On Writing were my Bibles. I’ll talk about this more below, but this might’ve been the most important preparation I took during my Year of Nothing. Writing is the foremost superpower. It turns mediocre papers into acceptances, gets you into good grad schools and behind other closed doors, and keeps eyes on your emails/websites/etc. Writing is the act of structured thinking, of seeing your thoughts on page and rearranging them. We lose that muscle at our own risk. In a world where we increasingly see basic functionalities be outsourced to source code, asking software to write for you is like asking it to work out for you. Worst of all, if you suck at writing and you’re a cornball like me, you’ll look back at old poems and letters you wrote to girls and want to hang yourself. Trust me. Some of those “masterpieces” are floating around out there in the world, and if they ever resurface, I’ll have to relocate to the wilderness for good.

Out on the Trail

I felt pretty good about my preparation levels coming into the program. My unequivocal goal, which I knew from the first day I got here, was to get into a top PhD program, preferably at Harvard. Note that in IACS, this isn’t rare but you will be a minority if your goal is a PhD: most people are there for a pay bump, a taste of the Ivy League life, and the Harvard brand stamped on their forehead. If this is you, it helps to be honest about your goals coming into the program. Don’t lie to yourself if you don’t want a PhD. Academics will obviously push you towards the PhD, so be careful around the Kool-Aid jar. There were some folks like me, others on the total opposite end of the spectrum, and some in-between. My bucket was probably the most self-loathing and miserable, the opposites were the happiest and most able to enjoy the networking aspect of Harvard, and the in-betweens wasted their time on dead-end applications for places they weren’t sure they even wanted to do a PhD at. The best PhD applicants I saw already came in with research experience under their belt, and there were others like me with weaker research backgrounds that made up for lost time by immediately getting involved as soon as they arrived in Cambridge and staying over the summers to get work done. Remember that the rec letters (and a publication or two) are the most important aspects of a doctoral application: these can take several years, at minimum, to obtain.

Coming in, I still had to figure out what subfield of deep learning I wanted to specialize in and what extra sorts of skills I’d need. I would realize how slow of a programmer I was when trying to build more complicated neural architectures in a reproducible, scientific fashion in Pytorch, but my math background was a big help in breaking problems down into simple pieces. Most engineering math is quite simple–the jargon coating the problem is the only barrier to its understanding.

One “shocker” I didn’t learn until after my first semester: grade inflation here is wild. I put in work for my 4.0 at Harvard, but I feel that I worked just about as hard, if not harder, for a 3.7 at UNC-CH. Granted, the courses were different (I had math and physics exams at UNC-CH where the averages were in the 20s and 30s) and perhaps I had just figured out the game, but it alerts my Spidey senses nonetheless. The point isn’t about grades. Especially now that I’m in a PhD program, who cares about grades? My point is this: take classes that sound cool, and worry about the outcome later. Of course, you try to build some balance into your schedule to allow for research and personal upkeep, which I’m still working on after 6 years of schooling.

That first semester was hard in a way that I didn’t expect, couldn’t prepare for with all of the new stimuli. I tried to take four classes and conduct research, and it was impossible. I dropped one class to make more room for reading papers, and it still hurt. I probably worked 80-90 hours a week and felt like I was menstruating for all of them. My friend Bill and I had some fantastic study sessions working on our neural computation homework together. However, one “night” in November, I came out of the iLab at 6AM, and for the first time in my life, I felt terrified that I couldn’t turn my brain off. In hindsight, I didn’t even need to do that, I just felt obligated to as if it were a badge of honor. The sun was rising, and I was in the same clothes I was in from the day before. My pursuit of intellect, whatever that was, had begun to destroy my ability to live as a multi-dimensional human being. The tether to Earth had snapped. I hit one of the first low points of my time here, and I resolved to take winter break completely off to reflect on my acclimations to a new city and school. Nothing dumber than being smart.

That January I wrote a list of personal Ten Commandments (I may post them here soon) based on my own experiences that became my guide during that spring, which was much smoother. I had two classes and a research component, and felt I could comfortably breathe. My research heated up and I even had some preliminary results that I could further explore for my thesis. We took some more time to enjoy ourselves as well, my friend Eric and I bumming around Boston over spring break, watching Dr. Strangelove and The Dark Knight on the big screen of the SEC. As much as I had wanted to push as hard as I could for my goal, the three weeks off over winter break had set me up with the energy and perspective to grow as a student. Ambition would soon again become a prison for me, but when you’re well-rested and taking care of yourself, you can largely turn off your own worst tendencies.

The summer at least started off well. The Celtics had squeaked into the NBA Finals, and it became a bit of a ritual for the last of us on campus to watch the games as Harvard became a ghost town. Good times. The winters here are truly awful (and windy, so damn windy), but the summers are complete heaven. But the PhD was still front and center in my mind. As a result, my whole summer would be spent doing research, but I spent that June mostly reading and working from the bottom up, reproducing results from papers. Since I needed work to my name, the goal was to submit to several NeurIPS (the biggest AI conference) workshops in September. If you’re thinking about a PhD, this is a great, low-stakes way to present your work and get some paper titles on your Scholar profile. The acceptance rates are at least a fair coin flip depending on the workshop.

The main problem was that I was juggling two projects in June, which somehow ballooned to three by mid-July. Even worse, I was working alone on two of them. Despite having no classes, I couldn’t make it work. It was a tough lesson I had to learn going into my PhD: you get, at most, one project at a time, especially when you have a terribly short attention span like I do. The funny thing is that if I had stuck to that principle, I probably would have gone further on all three, being able to circle back around after completing them in sequence. I see some people work on multiple projects at a time, but I can’t believe there’s no quality cost you pay for this. The primary lie I was able to spin into a positive was “being able to shift focus when I got stuck on one project.” All this impetus means is that you’re procrastinating and not attacking from a new angle, and you’re burning yourself out by not just simply doing nothing instead. Never distract yourself from work with more work. At this point in my research career, I’d much rather see one deep project through to completion. I talked to a former CSE grad who’s now at MIT, a much better researcher than me whose master’s thesis got nearly a hundred citations (a very impressive feat at that level), and he largely confirmed my lesson. His workflow is one paper per year. And when a PhD is essentially 3-4 high quality papers in top journals, he’s exactly right for American doctoral programs. One researcher here told me early on that the workflow is as little as 3 months (!) in machine learning, given how fast the field moves. These people are liars, or they’re at least not the folks moving the field forward. True understanding takes time and intense focus. Research is hard and arduous. Part of my drive to take on multiple projects was to “hedge my bets.” The mental switching time that it takes between projects, however, will blow you up before you even begin to gain any insight. Insight is very nonlinear: you’ll get nothing for months upon months, maybe even a year, and then you suddenly won’t have enough time to implement all of your “great ideas.” It just pours out of you. That’s what a good PI is: someone with a lot of intuition from staring at and working on a problem for a long time, but not enough time to prove their intuition in a rigorous fashion.

Nonetheless, I punched my ticket to New Orleans and got into two workshops by the end of the summer, but I had to cut down almost all of my work and abandoned two of the projects, only submitting a preliminary version of my thesis. Still, I will maintain that my writing skills were what got me into these venues in the first place. By the time I submitted further results to Computational and Systems Neuroscience in November, which was accepted in January, I knew the story inside out and the reviewer scores went from about a 6 average to an 8. Almost the same central results, but the significance was uncovered more clearly in words. However, I was also completely spent by this point in September–before the school year had even really started–and having sat at my desk for 16 hour days, 7 days a week, I started developing some nasty back pain that would plague me for the entirety of the school year, even until graduation in May. I wasn’t exercising, I wasn’t looking after my mental health. I was on so much adrenaline and anxiety that I was barely eating, subsisting on protein shakes. I just wanted to reach my goal at any cost. And while I was guzzling down painkillers and caffeine everyday to enable my lifestyle, I was not only able to temporarily hide my problems but do exceedingly well. Besides the papers that semester, I was on a panel in front of The Zuck, here to donate $500 million for the new Kempner Institute. I killed my classes. I gave a talk over at a very famous neuro lab at MIT, and met plenty of famous researchers along the way. I felt that I was making a name for myself big time, and it couldn’t come at a better time with PhD applications in December and January, which are themselves a 50+ hour time sink if you apply to ten places like I did.

After NeurIPS rolled around, winter break was like hitting a brick wall. I’d been going to physical therapy since August, but I wasn’t putting any work in besides those sessions, and had only gotten worse as I continued to lie to myself, that I didn’t need to change. If you ever go to physical or mental therapy, make sure you prepare some notes or some questions beforehand. Try out some exercises or techniques on your own time. It’s not something you just show up to and expect to be healed. My lack of investment came at great personal cost. I never slept well that entire semester, often waking up for several hours at a time with my spine on fire, and I started to need stronger pain killers. Completely unaware of its subject matter, I watched the movie Waves that January. The first half of the movie centers on Chris, an ambitious, talented, and well-liked athlete who slowly ruins his life with percocet after a labrum tear jeopardizes his wrestling career. It’s not a horror film, but I’d never been so deeply terrified in my life. I had a bottle of hydrocodone on my counter that I threw in the trash the next day. I survived the fall, barely, but I was in the darkest mental state of my life. When I got diagnosed with muscular dystrophy in high school, I had some recurrent suicidal thoughts for many years, but as Neitszche once said, “The thought of suicide is a great consolation: by means of it one gets through many a dark night.” This felt different, like a wild animal running around a museum, like my brain was going to explode and kill me without my own input. Harvard was my biggest goal, and it was starting to nauseate me.

One of the problems about deciding to change “at rock bottom” is that you’re usually realizing it on the way down, not actual rock bottom. People who’ve actually hit rock bottom usually ain’t coming back. There’s an inflection point where the mind collapses in on itself, and you’re no longer you. So when you do decide to change, expect more difficulties ahead (in the most immediate sense). This past spring was completely brutal even if I did begin to get back into exercise, a more healthy diet, and better habits. If you do apply for a PhD, expect to be brutalized in the application process. Even the best students today get a majority of rejections, and the most thoroughly impressive PhD candidate I met here at Harvard “only” got into Harvard. I was incredibly fortunate to be accepted to Columbia, very much needing a fresh start in a city I had been traveling to over my summers and winters whenever I could from Boston. It checked all of my boxes research-wise, and I was even more excited that GSAS funds their PhDs, not their individual PIs. In my experience at Harvard, the latter funding situation creates perverse incentives where the advisors look at students like employees. Columbia was a no-brainer for me. Another audible. I did it, I achieved my goal, but I was absolutely spent. It reminds of the final scene of Robert Redford in The Candidate: he’s spent the entire film doggedly pursuing the election, which he’s finally won. Now, in the Senate for the first time, he turns to his advisor and asks “What do we do now?”

Goodbye Brown Brick Road

After two straight years of madness, I crawled to the finish line on May 25, 2023. Harvard is a much darker place than it projects to the general public, and boy does it project well. Graduation is a decadent tradition here, one with all of the trappings, and it leaves a sweet taste in your mouth about the place even if it IS the Belly of the Beast.

I woke up on May 26 feeling like it was indeed all over for the very first time. Like I had been doing in my brighter days, I felt ready to reflect, now that I was detached from it, at least at an emotional level. Questions!

Was it all worth it? Was my victory Pyrrhic? I’ll never be naive enough to think that any accomplishment doesn’t come without tremendous sacrifice. If you want to win in this world, you nearly have to die (in some sense) to get to the absolute top. A big part of my ego died in Boston, and that’s both a good and bad thing. Yes, I was dumb when I came here. Yes, I’ll probably wonder what I was doing now, at 24, when I’m 30 or 40. But the naivete of youth is a hard thing to lose. There’s nothing so brash and so pure as the sort of confidence I came here with, one that is now modulated by the difficult lessons I bought with my experience. I came to Harvard eager for its prestige and its shine. There’s nothing like seeing The Yard for the first time: it inspires a sense of wonder I’d never felt before.

Part of me can’t feel that way again without a few years of distance between me and this town. Maybe ever. You know what Harvard is? It’s not the best place on earth, not one of the few places reserved for the best of the best. When I was at UNC-CH, I said the same thing about Michael Jordan every time I passed Old East Hall, and being from my crappy hometown in Carolina during high school, it was the great saxophonist, John Coltrane, that I wanted to emulate. Harvard is this: a foot in the door. It’s a place at the table. It’s not the food, or the company, or the restaurant. It’s just a ticket. And man, I’ve some people eating themselves alive at Harvard. Miserable people with no sense of humor, only able to see the next achievement, the next promotion, the prestige of it all. They will destroy you, they will destroy themselves, because they want nothing more than to NOT be themselves. That’s why people accomplish anything, anyway. Most things, that is. That’s why academics are some of the most brutal people: we don’t care about money or wealth. We want worship.

I’ve also seen some of the most brilliant, kind people I’ve ever met here. All of them shared one thing in common: they don’t see themselves in their job. A lot of things, including research, are art. Art is just an eye for what was previously unnoticed and unexpressed, giving form to what is valued. When you try to be a screenwriter, or a doctor, or a scientist, or a photographer, you are no longer an artist, but a word. And the word is never the thing it describes.

Passion and romance are necessary in this world. Sometimes it pays to just put on the rose colored glasses and be in awe of it all, totally uncritical and immersed in the experience. But most of the important battles in life are fought internally, and this requires a healthy mixture of skepticism and trust, not prestige or a rock-solid resume. The next wave of bright-eyed Harvard students come and go, and the world moves on. Veritas.