Arduous effort is needed for arduous goals

Arduous is an adjective that means requiring great physical or mental effort, endurance or labor.

This was one of those demotivating weeks. I fell sick and my research progress stalled because of it. I couldn’t read any research papers this week and got completely derailed. Now I’ve recovered and I’m all good. But even though I took a whole week off, nothing fell apart. That’s when I understood that you can do just the bare minimum and get away with it. Then why all the effort? What’s the use?

The goal is still the same: become the best possible computer guy, aka a world-class engineer, and swim in this huge world of computer science as much as possible. I want to build cool stuff and come up with cool ideas that can at least make my life better. A computer is a beautiful tool and I want to use it to create some sort of impact with whatever knowledge I gain. The goal is audacious, the goal is arduous and all of that is fine. But when you have goals like these, the effort needed to get there is also insanely high.

I recovered from my fever in 2-3 days and started reading a paper called Megatron from NVIDIA’s lab. I wasn’t able to understand anything from it. I couldn’t even get through a single page. For the past 2-3 months I’ve struggled a bit with every paper I read, but over 8 to 10 hours I’m able to understand what’s in the paper and the concepts behind it. Then it takes another 6 to 10 hours to really get the essence of it. But with this one, I got stuck in the second or third paragraph of the introduction and couldn’t move forward at all.

Part of the problem is how I learn. I don’t care about marks anymore. Pass, fail and all of that doesn’t make sense to me. I have my own logic and my own pass or fail criteria, and I have to satisfy that. That’s what matters to me. One big part of it is that if I don’t know something, I don’t move forward until I understand it completely. If I start skipping things, the odds of me coming back to the difficult part are very, very low. That’s why I do something I call onion learning (yes, I named it xd). I read something, and if I don’t know the depths of it, I start peeling the onion one layer at a time until I get the reasoning. I won’t move on until then. I’m bad at memorizing stuff, so this is the only way I stay conceptually tight. A high-level reading just doesn’t stick in my mind.

But with this paper I couldn’t move forward because it involves a lot of things: machine learning, systems programming, NLP, computer architecture and a bunch of GPU stuff like tensor cores, GPU internals, blah blah. Since I don’t know all of it at once, my understanding of the paper was very shaky. When somebody talks about some NLP thing, I have to dig into why they say it. Same with architecture. If somebody mentions pipeline parallelism, I need to know what exactly it is, how it works, what its pros and cons are and why it was designed that way. I felt super demotivated. I spent the whole day on it, but because I was so down, I’d say I effectively spent about 4-5 hours reading. And I still couldn’t get through a single page.

Coincidentally, this paper was part of my PhD 6000-level class, so I went to class the next day. A lot of folks were still clueless. But this class is not like other classes where you can just fool around. There are about 15-20 folks and they’re extremely ambitious and very focused on depth, not fluff. The class also isn’t the professor teaching while we listen. It’s more like we teach the professor and the professor teaches us back. It’s a continuous back and forth.

Even these guys had trouble explaining one of the other papers. They felt the second paper was even harder because it wasn’t written well. Meanwhile I couldn’t even get past the first one. It was super demotivating that everybody at least understood the first paper while I couldn’t get through a single page.

My professor is also my operating systems professor, so I could just go talk to him after class. I went and ranted for about 5 minutes. I told him I couldn’t understand a certain hyperscalar NLP part, and that my understanding of GPU internals wasn’t clear either because I’d completely forgotten some computer architecture concepts. So I was having trouble understanding the paper at all.

He listened to all of it patiently. Then he asked me why we should read this paper at all if it’s way more than we can chew. I told him it’s a hard paper because I’m venturing into unknown territory and not a lot of folks know all of these things.

He said there’s a reason not a lot of folks know all of these things. There’s a reason very few folks really know what they’re doing, and those folks control the whole industry. Take something like React. Say you want to learn front-end React, back-end Python, connect to a database, or Next.js, skills like that. They’ve been around for years now. You just pick up some concepts from the internet, try some stuff out and write about it. Take something more recent too, like MCP. It’s been around for about a year. If you want to know about MCP, you can just Google it. The information is very straightforward, it takes half an hour and then you can build stuff with it.

But when you want to go deeper into the so-called cutting edge, you have to start thinking from first principles and come up with your own ideas. That’s where the industry is within reach. You could literally ping someone at whatever company you want and say: I have this idea, can we work on a paper? Or: I have this concept, I put in the work and I think it’s novel. Or: I took an idea from a paper at this conference and built on top of it, and when I combine it with an idea from this other paper, I see a gap we need to fill. When things change every few months, not a lot of folks know all of this stuff. Getting to a place where you know it all takes a lot of effort.

And then he told me this. If you feel it’s hard, the guys who are moving the needle felt the same difficulty. The only difference between a normal guy and those guys is that they put in an insane amount of dedicated, focused work. They solidified their basics and skills over years, plus a little bit of luck. The only way to increase your surface area of luck is to have the skills and keep putting in the effort. Then some random day an opportunity shows up, you have the skill and the two match. Take luck out of the equation and it’s all about your skills, which take a lot of time to build. Even they went through the same cycle of despair, of not feeling great, of constantly putting in more effort than everyone else. They walked the same path you’re on.

Taking the easier path is always easier and there’s always some benefit to it. It’s just a trade-off. You take the easier path and you get a so-called better quality of life, if that’s what you mean by quality of life. It’s a trade-off you have to pick. But everyone who’s in the spot I want to be in had a miserable number of crash outs and months or years of silent hard work. Only then did they get there. There’s no way around that kind of ruthless prioritization and focus, giving up some parts of life. If everybody else is putting in 4-5 hours, you need to put in 8-9. They put in that kind of work and that’s the only way to catch up.

So he told me that if I find this stuff difficult, all I need to do is put in more hours. That’s it. There’s no magic pill. You can be smart or dumb or whatever, you just need to put in more time. As simple as that. He said this and left, but it completely made sense to me.

If you have insanely ambitious goals, they need an insanely high amount of effort and hard work. And I have to get out of the habit of comparing myself with others, because their goals are not the same as mine. There’s no point running the same race as everyone else. Everybody runs at a different speed and everybody’s finish line is different. So it’s high time I enter a phase where I shut out the outside noise and just focus and move forward. That’s a trade-off I have to accept and chew.

Working hard doesn’t guarantee success. I could fail miserably after wasting months locked in. But if there’s even a small chance for me to do good work, it’s only possible with that effort. I still believe in periodization of hard work, like I wrote about in this blog.

Life eventually becomes a pursuit of ruthless prioritization. Every useless yes is a no to something useful. You have to put yourself first. The season becomes a season of silence, of focus, of choosing yourself unapologetically. You stop explaining, you stop people-pleasing, you stop shrinking to make other people comfortable. You have to protect your time, guard your energy, say less and do more. Have insane urgency in whatever you’re doing, with a lot of no’s to a lot of things.