The Future Of Computer Science

Computer Science
96 followers
2 Video Views·Apr 19, 2026
I'm graduating with a computer science degree this spring and I've watched the field change
drastically right in front of my own eyes these past four years. So I want to lay out where I
think it's headed to hopefully help some of you. I think AI slop will disappear or at least start
to get filtered out because we pretty much saw this huge trend of companies adding AI into
everything. Like AI video generators, AI images, there's an AI chat bot in every website now. But
I think companies are going to realize that no one actually uses these things and I think most
people are tired of getting AI pushed down their throats. So this means the companies pushing the
AI slop are either going to get wiped out or get forced to pivot to making things that actually
impact people. So I can see the useless parts of the AI bubble start to pop as the years go by.
But what does that mean for a CS student or a developer? Well, I think you're going to be
expected to show a higher level of proof as the years go on. I think companies are going to want
you to use AI, but to make something actually meaningful. You're going to be expected to be
able to leverage AI till its fullest in order to learn and ship real things. So essentially the era
of building projects or learning something just for the sake of it is over. And obviously if you're
just starting out, you're going to have to make some projects just for the sake of it. And I
definitely encourage that. But as you progress, you really need to have some sort of intent to
show competency. Like if you can make a project or even just write code that someone else had to
approve, like a pull request that got reviewed and merged or a project where users depend on it,
if it's just 10 people from your school or 5 people from a discord server. Basically,
if you can make anything that shows you didn't just vibe code some AI slop, that's the best
new signal for competency as a student. Not how big your project is or what tech stack it uses,
it's can your work hold up under someone else reviewing it or when people actually try to use
it. Because that's the new best indicator that you actually know what you're doing and you didn't
just push some random slop. And so I think that's what's going to matter more than ever.
The generalist developer I'd say is the most at-risk role in software right now. Because AI
is simply better than you at bre- and probably not just slightly better, but way better. It can
do a little bit of everything pretty good and very fast. However, where I think it really struggles
is simply just going in depth, going deep enough in a specific domain to actually understand what's
really happening. Because think about it, when you want to do something really good,
you really can't rely on AI. For example, if you want a really good looking website,
you can't rely on AI to make the design because it really only makes generic UIs. Or if you want
to write low latency code, you also can't rely on AI because you know you have to benchmark or
triple check your code. And like, let's just say you're doing something like embedded. Do you
really think AI can replace a specialist embedded programmer? Or what I'm trying to say is, do you
really think AI can replace any specialist? And I think if you say yes to that question,
then you don't really know how deep a lot of these things go. So as a student or any developer,
that's where you can come in because the only real moat left is specialization. I think in the
world of AI, you need to pick a lane within computer science and go deep enough that you
can answer any questions without help, know how to debug without guessing and know how to look at AI
generated code and just feel when something's wrong. And I think a good sign of going deep in
a domain is being able to make things that test your fundamentals just at the top of your head
without any help. So if you're front end, you should be able to build something like Wordle
in React or whatever framework. If you're back end, you should be able to build a basic API
server from scratch. If you're into systems, you should know how to build like a simple shell or
a file explorer, understanding how programs actually run. And I think if you can get to
that point where you've specialized so good that you have a good enough baseline competence to
write simple things in your field, then you'll stand a great chance with getting a job. But
here's the problem with computer science right now. It's that schools are still producing
generalists. They'll essentially give you an intro course, or maybe if you're lucky, you'll get like
two courses on your specific field throughout your entire four years of college. And that simply
isn't enough because you're going to end up with broad exposure to everything, but really shallow
depth. And then you graduate and you're supposed to somehow compete with people who have years
of experience. And it's almost like you're just not even going to have a chance. And so I think
there's a huge widening gap between what school teaches you and what actually makes you valuable.
And I really believe it's going to get wider until the curriculum starts getting better. And so in
my honest opinion, as a computer science student, I would say to use AI to get through all of the
irrelevant stuff as fast as you can and spend that saved time going deep on something and trying to
get any experience that you can get. In the upcoming years, interviews are going to get a
lot harder. But what I think people are getting wrong is that they think that the data structure
and algorithms interviews are going to go away because AI can solve the code problems. But this
style of interview was never really about finding the solution. Anyways, the interviewer already
knew the solution. All of these problems have been answered online for the entirety of leetcode. The
whole point was always to see how you think and communicate under pressure. And I just can't
really see that changing that much. What I do think is changing is that smaller companies and
more mid-level companies are going to start adding more practical interviews and they're going to
become a lot more common as the years go on. I think more companies are going to want to watch
actually build something alongside an engineer. And so that's a completely different skill than
solving leetcode questions. So I think you need to start doing both. Keep doing leetcode, work
through neetcode 150 as early as you can, get to the point where you can really think of at least
some sort of approach to any medium problem that you see. But you should also try and practice
building. And I don't want that to scare you. It really shouldn't be you building anything crazy.
You can just ask AI to give you something to make, something that just shows baseline competency for
whatever role you're going for. And then just close everything. Don't get any outside help.
Just use your code editor. And if you do that once or twice a week, you will find your gaps
faster than anything else. And also on the topic of interviews, just make sure you have at least
two things to talk about for every bullet point on your resume. Practice the star method, even when
you don't have anything coming up. Because the worst thing that can happen is you finally get
a call and then you fumble what should be the easiest part of the whole process because you
couldn't remember what you actually did. Okay. This one is probably the most optimistic prediction
that I have for the future of computer science. I think the whole computer science grind is about
to get a lot more fun, which I know a lot of people are going to disagree. But in my view,
as AI takes over the boring parts, like doing random prereq homework or helping you learn the
syntax of a new language, what's left is really the fun stuff. And the only way I can argue this
is really that you need to have like-minded friends. Because when you're building things
that you care about with people you like, it really stops feeling like a grind. Like getting
no sleep at a hackathon with your friends, working on something cool, in my opinion, that's pretty
fun. Or joining a club and working on a genuinely cool project with people who actually care, that's
fun too. Or studying at the library with your friends is also pretty fun for like a final or
a midterm coming up. And I don't know, maybe I just have a warped view of fun. But in my opinion,
I really think the whole grind culture can be fun if you just find the right people. So this is
honestly my biggest piece of advice for computer science moving forward and in the future. Find
your people, people who are curious and who want to get better and who actually care. Because if
you have that, then you really don't even have to force consistency. It just kind of happens.
And if you're struggling to find those kinds of people in person, there is a subreddit called
r slash programming buddies, where people post what they're working on and look for people to
build with. But definitely try finding friends in person at your university first. Join a club or
go to a hackathon or talk to the people in your classes. And also on the topic of fun, I would say
try and look at all this AI stuff, not as a threat, but just as endless opportunity. Because
the barrier to building something real has really never been lower. You can ship something with
actual users faster than any generation of computer science students before you. And I think
the students who genuinely get excited about that and run with it, those are the ones that are going
to have a really exciting few years ahead. And in order to ensure that you're part of that future,
you actually need to have the fundamentals to back it up. And that's why I think Brilliant
is worth talking about, because it's genuinely one of the best ways to build those fundamentals
in a way that actually sticks. Brilliant today's sponsor is an interactive learning platform that
teaches you math, CS, and programming through actually solving problems, not just watching
videos. And I think that distinction really matters a lot because there's a big difference
between watching someone explain an algorithm and actually working through it yourself. Brilliant
makes you do the latter, which is why concepts actually stick instead of just feeling kind of
familiar. For CS students specifically, their algorithmic thinking course is definitely the
one that I would point you to, especially if you're trying to understand the concepts behind what
you're studying and not just memorize them for an exam or a quiz. And I've been trying to really
master Python. And so if you're like me, they specifically have a thinking in Python course,
which is another great one that I definitely point you to as well. And the courses are built by
MIT, Harvard, and Stanford. And it works for pretty much anyone, whether you're just starting
out or trying to fill gaps in your existing knowledge. So if you want to try it out,
Brilliant is giving you a full 30 days for free at brilliant.org slash lattice. And if you decide
you want to keep going, they're also giving my viewers 20% off an annual premium subscription.
Link is in the description. So thank you so much to Brilliant for sponsoring this video.
And thank you so much for watching.
