For years now I've seen young programmers ask the same question: "Should I go into frontend or backend?" Today, my answer is categorical: backend. Let me explain why.
In the excellent film Margin Call, which depicts the 2008 financial crash, the bank's CEO describes three ways to succeed: be a genius, or be first. (Or cheat, but that's for a different article.) While being a genius is always a good thing, it's also rare and hard. It's much easier to understand the trend and simply be first. (That's how his bank survives the 2008 collapse: they were the first to get out of mortgages, by a few hours.)
When I was working in Silicon Valley in 1998, the hottest word was "portal." Yahoo was the hottest internet giant, the darling of Wall Street, the company everyone wanted to be. Their model: they were the gateway — the "portal" — to the internet. You started your internet session at Yahoo.com, and from there you went to whatever you were looking for. If I wanted, say, a recommended sushi restaurant in Palo Alto, I'd go to the Restaurants category, open California, open Bay Area, open Palo Alto, open Sushi, and there find links to various Palo Alto sushi restaurant websites, other sites that rated restaurants, and so on. There was also a search option on the homepage, but search was very primitive back then, and Yahoo didn't pay much attention to it. It returned endless irrelevant results and mostly tried to sell you things.
So Yahoo was to be the portal to the internet. Everyone would go through it to get to what they were looking for, and it would make money by selling real estate on its pages: ad placements in the highest-traffic locations on the internet. Its homepage, and all the secondary pages, were so busy it made you dizzy just looking at them. (Portals weren't anything new, by the way. Before them there were directory services, which portals were an upgrade of, and which partially overlapped with an earlier service called Gopher. So there wasn't much originality there.)
It seemed like all of Silicon Valley was obsessed with portals back then. It was standard in every interview to reach the part where they'd say, "So tell us about your experience building a portal." If you didn't live and breathe portals, you probably didn't belong in tech. Have you considered becoming a personal trainer, or a yoga instructor, or opening an artisanal burrito shop for the real tech people? Have you maybe thought about leaving Palo Alto, moving to Arkansas, and retraining as a teacher? There you could make a difference…
Into this situation came one person with the simplest insight in the world: you're treating search as if it's just another feature, but it's the heart of the matter. A portal is a dumb thing. Why should I walk down a tree, opening more and more branches until I finally reach my area, when I can simply type "sushi 94301" and immediately get all the sushi options in downtown Palo Alto? Give it one minute of thought, and you'll clearly see that search steamrolls the portal concept.
It seems amazing today. Such simple, obvious logic, and yet nobody thought of it: search beats portal. Over the years I've learned that most of the biggest ideas are remarkably simple, and most people won't see them. Everybody in tech at the time loved to say "paradigm shift." But real paradigm shifts are hard to recognize while you're in the middle of them.
Larry Page built Google around this simple concept: search, search, and more search. The portal that everyone was chasing without asking too many questions had had its day. From a business standpoint, would search generate as much revenue as selling internet real estate? Page said, essentially: when the numbers from advertising on search results come in, we'll see, but I know search is the future. The numbers did come in, and the rest is history.
The reason I bring this up is that very simple ideas can change the entire tech industry, and this was one of them. Just as it was inconceivable in 1996–1998 that search would be the way the internet works, it's inconceivable to people today that frontend is dead.
The fact that you've invested endless time learning frontend doesn't matter. The fact that there are countless podcasts on UI/UX, that people have built careers on it, that there are endless conferences and people fly around the world to hear the latest UI/UX expert speak, doesn't change the fact that the user interface is going exactly where the portal went: into the dustbin of tech history.
Think about what AI agents have achieved in the short time since they appeared, and think about where we're headed. The future of software in, say, the travel business isn't in the interface for finding you a hotel on Booking.com or Airbnb. It's not about how beautiful your site is, how convenient, how welcoming to a new user, how attractive, how it makes the user feel at home, how it builds trust, and so on. The future here is saying one sentence to an AI, "Plan me a vacation from September 15 to 22," and how well it does it. In the future, which is already here, you'll just say things like "Do my taxes" or "Transfer money from one account to another." Eighty to ninety percent of frontend will be gone. Of course, some of it will stay, because agents will sometimes need to show you images and videos, for example. That's the 10–20% that will remain. But no more radio buttons, drop-downs, widgets, or UI components. No more debugging around all the CRUD your beautiful interface performs, no more enable/disable logic, modals, tooltips, accordions, carousels, date pickers, autocomplete, pagination controls, or breadcrumbs. No more obsessing over bounce rates. No more state management libraries, prop drilling, re-render bugs, multi-step wizards, onboarding tours, checkout funnels, A/B testing button colors, usability sessions watching people fail to find a certain page, and starting all over again with a redesign every two years. The interface is: you talk to an AI, which beats the most beautiful web page you ever think you'll build.
It's not just that Claude Code does excellent frontend work. Yes, it does. But frontend right now is a kind of hybrid that will be here only very temporarily. Using AI to create UI is like the first cars that replaced carriages. They put the driver's seat up front, high up, because the carriage driver sat above the horse, steering it and whipping it. The engine had just arrived, and people still hadn't grasped the full meaning of the fact that… there's no horse anymore. You don't need to sit up high to drive. Over the following years that went away, and gradually the driver's seat moved to where we know it today. That's how working on frontend looks to me today.
So in short: you don't need to be a genius. Just be first. Understand where we're going: to a world where we talk to software in English and it simply does what we want. For a career over the next 20–30 years, invest in backend, or algorithms, or anything that isn't the old way of interfacing with software, even if frontend is what you love and you've invested so much time mastering it. Reality will march on. Don't get left behind.
Top comments (0)