Replit's Founder Found His Next Product at the Chessboard

Source: a16z | Published: 2026-09-23T14:30:06Z

Universities systematically teach silence to young people at the exact age they should be becoming heretics, says Amjad Masad — and he frames it as a civilizational loss, not just an economic one.

One day, the machine glitches, and a truly new person is born. He starts causing trouble from day one in the training camp: doing strange things, asking strange questions. The strangest of all: what's beyond the city walls?

Everyone says: we don't talk about what's beyond the walls.

Amjad Masad, founder of Replit, thought of this story when talking about higher education today. "That person needed a malfunction to exist," he said. "But every society needs people like him."

Universities Are Doing Exactly What Young People Don't Need

Amjad's take is blunt: the most valuable thing young people offer society is the willingness to question deep-seated beliefs — to be heretics. But modern universities do the opposite. They tell you: we've already figured everything out. Just learn our methods, get ready to find a job, buy a house and a car, and live a happy life.

The problem is that world no longer exists. You can excel in college and still end up unemployed and drowning in debt.

But the bigger loss isn't economic — it's civilizational. People who might have asked disruptive questions are systematically taught to be quiet, at precisely the age when they'd be most likely to ask them.

The Steve Jobs Elective Logic

If he could redesign the four years between 18 and 22, Amjad says he'd want a school that doesn't tell you what to study.

He brings up Jobs' famous Stanford commencement speech — Jobs enrolled at Reed College but couldn't afford tuition, so he dropped out, then stuck around auditing classes. One of them was calligraphy. He didn't know why it interested him; he just went. Years later, when building the Mac, they decided it needed proportional typefaces and multiple font options — ideas that came directly from that college calligraphy class.

"This kind of aimless exploration — modern society thinks it's useless," Amjad said. "But I think it's exactly the most valuable thing college has to offer."

The other thing he'd want: when you get genuinely consumed by something, the school gives you space to pursue it. When he was a student, attendance was mandatory — miss more than six times and you were out. "If I have an idea and I'm grinding on it, you can't expel me just because I skipped class."

Premature Optimization Is the Root of All Evil

On one side: four-year universities, rigid and structured. On the other: someone handing Stanford students million-dollar angel checks before they even have an idea. Amjad thinks both paths have problems, but the latter carries a particular danger — what he calls "premature optimization."

There's a saying in programming: premature optimization is the root of all evil. It means that when you start optimizing a program before it's actually running, before you know where the bottlenecks are, you end up wasting time on things that don't matter.

The same applies to startups. He's seen plenty of kids take funding because their friends are raising rounds — some even becoming unicorns — driven not by genuine curiosity but by envy. They commit prematurely to an idea they don't actually care about, then try to optimize along a path they haven't earned yet.

"For me, Replit almost accidentally became a company," he said. "I was actively trying not to start it. But it became what it became on its own. The feeling was: I have to do this, because it already had so much momentum."

That, he says, is the signal worth following. A VC handing a Stanford student a million dollars on campus is — in a certain sense — the same logic as a record label finding a young artist and locking them in. Kanye has talked about how predatory that is.

From Chess to ML Researcher

Amjad came to chess late — something he admits is typically learned at age six. He can't explain why he suddenly got interested; he just followed the feeling.

Then he started writing programs to "solve" chess, and noticed something strange: ChatGPT is terrible at chess, hallucinating illegal moves. That sparked a question: ARC-AGI scores keep climbing and these models are supposedly approaching AGI — so why can't they play chess?

He spent a lot of time researching how to train models — a field he'd never touched before. Along the way, he realized AI already has the capability to do machine learning research. Before Replit Agent launched in 2024, the idea of an agent writing code was still a novelty. He feels ML is at a similar inflection point now — Karpathy calls it "auto research." You could call it "vibe research."

That exploration eventually became a new product direction for Replit: letting anyone — including business owners with nothing but their own data — fine-tune models to solve specific problems.

From chess to a new Replit product line, Amjad says he couldn't see the path when he started. "But as long as you follow what interests you, it eventually leads somewhere — at least that's how it's worked for me."

The Boundaryless Company Is a Technology Problem

Holacracy, Medium, GitHub — all tried flat, managerless organizations. All failed spectacularly. Amjad thinks this is a technology problem, not an ideological one.

He's now experimenting at Replit with what he calls a "self-driving company": agents as the connective tissue inside the organization, bureaucracy running silently in the background, humans with genuine autonomy to work on what they care most about, and AI evaluating impact.

He admits he doesn't have a fully formed vision, but he's certain of one thing: all of this is going to change. "If you don't understand it natively, that transition is going to be hard."

How to Be Just Disagreeable Enough

Amjad frames it this way: you need to be contrarian enough to have genuinely interesting ideas, but some people are so contrarian that nobody wants to work with them.

On the idea of "learning how to follow," he reframes it: you need to learn how to collaborate, how to navigate systems, how to get what you want — the optimal outcome.

Every conference room at Replit is named after a legend of computer science and technology: Marc Andreessen, Linus Torvalds, Robert Noyce. "We respect the past, but we'll change it without hesitation."

Last year, he was piled on for saying "you shouldn't learn how to write code anymore." A year later, the same people who pushed back are saying "I don't read code anymore — I just prompt my code."

The Only Thing That Impresses Him at 18

Someone asked: when you meet an 18-year-old, what makes you think "I need to be around this person"?

His answer was simple: they've gone deep enough on something that they can teach me something I don't know.

"Most people get bored easily. Most people do conventional things," he said. "But have you actually followed a thread all the way to the end? That's rare, and it reflects every good character trait — persistence, intrinsic motivation, willingness to work on something without immediate reward."

One more thing: strong opinions, but not delivered in a way that makes people feel like idiots. He's met plenty of smart young people who talk to others with a "you just don't get it" attitude. That makes people unwilling to invest time in you.

Being able to go completely down the rabbit hole in some niche domain — while still making people feel like you're a good person. He says both are real and visible. And neither alone is enough.