Replit's CEO: The Future Company Has Just Two Jobs

In Brief

Replit just raised $400 million at a $9 billion valuation. On the back of that round, co-founder and CEO Amjad Masad sat down with Y Combinator general partner Andrew Miklas and made a striking claim: the company of the future has only two kinds of jobs left, builders and salespeople, and almost everyone inside it acts like a founder.

Miklas pushed back gently throughout the interview. Where do real people fit in a company that's "agents all the way down"? Masad's answer is less science fiction than it sounds.

The claim: builders and salespeople

Late in the interview, Miklas asks the question everyone wants answered: if AI keeps eating tasks, what is left for humans inside a company?

Masad's reply is blunt:

"...the company of the future is made of builders and salespeople."

That's the whole frame. Two jobs. Everything else, he argues, gets folded into one of those two roles or handled by AI.

He doesn't mean salespeople in the old sense. The job description shifts. Sales becomes mostly evangelism and education, helping other companies understand what AI can do for them. The builder role expands the other way. A builder isn't just a software developer. It's anyone with an idea who can deputize agents to make that idea real.

What does "builder" actually mean?

This is where Masad's vision becomes concrete, and where the pedagogical shift happens:

"Almost everyone is a founder. They wake up in the morning and they think how can I make the company more successful? How can I make the company make more revenue?"

The builder is a generalist with a founder's mindset. They walk around the company looking for problems. When they find one, they don't file a ticket and wait. They build a small piece of software that solves it, often by telling an AI agent what to do.

Replit calls this vibe coding. You describe what you want in plain language, and the agent writes the actual code in the background. You stay in the driver's seat. The agent does the typing.

The shape of the role is recognizable. It's the PG product person, the resourceful generalist Y Combinator has spent two decades describing. What's new is that this person doesn't need to know how to code anymore. The technical wall that used to separate "people with ideas" from "people who can ship" is dissolving.

The case study: Replit's own internal team

Masad gives a real example from inside Replit. They have a small team with what he calls a "vague mission": walk around the company and make it better.

This isn't engineering work in the traditional sense. The team's job is to spot a problem, build the fix, and move on. The actual coding gets done by a Replit agent. The humans bring the judgment, the conversations with the affected team, and the taste for what should exist.

That's the builder role in practice.

Why salespeople survive

Sales is the more surprising survivor in Masad's model. AI is supposed to eliminate friction, and salespeople are pure friction in the old framing. So why keep them?

His answer:

"A lot of other companies will want to talk to someone, will want to learn from someone. That's how a lot of people learn, and they trust other humans. So the sales part is one of the more defensible jobs."

Trust doesn't transfer to bots, at least not yet. When a buyer commits real money, they want a human across the table. They want to be taught how the product fits their world, what to try first, what to avoid.

So the sales role morphs. Less "pitch and close." More "evangelist and educator." Masad describes Replit's own enterprise sales motion this way: come in, do a hackathon with the most curious team, train leadership on what AI can do, and let the product spread from there.

Where the model breaks (Masad's own caveats)

Miklas keeps pulling Masad back to reality. What's still missing for this future to work? Masad lists two big gaps openly.

Computer use models are disappointing. These are AI systems that operate a computer the way a human does, moving the mouse, reading the screen, clicking buttons. The dream is an agent that can use any old software, including the legacy systems most large companies depend on. The reality is that these models are progressing slowly, and Masad calls the slowness "kind of a mystery."

Continuous learning isn't real yet. Today an AI agent doesn't actually learn from its job. It picks up patterns within one conversation, but the next conversation starts fresh. The current workaround is to write notes to a file the agent re-reads later. True on-the-job learning, where an agent gets steadily better at one organization's quirks, "still seems far away," Masad says.

Both gaps matter. A future where everyone is a builder needs agents that can drive any tool and remember what worked last week. Neither is here yet.

What this means if you're not a developer

The most important sentence in the whole interview is buried in the discussion of what skills matter now. Masad puts it this way: the moment you understand that you can solve a problem with code, it changes how you see the world. Suddenly the broken process at work, the missing app for your hobby, the spreadsheet your team hates, all of those look fixable.

That mental shift used to require years of programming training. Vibe coding compresses it to an afternoon of trying things in a tool like Replit.

If Masad is right about the future, the practical move is small and immediate: pick one annoying problem in your daily life or work and try to build the fix yourself. Not to become a developer. To get the muscle memory of looking at the world as a builder.

That muscle is what either of his two future jobs needs.

Glossary

Sources and resources