Jordanian-American software engineer and founder/CEO of Replit, an online coding platform; previously worked at Facebook.
Amjad Masad says Replit's revenue sat at roughly $2-3M for years, then jumped 100x to $250M between 2024 and 2025 after launching Replit Agent.
“We went from $2.5 to $250 million in one year. That was between '24 and '25.”
During Replit's darkest stretch (a layoff, a cold empty new office, daily resignations), Masad describes a schizophrenic split: the whole company lost faith except one room building Replit Agent, where the mood was electric.
“So whenever I go to the office, everything is depressing except this one room we call the war room. And I walk into the war room and the mood is intensely different. Everyone is super pumped. They know we have created something that's going to be super valuable.”
Masad borrowed the video-game 'early preview' framing to launch Replit Agent semi-broken: tell users it may be buggy, opt them in knowingly, and let a product that works 50% of the time still amaze the market.
“I was like, we got to launch this. I don't care if it is semi-broken. If you get 50% of the time, get amazing results, it's going to wow the world”
Steal thisFrame a half-finished launch as an opt-in 'early preview' so buggy software can still create wow without breaking trust.
Masad borrowed the video-game 'early preview' framing to launch Replit Agent semi-broken: tell users it may be buggy, opt them in knowingly, and let a product that works 50% of the time still amaze the market.
“I was like, we got to launch this. I don't care if it is semi-broken. If you get 50% of the time, get amazing results, it's going to wow the world”
Steal thisFrame a half-finished launch as an opt-in 'early preview' so buggy software can still create wow without breaking trust.
Riffing on Clay Christensen's sustaining vs. disruptive technology, Masad distinguishes businesses fighting over existing demand in a near zero-sum market from 'market creation' moments where a new capability triggers an explosion of demand with no alternatives.
“Most businesses are trying to compete in a more or less zero-sum market. So they trying to capture market from existing demand on existing competitors. Right now there are market creation moments. And I think for Replit, this was a market creation moment.”
Steal thisDecide whether you're capturing existing demand (slog, out-execute) or creating a new market (pivot until it pulls) and play that game accordingly.
Masad's rule: if you're entering a known market with a business plan, prepare to slog and out-execute on team and process; if you're inventing something that didn't exist, pivot relentlessly until it hits a hidden truth about human nature.
“If you're a young guy trying to make something that didn't exist before, or, you know, like, like Sean and some of the guys in social media, they're trying to create a new clubhouse or some kind of thing that I didn't know I would enjoy talking to people, taking a walk on my earbuds. This is just like, didn't occur to me. So if you're inventing something like that, you will, the point is pivot, pivot, pivot until it hits.”
Steal thisMatch your strategy to your bet: out-execute in known markets, pivot relentlessly when inventing something new.
Masad points to MedVie, a New York Times-covered one-person company selling GLP-1s, as the biggest app running on Replit. The founder builds back-office interfaces and automations on the fly, going from idea to prompt faster than anyone Masad has seen.
“MedVie, the billion-dollar one-person business covered by The New York Times, basically they were a company that sold what? GLP-1.”
Masad points to MedVie, a New York Times-covered one-person company selling GLP-1s, as the biggest app running on Replit. The founder builds back-office interfaces and automations on the fly, going from idea to prompt faster than anyone Masad has seen.
“MedVie, the billion-dollar one-person business covered by The New York Times, basically they were a company that sold what? GLP-1.”
Masad's top idea for solo builders: because software is now cheap to make, target local-style businesses that still have no software at all. He cites a UK founder at $100K run rate building ice-rink management software.
“He is building software for ice rink management. So skating rinks and all that. I think that you go around in your life and there's a lot of things that are just not computerized yet. There's no software running it. So that, that would be my first approach.”
Steal thisFind a local-style business category that still runs on paper and build the software no one has bothered to make.
Masad invokes the programmer's adage that laziness is a virtue: hunt for the chores you repeat constantly and automate them. With AI as a magic automation machine, a million others likely hate the same chore.
“What are the kind of things that are so annoying that you have to keep doing over and over again? There's probably a million, maybe a billion people that feel the same way, and perhaps you can make a piece of software to automate that.”
Steal thisList the chores you repeat daily, automate the most annoying one with AI, and assume millions share the pain.
Using Hamilton Helmer's Seven Powers, Masad argues LLMs are fundamentally commoditizable (one click to switch in Cursor) with no network-effect or scale moat. The only natural moat may be continuous capital to keep training the next model, which is good news for app-layer entrepreneurs.
“We haven't seen any of those moats sort of emerge from these foundation model companies. And so my hypothesis is that maybe the one natural moat is capital. You need not just one-time capital to enter the market. You need continuous capital to train the current model, the next generation model, and the next generation model.”
Masad explains why he unexpectedly loves sales: consumer growth is like weather (A/B tests, hype, viral spikes outside your control), while sales is a contact sport where applied effort directly affects whether you win the deal.
“It's more of a contact sport than consumer. Like consumer, you do something and then you do an A/B test and maybe 2 weeks later you find out whether it worked or not. And there's this huge delay and there's so many things outside of your control.”
Masad recounts the Vercel breach: an employee at vendor Contacts.ai installed a backdoored Roblox cheat, compromising that company; a Vercel employee had Contacts.ai connected via Google OAuth, giving attackers an entry point to escalate privileges into Vercel's databases.
“Companies that downloaded a Roblox cheat software, and that Roblox cheat software was infected with some kind of backdoor. And then that company became fully compromised, and someone at Vercel installed Contacted AI through Google OAuth on their workspace, and then they had an entry through contacts into Vercel.”