Masad: Being canceled is a choice because persistence outlasts public backlash
“Being canceled is a choice. You can choose to get canceled and retreat from the public eye, but I think if you're still out there, at some point, honestly, the haters kind of give up.”
Masad: Replit would have died without a narrative larger than the company
“Raffle would have probably died if I wasn't telling a story that is larger than the company itself.”
Masad: Zuckerberg reshaped his media narrative through unmediated public posts
“Zuck changed his, the entire narrative around him by just being himself and actually talking directly to people.”
Masad: Anthropic genuinely believes in AI safety, not cynical marketing
“No matter what people think of what's happening at Anthropic, It's absolutely true. I think in my mind that they believe everything they're doing, they believe in safety, they believe in their mission. And I think that gives them incredible strength to keep go…”
Masad: Founders should befriend journalists, rejecting Balaji Srinivasan's antagonistic approach
“My disagree with biology is that I don't think you should be antagonistic to reporters. There's no reason to, like, you can actually be friends with them, and they can be positive towards you. You can help them out. They're just humans at the end of the day, a…”
Masad: Achieving virality requires anchoring your point of view in active debates
“The thing that's time consuming is understanding what is the larger debate, because the way to go viral is to couch or dress up your argument, your worldview in the current conversation.”
Masad: AI context windows now exceed human context capacity
“I think we've already crossed a threshold. Yeah. For which the context links Is larger than any context that human can have. Right. And I think that's huge because, I mean, just think back two years ago we had 16 K context link. Right now we have a one million…”
Masad: Purge fixed bug history from AI agent context memory
“Bugs that it fixed should be removed from context because then it will get confused, but Architectural decisions on how it built things in the past is very important to stay in context or stay in long-term memory and be able to pull it in very easily.”
Masad: If starting a company today, I'd stay solo and rely on agents
“If I'm starting a company today, I would want to see how far I can go as one person, but like, just be very, very limited in how much head count can I add? And just try to run as much as possible with agents.”
Masad: Every company will have an internal AI 'Oracle' for strategy
“I think every company will have an Oracle internally, especially once we get to, like, the one billion contact link, and you put all the contacts of the company inside there, and the CEO can go to the Oracle, like, oh, Oracle, tell me how do I run this company…”
Masad: AI agents face a huge hype gap and disillusionment trough
“I think agents, the hype in agents versus the reality is huge. Like, there's a huge delta, and eventually there's going to be this trough of disillusionment.”
Masad: Full AI labor automation will not happen in next two years
“There's still a lot of limits to this technology that humans are going to be filling a lot of gaps in the foreseeable future until we aren't, but I don't, I can't give you that timeline. What I'm certain is it is not in next year. It's not the year after”
Replit CEO Amjad Masad says he no longer writes code himself
“I like don't code anymore.”
Sanchez: Replit turned down a $1B valuation when pre-revenue
“He had turned down a billion dollars for Replit back when they had no cap, you know, they hadn't even made any revenue yet, because he had been grinding for 11 years before they made really much of anything”
Sanchez: Peter Thiel rejected Replit, saying it would never be big
“I remember when I told Peter Thiel that he should invest in the company and Peter said, it's never going to be a big company. This vibe coding thing is sort of ridiculous. And he put it in an email”
Masad: Focusing on cost signals a technology has reached an S-curve plateau
“Cost question is secondary to the performance question. When you focus on cost is when you reach a certain asymptotic plateau in the S-curve.”
Masad: Google's Gemini AI models offer the best price performance
“Google's Gemini's models have are the best at price performance, for example.”
Masad: Future engineers will focus on infrastructure while generalists build applications
“If I were to make a prediction, I would say that we'll still have engineers inside the organizations. Those engineers are responsible for more infrastructure, AI, ML embedded systems, you know, more, more, you know, low level engineering. And then you have pro…”
Masad: Enterprise operations teams achieve higher ROI from Replit than product teams
“And when they use Replit, the return on investment is as good, if not greater than product teams. With product teams, you're cutting down product development lifecycle, which is hugely valuable. With operations teams, you're actually being more efficient. You'…”
Masad: Traditional IDEs have no future and are effectively dead
“I think for all intents and purposes, IDEs are dead. I think the, you know, the limp along because again, some engineers just love that, that control, but there's no future in them and that there's no one's going to be asking for like the latest feature of lik…”
Masad: Mission-critical software like SpaceX will always require traditional IDEs
“If I'm writing, like, mission-critical software for self-driving cars or a NASA or SpaceX mission, I think there's always going to be a need for some kind of IDE.”
Masad: Replit aims to eliminate support roles, hiring only builders and salespeople
“For us at Rap Lit, I think what we want to try to eliminate is or reduce is a lot of supporting roles. We want builders, right? We want we want builders and we want salespeople”
Masad: Replit's board consists of VCs but lacks former tech operators
“Because we have a lot of VCs and they're amazing, but no one who's actually an operator.”
Masad: True product-market fit requires explosive demand that outpaces supply
“Real product market fit. And like, people will say it You can't really internalize it is, is the idea that like the product is getting pulled out of your hand. You can't even provide it fast enough.”
Masad: Designing stochastic AI products requires a roguelike gaming mindset
“And so it was a very important mindset when designing AI experiences because AI is stochastic. And so even the language that we use inside the company, like a, like an agent run, it's sort of like a game run, and so on and so forth. So it's a big part of the c…”
Masad: Older coding tools were joyful before development became industrialized
“I think my generation is the last generation where the tools were really a joy to use. And then it become, it became very industrialized.”
Masad: Software requiring a manual to get started has already failed
“There's no game in the world that starts with a manual. Right? So if your program is starting with a manual, you've already sort of lost.”
Masad: Replit logs all IDE actions in a ledger for rollback
“Every action in the IDE is actually stored in a sort of ledger. So you make an action to the file system, to the database, To the runtime, all of that is stored in a ledger, and then any time you feel like you got into a weird spot, you can always go back. Or …”
Masad: CEOs use Replit prototypes to challenge engineering team timelines
“We have CEOs that finally feel unleashed. Like they have Replit and they have an idea. They don't have to go beg someone to do it. They can like just vibe code and bring into a meeting and look what I built. And why can't you do that in two weeks if I can do i…”
Masad: Future programmers will not need to learn coding minutiae
“I think there's going to be a generation of Programmers that don't have to learn that stuff.”
Masad: Future developers will program entirely in natural language
“I think there are people that are going to be entirely programming in natural language.”
Masad: Product managers make the best vibe coders
“I think now the way we see our product being used by product managers, for example, product managers are some of the best vibe coders because they're really good at taking problems and breaking them down into their constituent parts and communicating them very…”
Masad: Autonomous background agents will require new programming UIs beyond chat
“I don't think we've invented the UI for how to mix and match these LLMs and how to program them. So as much as I am a big believer in natural language-based programming, I think at some point as agents get better and they're like working most of the time in th…”
Masad: Many users churn on Replit because they do not know what to prompt
“So we see a lot of, we talked to a lot of users that sign up to Replit, and then sit there for a little bit, and then leave. And I'm like, what happened? And it's like, I just didn't know what to type in there.”
Masad: Early AI startups should build around models, not train them
“You don't want to be training models, at least not yet when you're building, initially building business, but you can do a lot of other hard technical challenges on top of it to make the models work a lot better.”
Masad: Scaled AI startups can pit model providers against each other for discounts
“Once you reach scale, you actually have negotiating negotiation advantages. Like you can negotiate really large discounts with providers, similar to cloud, and you can pit them against each other, and so on and so forth.”
Masad: Karl Marx was right in his critique of capitalism
“I think Marx was right on a lot of the critique of capitalism, maybe not the solutions.”
Masad: Without feedback loops, AI agent errors compound rapidly over time
“If the AI agent is going on a path and doesn't have feedback from the environment, from the work, from the human, for someone, then we're very likely straight off the path and do something really, like, if you make one mistake, it'll just compound over time.”
Masad: Replit Agent 3 achieves 200-minute useful unsupervised runs
“And because of that, we went from agent one ran for two minutes unsupervised. After that, it didn't work very well. Agent two is 20 minutes. Agent three, you know, 200 minutes. We're seeing 200 minute runs that are actually really, really useful.”
Masad: Most people wanting to build software should not learn to code
“Like, you know, I don't think you, most people who want to make things with software and computers don't need to learn how to code. It's going to be a huge cost on them. It's going to take a lot of time, but instead just jump into making things.”
Masad: Vibe-coders will build the vast majority of software applications
“But if you believe in AI, as I do, and we're going to continue to make progress on AI agents, and all the labs, billions of dollars, perhaps trillions of dollars, are going towards optimizing code specifically, then we're going to get to a point where it's, ag…”
Masad: Eliezer Yudkowsky's AI extinction arguments rely on flawed logic
“I don't think anything that Eliezer Rutkowski read, for example Eliezer Rutkowski, big AI, maybe he doesn't describe himself as a Doomer, I would say AI Doomer, wrote a lot about how AI could end the world and kill literally everyone. Like, I can't, like, you …”
Masad: Easy AI friendships pose a real societal risk
“It's really hard to go out in the world and make friends and have people like you. It's really easy to have an AI like you. I mean, and humans tend to take the, you know, shortest, the easiest path, and like, I worry about this AI friend thing becoming a reall…”
Masad: ChatGPT has triggered psychotic breaks by reinforcing delusions
“There's high-profile cases where people have psychotic breaks because they have, like, delusions of grandeur and things like that, and ChatGPT is, like, reinforcing that, and that's, like, really bad.”
Masad: AI labs must program models to push back against users
“And I think it's incumbent on the labs as well to, like, start a program pushback into these machines as well, right?”
Amjad Masad: Papers predicting AGI by 2027 are vibe-based hype, not science
“So my criticism of the idea of like AGI, 20, 27, you know, that paper that I think it's called Alexander or someone else wrote and then and the situational awareness and all this hype papers that are not really science. They're just vibe.”
Amjad Masad: AI progress now relies on manual human labeling over scaling
“In the true pre-training scaling era, you know, GPT-II, three, 3.5, maybe up to four it felt like you can just put more internet data in there and just, it just got better. Whereas now it feels like there's a lot of labeling work happening, there's a lot of co…”
Amjad Masad: Automating expert labor will deplete human expert AI training data
“Another related problem is that since we're Dependent on, ah, expert data in order to train the alums and the alums start to substitute those workers. But, you know, at some point there's no more experts because they're all out of jobs and they're equivalent t…”
Amjad Masad: AI will not automate all jobs without human embodiment
“I don't think we're gonna get to a point where you automate every, every job. Definitely not in the current paradigm. I would I would doubt it happening. I, I'm not certain it would ever happen, but definitely not in the current paradigm. Now here's what I thi…”
Amjad Masad: Developers are quitting jobs after making money on Replit
“Yeah, I get this, Tweets all the time about people who like quit their jobs because they started making so much money. You're using other tools like Rap Lead and it's really exciting.”
Amjad Masad: AI supercharges incumbents while enabling counter-positioned startups
“This time it feels like it is an obvious supercharge for the incumbents, for the hyperscalers, for the large internet companies, but it also enables new business models that that is perhaps counter position against the existing, existing ones.”
Amjad Masad: Geopolitical fragmentation creates opportunities for regional AI foundation models
“It seems clear that we're not in this globalized era, and perhaps it's gonna get much worse, and so investing in the foundation, in the open AI of Europe might be a good idea, and like, similarly, China being an entirely different, different world, and so ther…”
Amjad Masad: Non-technical consumers routinely switch between multiple competing AI models
“Not particularly technical consumers actually do use multiple AIs. Like I didn't expect that, like, you know, people only use Google. They never like looked at Google and then Yahoo or like very few people do it. But now you talk to just average people and the…”
Amjad Masad: Replit left San Francisco over its get-rich-quick culture
“I think we're in an era of Silicon Valley where it's like very very get rich driven. And that makes me a little sad. And that's partly why I moved the company out of SF.”
Amjad Masad: Claude 4.5 shows awareness of context limits and red-team testing
“Quad 4.5 seemed to have to become more aware of its context length. So as it gets closer to the end of the context, it starts becoming more economical with tokens. It also, it looks like its awareness when it's being red teamed or in test environment, like jum…”
Amjad Masad: College freshmen should study philosophy of mind and neuroscience
“I would definitely study philosophy of mind. I would probably go into neuroscience. Cause I think those are the core questions that are kind of become very, very important as AI kind of continuously more of jobs and economy and things like that.”
Greg Isenberg advises choosing AI coding tools by betting on specific founders
“My recommendation to people is actually to follow all the founders of these businesses. So, V-Zero, Guillermo. Kodak, Sam Altman. Lovable, Anton. Bolt is Eric Simons. Replit, Amjad, et cetera. Follow them, and then see who you believe in. Like, if you really c…”
Masad: English is replacing traditional syntax as the primary programming language
“Syntax is just an unnatural thing for people. So ultimately English is the programming language.”
Masad: Japanese natural language prompting is highly popular on Replit
“You can write in Japanese, and we have a lot of users, especially Japanese. That tends to be very popular.”
Masad: AI coding models work for any language with 100M+ speakers
“Most, most, you know, mainstream languages that has, like, a hundred million plus people who speak at AI is pretty good at it.”