Agent 3 as a Junior Programmer Substitute

We invited Amjad Masad, CEO and founder of Replit, fresh off raising $250M at a $3B valuation and hitting $150M ARR, to share why point solution model is dying, how his own HR person built their org chart software in 3 days (killing the need for expensive vendors), and why consumers still don't understand they're hiring employees, not buying software.

Amjad and Manny explore the resurgence of the small entrepreneur enabled by AI agents. From the YouTuber doing Upwork arbitrage with Agent 3 (45 minutes, $25, full app delivered) to the VC CFO building software for other VC CFOs, this episode maps out the future where domain experts build hyper-specific micro-SaaS instead of Silicon Valley creating billion-dollar point solutions. Amjad also gets refreshingly honest about VCs leaking his numbers twice, why management fees attract short-term thinkers, and how Replit does seasonal intensity with 12/12/7 sprints during launches but sustainable hours the rest of the year.

"Agent 3 is an employee, it's not a piece of software. So instead of paying a hundred thousand dollars for a worker, you pay a thousand bucks."

The Pricing Model That Confuses Everyone

Amjad reveals why Replit's token-based pricing creates sticker shock for consumers but makes perfect sense to businesses. A doctor saved $99,700 building an app quoted at $100K. Someone in government built an education app quoted at $10M for a few hundred dollars.

"The difficult part is how people think about value here. When people come to Replit, they come with a mindset of I'm buying a piece of software. In reality, you're hiring a piece of software."

This mental model gap between technology capability and consumer expectations is the fundamental challenge every AI company faces right now.

How Agent 3 Actually Works Under the Hood

Most people don't know this, but Agent 3 isn't a single agent. Amjad pulls back the curtain on the multi-agent architecture that makes it the most advanced system in market:

"Agent 3 is actually a team of agents. There's an agent that is called the architect that continuously monitors the architecture of the system and gives feedback to the main agent. You can see the architect saying like, I don't like this is not very secure. And then the other agent kind of arguing back with that. It was like, it's fine for now."

Users can watch these agents debate in real-time. Some love it because it modernizes their codebase. Others hate it because they're spending money while the agent refactors instead of shipping.

The 3-Day Death of Generic SaaS

Amjad shares what happened inside Replit that proves point solutions are doomed:

"One person on our very small HR team didn't find the exact org chart software that they want to use from vendors. And it was like costing tens of thousands of dollars. And she built it over three days in Replit. And that's what we use right now."

When a single HR person can build exactly what they need in 3 days instead of paying $50K/year for something that almost works, every org chart SaaS company should be terrified.

The Silicon Valley Point Solution is Dead

Amjad's most controversial prediction challenges the entire VC-backed scaling model:

"I think the idea of Silicon Valley companies being able to create these point solutions and scale some to billions of dollars of revenue, I think that's going away."

The future belongs to domain experts building hyper-specific software. A CFO at a VC firm built software specifically for other VC CFOs. Someone who worked HR in hotels can now build HR software for hotels in Istanbul.

"If you have domain knowledge in your head, you can put that into a piece of software and sell it. Maybe that's like a $10 million business, but it is not like a think that's going to IPO."

From Upwork Arbitrage to Micro-SaaS Millions

Amjad maps out the new entrepreneurship ladder enabled by agents:

"We're seeing a YouTuber, name is Nick Conley, after Agent 3 came out, went to Upwork, copy pasted an entire job, put it into Replit at 45 minutes, $25, and he had a full app. Now he could kind of go back to Upwork, sell it back for a couple thousand dollars."

That's level one arbitrage. Level two is building defensible micro-SaaS using your domain knowledge. Level three is making small businesses radically more efficient without hiring.

The Societal Vision: Fewer Billionaires, More Millionaires

Amjad sees a fundamentally better economic structure emerging:

"I think there's less billionaires and a lot more millionaires. I think that might be a good thing. The conglomeration of the world where things are centralizing into a few big companies that are government sanctioned monopolies, I think created a lot of societal ills."

More small businesses, more freedom over where and how people work, more people getting rich on smaller scales.

Why VCs Keep Leaking His Numbers

Replit's revenue numbers leaked twice. Both times from VCs. Amjad explains the systemic problem:

"I think the VC market has become so lucrative, especially lucrative on the management fee side. So you can be a shitty VC and make a lot of money. And so you have a lot of short-term thinkers that are going in and I don't know what's their incentive to be leaking these things. Maybe they're trying to carry a favor with a journalist."

Silicon Valley's trust culture is eroding because the incentive structure changed.

The Seasonal Intensity Model That Actually Works

Amjad rejects both 996 and anti-hustle culture, offering a third way:

"I think we've always taken the view of seasonality. Like I think even in early human history, you have the season in which you're like planting the crop, the season when you're harvesting. Especially around the major releases, we ended up doing a lot more than 996. We ended up doing 12/12/7. But for the most part, it's a lot more sustainable."

During Agent 3's launch, the team went all-in. The rest of the year, people pick their schedules, go home for dinner, and work flexibly.

Which SaaS Companies Will Survive

Amjad draws a clear line between winners and losers in the AI transition:

"Those systems, I think have very lasting advantages just because this is where all the data is. There's network effects, there's developer mind share. They need to start thinking of themselves as like more of an open system, like MCPs and being able to be a platform where other agents can sit on top of."

System of record players like Salesforce survive only if they open up. Point solutions die as companies build custom. Closing APIs (like Salesforce just did with Slack) is the fatal mistake.

The Call to Action

"Jump head first, like don't think too much about it. I think the future really benefits those people who are more action oriented."

Companies Mentioned

• Replit
• Cursor
• Devin
• Lovable
• Bolt
• Retool
• Zapier
• Make
• Figma
• OpenAI
• Salesforce
• Workday
• Upwork
• Klarna

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.