Amjad Masad

Co-founder and CEO of Replit

Amjad Masad is the co-founder and CEO of Replit, and he says all app software will go to zero. On 20VC in April 2026, he agreed the drop in SaaS stocks is justified. Big companies still keep Salesforce and Workday and build their own tools on top, he said, while narrow tools that do one job get replaced. In his view, ops teams get as much back from building tools as product teams do, or more.

20VC 00:04 19:29 20:12 20:22 21:00 21:20 Y Combinator 15:55

Amjad Masad's key ideas, in short

  • Product strategy: Pick an industry you know and automate its slow work. My First Million, December 2024 54:13
  • AI coding: Levels of AI Autonomy
  • Scaling: Let agents do the busywork that sank teams with no managers. a16z, September 2026 24:50
  • Product strategy: Build for a problem you feel yourself. My First Million, December 2024 07:07
  • AI coding: A small model run many times can beat one large model. Y Combinator, July 2025 08:58

By , founder of Gavel

Updated

Stuck on a real call of your own? Gavel answers startup questions with what a named founder did, and links to the minute they said it. Ask Gavel free

Frameworks
1
Problems
3
Ideas
9
Recordings
6

The SaaS apocalypse: his calls on what to build and what to buy

In three interviews from July 2025 to April 2026, Masad makes the same bet: the software companies buy today will get cheap enough to build. Each call in the table links to the minute he said it.

What this means for what you build

Amjad Masad's calls on SaaS, July 2025 to April 2026
Y Combinator Jul 2025 at 34:10 A worker was quoted $150,000 for a tool. He built it on Replit for $400 and sold it to his employer for $32,000.
“they got quoted $150,000 for a piece of software”
Y Combinator Jul 2025 at 34:23 He sees software with a developer community and plug-ins as safe.
“Companies that have a platform developer community around it”
Y Combinator Sep 2025 at 15:55 He predicts that apps of every kind will become close to free, over a span of years.
“My prediction is that all application software will go to zero.”
Y Combinator Sep 2025 at 17:01 A business can replace maybe 15% of the software it buys today, he says. He expects that to reach 100% in a few years.
“maybe 15% replaceable to 100% replaceable”
20VC Apr 2026 at 19:01 Operations teams buy a lot of SaaS and dislike it, because each tool keeps its own data apart.
“typically they'll buy a lot of SaaS software. They're typically not happy with it”
20VC Apr 2026 at 19:29 In his view, operations teams get at least as much back from building tools as product teams do.
“the return on investment is as good if not greater than product teams”
20VC Apr 2026 at 20:12 He sees big companies keep Salesforce and Workday and build their own tools on top, through those tools' APIs.
“We're not seeing people rip out Salesforce, Workday”
20VC Apr 2026 at 20:38 Some teams skip SaaS and build on their own data warehouse.
“the system of record is actually your data warehouse”
20VC Apr 2026 at 21:00 The host said SaaS stocks fell for a fair reason. Losing 20 to 30% of buyers to the data warehouse is enough to slow growth. Masad agreed.
“agree with that. I would agree with that.”
20VC Apr 2026 at 21:20 Narrow tools, such as survey software, are being replaced outright with Replit.
“we see that getting replaced wholesale”
20VC Apr 2026 at 23:10 Software you build needs upkeep. Replit spends as much on keeping code working as on writing it.
“we spend a lot of tokens on maintenance as much as we spend on creating that software”
20VC Apr 2026 at 24:40 Say an operations manager saves $10,000 on SaaS and $200,000 on hiring. Then $1,000 more for safe code is an easy yes. He puts the return for his clients at a hundredfold.
“you're an operations manager using Replit, and you just saved $10,000”

What other operators say

Key ideas

Amjad Masad's advice for founders deciding what to build, sorted by the problem it solves. Each quote names the speaker and links to the minute it was said.

Product strategy

When AI changes what's worth building

AI coding

When your AI-built app has to work

Shipping

Levels of AI Autonomy

Grade a coding tool by how long it runs before it needs you. Level one is a language server, level five is a thousand agents.

the state-of-the-art code assist was a language server, right? That's IntelliSense if you're using VS Code. And you can think of it as level one autonomy.
Y Combinator · The Future of Software Creation with Replit CEO Amjad Masad Watch at 09:21
Read the framework

Scaling

When your team is growing

Disagreements

Does climbing the ladder retire the engineer?

Amjad Masad says apps will get close to free as agents do more of the work alone. Then the hard part is knowing what to build, so people who can do many jobs win.

Michael Truell says building apps by asking an AI is not enough to grow as a programmer. Coding and math still matter, and taste pays off only when you know the tech under it.

From Levels of AI Autonomy

Sources

Stuck on a real call of your own? Gavel answers startup questions with what a named founder did, and links to the minute they said it. Ask Gavel free