Everyone keeps telling me SaaS is dead. So I asked a product lead at Replit, a company now worth $9 billion off this exact trend, and his answer flipped how I think about it.
The company built on the trend that's supposedly killing software
Replit lets you build with AI even if you can't code: apps, designs, slides, even videos. And people love it. They raised $250M, then $400M more, they're past 40 million users, and around 85% of the Fortune 500 use the platform. Now execs build websites, marketers write code, and sales teams build their own CRM, all inside Replit. So at their Time to Build event, I asked the thing that's been bugging me.
If models keep getting better, does SaaS just die? Do people build their own tools instead?
So I asked: does SaaS just die?
He paused. "Great question. It's a real worry in Silicon Valley. Software founders are scared." Part of me hoped he'd say yes. I've seen 100 Notion clones, maybe 1,000 CRMs with an AI layer slapped on. SaaS was finally done. Then he said the opposite.
Vibecoding isn't cheap
Vibecoding isn't cheap.
And he's right. Uber burned through its entire 2026 AI budget in four months. Some engineers spend $2,000 a month just on tokens. Building with AI feels free, until the bill shows up.
People don't want to do things themselves
Then he said: "People don't want to do things themselves. Look at DoorDash." In a single quarter, DoorDash did 776 million orders, about $25 billion in food. Every one of those customers could cook, or drive over and grab the food themselves. They pay extra anyway. Convenience beats ability. That's why people still pay for software.
It's a gold rush, if you can stay defensible
"It's a gold rush right now," he said. "So many new markets to build for." So if you've been sitting on an idea, this is the window to build it. But he left me with one rule so your product doesn't get copied to death:
Find ways to be defensible.
I'm unpacking what defensible actually means over the next few posts. Follow along so you don't miss it.
