Developer Claims AI Cut App Build Time From Hours to 13 Minutes

So apparently we're at the point where you can build an entire SaaS application—with authentication, a database, and a full design system—in under 15 minutes. At least, that's what developer Hamish from Income Stream Surfers demonstrated in a video that's equal parts impressive speed-run and thought experiment about how we're actually using AI tools.

The headline number is wild: what used to take him 2-3 hours to set up now takes 13 minutes. But the more interesting part is how he's doing it, because it's not what you'd expect.

The Stack and the Speed

Hamish's workflow chains together several tools that feel almost purpose-built for this exact use case. He starts with Shadcn for generating design systems—literally just clicking a button until he finds color combinations and fonts he likes. Then he imports that into V0 (Vercel's AI design tool) to generate actual page layouts. For authentication, he's using Clerk. For the database and backend, Convex. And Claude Code ties it all together.

The demo moves fast—he's literally racing against a self-imposed 10-minute timer—but the basic flow is: generate design, set up auth, create database, use Claude to wire everything together, download the code, implement the design. "We're about, you know, 5 minutes into the project so far," he says mid-video, already looking at a working dashboard with user authentication.

Is it production-ready? That's... a different question. But as a foundation? It's genuinely functional. He shows a real client project (quot.info) that started this exact way: "You can see it's 99% the same. There are some changes just because I've since developed more and more, but you can see what I'm talking about, right?"

The Part Everyone's Missing

Here's where it gets interesting, and where Hamish's actual point often gets lost in the "omg 10 minutes" framing: he's not letting AI do the work. He's doing the work with AI helping.

"The main thing that's changed is instead of trying to get AI to do it, I've just kind of let I—I've done it myself with AI helping me instead of just saying just do this for me," he explains toward the end.

This matters more than the speed. The developer discourse around AI tools tends to split into two camps: people who think AI will replace developers entirely, and people who think AI tools are overhyped garbage. Hamish is demonstrating a third option that's way less dramatic but probably more predictive: AI as an incredibly fast junior developer who needs specific direction.

He knows exactly what he's building. He's following documentation he's memorized through repetition. When he tells Claude to "please make sure user accounts are created in convex and this app is set up for multi-tenancy," that's not magic—that's him knowing Claude consistently misses that step and preemptively fixing it.

"This is a really really common error with Claude," he notes. "Interestingly, [Google's] Gemini 2.5 did not make this mistake and instantly set up multi-tenancy." That's the kind of pattern recognition that only comes from doing this repeatedly, understanding where the tools fail, and developing workarounds.

What Actually Speeds Things Up

The technical stack matters less than the methodology. Shadcn ensures mobile-friendly, fast-loading designs without manual testing. V0 generates React code that drops directly into Next.js projects. Clerk handles authentication complexity. Convex provides database and backend infrastructure. Claude stitches it together.

But the real accelerant is something older: standardization. Hamish can move this fast because he's building the same type of thing repeatedly with the same tools. He's created a process. The AI just automates parts of that process that used to require typing.

He even mentions that Stripe integration—typically a multi-hour headache—is "actually a lot easier than I thought it was" once he realized he doesn't need to use Stripe products at all. "You can send a price. So, let's say they select five different things and the total price is $1.99. Instead of using Stripe product, you can dynamically just send whatever this amount is for them to pay in a Stripe checkout. It's literally that easy."

That's not AI magic. That's just him finally reading the documentation and realizing there was a simpler way all along.

The Design Question

"Yeah, you might say this is AI generated, whatever," Hamish admits while looking at his V0-generated homepage. "The really good thing about this is it's going to be mobile friendly just off the bat because it's using Shad CN, right? So you don't actually have to worry about any of the mobile friendliness, anything like that."

He's acknowledging the obvious tension: these designs look... fine. Generic, maybe. Not offensive, not inspiring. But they work, they're accessible, they load fast, and they don't break on mobile. For many projects, that's actually sufficient.

The interesting question isn't whether AI can generate beautiful designs—it mostly can't yet, at least not consistently. The question is whether the baseline quality is good enough that custom design becomes a second-stage concern rather than a prerequisite.

For client projects where the value is in functionality rather than aesthetic differentiation, that trade-off might increasingly make sense. For consumer-facing products where brand and design are the product... probably not.

What This Means (And Doesn't)

This isn't a "developers are obsolete" story. Hamish clearly knows what he's doing—he's just doing it faster. He's making architectural decisions, catching AI mistakes, understanding multi-tenancy requirements, knowing which documentation to follow.

But it's also not a "AI is useless" story. The tools are clearly accelerating his workflow in meaningful ways. The 2-3 hour baseline he mentions isn't nothing.

What's actually happening is more subtle: the bottleneck in web development is shifting. Less time spent on boilerplate setup means more time available for the parts that actually differentiate products—the specific features, the business logic, the user experience details that generic tools can't generate.

Or, alternatively, it means one person can now build multiple projects in the time they used to spend on one. Which has its own implications for the economics of software development.

Hamish hasn't solved web development. He's found a workflow that works for his specific use case—building similar SaaS applications repeatedly with common patterns. The tools enable that workflow. The workflow shapes how he uses the tools. It's a loop, not a revolution.

But 13 minutes is still 13 minutes. That's the kind of efficiency gain that changes what's possible to build solo. Whether that's ultimately good or just means more mediocre SaaS products flooding the market... well, we're about to find out.

—Zara Chen