Aravind Srinivas Predicts a Future of Custom Apps for All

magine a world where you don’t have to wait for a developer to fix your app’s bugs or add that one feature you’ve been begging for on Twitter.

Instead, you just tell an AI what you want, and it builds a custom app tailored to your exact needs—whether it’s a fitness tracker that knows you hate burpees or a finance tutor that explains things in a way that actually clicks for you.

That’s the future Aravind Srinivas, CEO of Perplexity, envisions in a fascinating conversation with Nikhil Kamath, co-founder of Zerodha.

The Rise of Personalized Apps

Aravind predicts that the next decade will be defined by personalized software.

Today, if an app doesn’t work the way you want, you’re stuck filing support tickets or tweeting complaints, hoping the developers prioritize your feedback.

It’s slow and inefficient. But with AI capable of writing code on demand, that’s about to change.

“I can have my own fitness app that knows what workouts I like, how I feel that day, and programs itself for me,” Aravind explains.

The same goes for health apps, tutors, or even something as simple as a payment-splitting tool better than Splitwise.

The catch? We’re not quite there yet.

Tools like Claude can write the code, but deploying it—figuring out where the data lives and making it seamless—still has bugs to iron out. Aravind believes someone will soon build a platform to abstract those details away, making app creation as easy as typing a prompt.

Picture this: you build an app for yourself, share it with friends, and maybe even monetize it through micro-payments or ads. A social layer could emerge where people trade custom apps like recipes. “No one’s cracked this yet,” he says, “but it’s a huge market waiting to happen.”

Tools for the Non-Tech Savvy

Nikhil, who admits he’s not a tech expert, asks for app recommendations to boost efficiency in business and life. Aravind suggests two tools that embody this AI-driven future:

Cursor: A coding assistant that helps you write software even if you’ve never coded before. Want a website? Tell Cursor what you need, and it generates the code.

Replit or Bolt: These take it a step further by not just writing the code but deploying the app for you. They’re not perfect yet, but they hint at a world where you don’t need to be a software engineer to build something useful.

Nikhil wonders if these tools could rival a pro developer’s work. “Not today,” Aravind admits. They might match a junior engineer’s output, but the best coders still have an edge. That said, he argues traditional coding skills won’t vanish—expertise in infrastructure, data storage, and debugging will remain critical, especially as AI handles the front-end fluff.

AI’s Utopian Promise—and Dystopian Shadow

Looking five years ahead, Aravind sees AI becoming as ubiquitous as the iPhone. “We’ll all have personal assistants,” he predicts, “and it won’t be a luxury for billionaires—it’ll be affordable.”

This democratization will unleash creativity: if you can dream it, you can build it. No more relying solely on someone else’s app or tool.

But there’s a flip side. Labor displacement looms large.

“You don’t need 10,000 people to build a trillion-dollar company anymore,” Aravind notes.

Big tech is already laying off staff, and new graduates might struggle to find jobs. Those who adapt—using AI to upskill—will thrive, but the transition could be messy.

“How people react to this is going to be interesting to watch,” he says. “No one really knows how it’ll play out.”

Regulation: Less on Models, More on Use Cases

When Nikhil brings up AI regulation, Aravind cautions against overreach. Regulating models themselves won’t work—people can still download and use them. Instead, he suggests focusing on applications.

One concern? Kids forming emotional bonds with chatbots, potentially worsening loneliness or mental health issues.

“That’s worth thinking about,” he says, advocating for apps that enhance knowledge rather than replace companionship. Beyond that, he’s skeptical of heavy-handed rules. “We’re still early in AI. Moving slow could cost us trillions long-term.”

On data ownership—could Indians own Indian data and charge models to use it?—Aravind thinks the internet’s global, fair-use nature will persist.

Paywalls might pop up for premium content, but he’s unsure how fees would work for models training on it versus humans just reading it.

Perplexity, he notes, sidesteps this by attributing sources rather than claiming content as its own.

A Learner’s Mindset in an AI World

Nikhil, feeling left behind by AI’s pace, jokingly offers to intern at Perplexity for free. Aravind’s response is heartfelt: “We’d be honored to host you.” But he also reassures Nikhil that physical proximity matters less now.

“The internet has everything out there, and X has the best minds talking in real-time. Spend time with AI models, see where they fail, and talk to smart people—that’s how you stay ahead.”