Everyone's Building an App Now. It's the New Starting a Podcast.

I saw a guy on Twitter last week bragging about having 153 apps on the App Store. All vibe coded. Real clients, real databases, real paying customers. He was genuinely proud.

I have two apps. DeutschWunder, an iOS app for learning German. And WunderType, a macOS writing correction tool. Two. And I can barely keep up with maintenance, bug fixes, and OS updates. Let alone new features.

So either this guy is a mass-production genius, or we have very different definitions of what "having an app" means.

Remember When Everyone Started a Podcast?

In 2020, the barrier to podcasting hit the floor. A USB mic, Anchor.fm, and you were live. 4.5 million podcasts got created. Everyone had one. Your dentist had one.

Today, only about 450,000 of those are still actively releasing episodes. That's 10%. The other 90% are digital ghosts. 72% of podcasters cited discoverability as their biggest challenge. Turns out, pressing record was the easy part. Getting someone to listen was not.

Sound familiar?

Now Everyone's Building an App

Andrej Karpathy coined "vibe coding" in February 2025. By the end of the year it was Collins Dictionary's Word of the Year. 63% of people using these tools aren't even developers. Cursor, Replit Agent, Lovable (now valued at $6.6 billion), Bolt.new. The tools are genuinely impressive.

YC's Winter 2025 batch: 25% of startups had codebases that were 95% or more AI-generated. iOS app releases jumped roughly 60% year over year. 46% of new code on GitHub is now AI-generated.

I'm not throwing stones here. I built WunderType with AI. I use Claude Code every single day. I'm part of this wave. That's precisely why I can see what's coming.

The Tools Got Easy. The Hard Part Stayed Hard.

Here's the thing nobody told me when I shipped my first app: building it was maybe 20% of the work.

There are 1.8 million apps on iOS. 3.7 million on Google Play. Over 90% of revenue goes to the top 1%. The idea was never the hard part. Distribution is. Finding users who actually need your thing, getting them to try it, getting them to stay.

None of that got easier because the code writes itself now.

That guy with 153 apps? I'd love to know the retention numbers. The monthly active users. The support ticket volume. Because shipping is a dopamine hit. Maintaining is a job.

Vibe Coded, Vibe Shipped, Vibe Broken

45% of AI-generated code contains security flaws. That's not my number, that's Veracode's. CodeRabbit found a 2.74x higher vulnerability rate in AI code compared to human-written code.

Tenzai audited 15 vibe-coded apps and found 69 vulnerabilities. Zero had CSRF protection. Zero had proper security headers. In March 2026 alone, Georgia Tech attributed 35 CVEs to AI-generated code, up from 6 in January. Real incidents: 1.5 million API keys exposed. A Base44 authentication bypass. Production database wipes.

When I read about the security disasters in vibe-coded apps, I felt a mix of schadenfreude and genuine concern. Because Claude is great at building software, but it's also great at breaking it. I've seen it in my own projects. The code works until it doesn't, and when it breaks, you need to actually understand what happened.

I Use AI to Build Apps. That's Exactly Why I'm Worried.

I want to be clear: I'm not anti-AI. I literally use spec-driven development with AI to build features. This isn't a "back in my day" rant.

But I have two apps. Just two. DeutschWunder needs iOS updates every time Apple changes something. WunderType needs macOS compatibility testing, feature requests from actual users, bug reports that require deep debugging. Dependency updates. Privacy policy changes. App Store review guidelines that shift every quarter.

Two apps. And some weeks I'm drowning.

So when someone tells me they're running 153 apps with real databases and real customers, I don't think "wow, impressive." I think: who's checking the security patches? Who's handling the GDPR requests? Who's monitoring the databases at 3am when something corrupts? Who's answering the support emails?

The answer is probably: nobody. And that's the problem.

Developer trust in AI code accuracy dropped from 43% to 33% between 2024 and 2026. Favorability toward AI tools fell from 77% to 60%. The people closest to the code are getting more skeptical, not less. That should tell you something.

The Graveyard Is Getting Crowded

The podcast parallel is almost too perfect. Millions of podcasts with zero listeners became millions of apps with zero users. But there's a crucial difference: a dead podcast costs you nothing. It just sits there. A dead app costs server time, database storage, potential security liability, and the slow anxiety of knowing real user data is sitting in something you stopped maintaining six months ago.

Gartner predicts 60% of all new code will be AI-generated by the end of 2026. That's not a future prediction anymore. We're living in it. And the graveyard of unmaintained, insecure, abandoned apps is going to make the podcast graveyard look quaint.

Indie devs are already feeling the squeeze. One report showed developers losing 40% of organic installs in two weeks just because a competitor updated their metadata. The discoverability problem that killed podcasts is killing apps too, just with higher stakes.

What Actually Matters Hasn't Changed

Distribution beats product. Understanding your users beats shipping fast. Maintenance beats launch day.

I'm not saying don't build apps with AI. I do it. I'll keep doing it. But I am saying: if you can't maintain it, you didn't build a product. You built a demo.

The vibe coding market is estimated at $4.7 billion in 2026, projected to hit $12.3 billion by 2027. That's real money chasing the dream that building is the hard part. It's not. It never was.

The hard part is still the same boring stuff it's always been. Do people need this? Can they find it? Will they come back? Can you keep it running when something breaks at 2am and the AI that wrote it has no memory of why it made that architectural decision?

I don't have 153 apps. I have two. And some days that feels like plenty.

If you're building something and want a developer who actually maintains what they ship, that's what we do.