People don't want auth, they want sign in with Google

The gap between thinking of building software and actually building it has never been smaller. That friend who used to message you with million-dollar ideas, begging you to build them, is suddenly one prompt away from shipping. Sure, we can argue how far he'll get with a "vibe coded" app, but that misses the point.

Developers have always been a tiny slice of the population - maybe 25 million worldwide. Now regular people can build software by describing what they want in plain English to tools that generate code for them. We're going from 25 million to potentially 500 million builders.

At Leap, I talk to hundreds of people building apps who would never touch a terminal or read API docs. They're not trying to become developers, nor are they trying to learn programming - they're builders solving problems they have.

Before you dismiss their "good enough" apps, remember: scrappy works. A basic todo app that handles payments and serves fifty users is a real business. The founder doesn't need enterprise-grade architecture. These builders represent a fundamental shift in how software gets made.

Here's what I've learned about how these new builders think: they don't want authentication - they want "sign in with Google." They don't want payment processing - they want "monthly subscriptions." They speak in outcomes, not architecture. They think in user-facing features, not technical implementation.

But our tools still assume you know what OAuth is. We design APIs for people who understand REST endpoints and database normalization. We write documentation for engineers who think in systems.

As more people gain the ability to build software, almost no tools are designed for how they actually think. We could keep telling them to learn OAuth and database schemas, but that's not realistic. Building better abstractions that code generation tools can actually use might be the better path.¹

---

¹ Yes, eventually LLMs might perfectly translate "I want monthly subscriptions" into working code. But while we wait for perfect AI understanding, we can build better abstractions.