"Learn to code."

For the past decade, that was the universal advice. Want to build a startup? Learn to code. Want a better career? Learn to code. Want to understand the future? Learn to code.

It was well-intentioned. It was also gatekeeping disguised as empowerment.

The implication was clear: until you've spent months or years memorizing syntax, understanding frameworks, and paying your dues in tutorial hell, you don't get to build things. You're not one of us.

That gate just swung open.

What Actually Changed

In 2024-2025, AI coding tools crossed a threshold. They went from "interesting but unreliable" to "I just built a working feature in 20 minutes."

The shift isn't incremental. It's categorical.

Here's what's different now:

AI can translate intent into code. You describe what you want in plain English. The AI writes the implementation. Not pseudocode. Not suggestions. Working code that runs.

The bottleneck moved. It used to be: can you write it? Now it's: can you describe it? The limiting factor is no longer syntax knowledge—it's clarity of thought.

Iteration became instantaneous. "That's not quite right, make it do X instead" gets you a new version in seconds. The feedback loop that used to take hours now takes moments.

I'm not theorizing here. I built this website—the one you're reading—with AI assistance. Not as an experiment. As a production site that needed to work.

What You Still Need

Let me be clear: "you don't need to learn to code" doesn't mean "you don't need to learn anything."

You still need mental models. How does a web request work? What's a database? How do the pieces of a system connect? You don't need to implement these from scratch, but you need to understand what they are.

You still need debugging intuition. When something breaks—and it will—you need to sense where the problem might be. Not fix it yourself necessarily, but describe the symptoms well enough that AI can help.

You still need domain knowledge. The most important question isn't "how do I code this?" It's "what should I build?" That requires understanding your users, your market, and the problem you're solving.

What You Don't Need

Here's what you can skip:

Memorized syntax. Nobody needs to remember whether it's array.push() or array.append(). The AI knows. You don't have to.

Years of tutorials before shipping. The old path was: learn fundamentals, build toy projects, eventually attempt something real. The new path is: build something real, learn what you need as you go.

A CS degree or bootcamp certificate. These still have value for certain career paths. But for building products? They're no longer prerequisites.

Permission from the developer community. You don't need to identify as a "real developer" to ship real software. More on this in a future post.

The New Path

If you want to build something today, here's what works:

  1. Start with a real problem. Something you actually want to exist. Not a tutorial project—a thing you'd use.

  2. Describe what you want. In plain language. Be specific about what it should do, how it should look, what constraints matter.

  3. Let AI translate. Use Claude, ChatGPT, Cursor, or whatever tool fits. Get a working first version.

  4. Iterate by describing changes. "Make the button blue." "Add a search field." "Show an error if the email is invalid." Each description gets you closer.

  5. Learn concepts as you need them. When you hit a wall, learn just enough to describe the problem better. Just-in-time learning, not just-in-case.

This isn't a lesser path. It's a faster one.

Who This Threatens

Not everyone is happy about this shift.

Traditional education gatekeepers built their value proposition on being the only path in. If you can build without their curriculum, what are they selling?

"10,000 hours" purists believe mastery requires suffering. If someone ships a useful product in a weekend, that violates the rules of the game they spent years playing.

Syntax snobs confuse fluency in a programming language with the ability to solve problems. They're not the same thing.

These reactions are understandable. Having your hard-won expertise suddenly become optional is disorienting. But the tools don't care about job titles or credentials. They just work.

The Honest Caveat

There are still domains where deep expertise matters.

Security requires understanding attack vectors, threat models, and defensive patterns that AI can't fully substitute for yet.

Infrastructure at scale involves tradeoffs and failure modes that need experienced judgment.

Performance-critical systems demand understanding of what's happening under the hood.

If you're building the next Stripe or AWS, yes, you need serious engineering depth.

But if you're building a SaaS tool, a mobile app, a website, a workflow automation, an internal tool for your company—the bar has dropped dramatically.

The Permission Slip

Here's what I want you to take away:

You don't need to spend six months learning before you can start building.

You don't need anyone's approval to call yourself a builder.

You don't need to follow the path that made sense in 2015.

The tools have changed. The gatekeepers lost their keys. The only thing stopping you now is the belief that you're not allowed.

You are.

Start building.

Continue Reading

This post is part of a series on the new rules of building in the AI era: