Programming keeps getting easier. AI can now write code from plain English. Every tech newsletter promises: "Soon, everyone will build their own apps."
This is Silicon Valley talking to itself.
Step outside the echo chamber. Most people don't want to code. They never did. AI won't change that.
1950s: Assembly language replaced punch cards. Experts predicted mass adoption.
1980s: C made programming "accessible." Same prediction.
2000s: Ruby on Rails promised "convention over configuration." Same prediction.
2020s: No-code tools like Webflow. Same prediction.
2024: AI agents write code from prompts. Same prediction.
Each wave lowered the barrier. Each wave flopped at mass adoption.
The bottleneck was never syntax.
Most brains don't think in systems.
Building software requires decomposition. You see a problem. You break it into parts. You model state, handle errors, plan for edge cases.
This isn't coding. This is logic architecture.
When your Uber app crashes, you don't debug the API. You delete and reinstall.
When your spreadsheet formula breaks, you don't fix it. You Google a template.
That's not laziness. That's specialization.
Most people want solutions, not tools to build solutions.
You don't grow your own food. You don't generate your own electricity. You don't sew your own clothes.
Why would you code your own software?
Division of labor exists because it works.
High agency is rare.
Building software is grinding. Your first ten attempts fail. You spend three hours on a typo. The error messages make no sense.
Some people push through. Most quit.
AI doesn't fix this. It just gives you broken code you can't debug.
GitHub has 100 million users.
Earth has 8 billion people.
That's 1.25% of humanity. And most of those accounts are inactive.
The real number of people shipping code regularly? A rounding error.
Three groups:
High-agency builders who get dopamine from creating things from scratch. They code because they love it.
People with specific motivations. Founders who need custom tools. Researchers who need bespoke analysis. Hackers who want total control.
Professionals getting paid.
That's it. AI might grow this group from 1% to 2%. But it won't flip.
If you're building a product: Don't assume users will customize it themselves. They won't. Build it complete.
If you're betting on AI coding tools: You're betting against human nature. The market is founders and engineers. Not everyone.
If you're a founder: The opportunity isn't empowering everyone to code. It's building software so good that no one needs to.
The other 98% will use what the 2% builds.
They'll use ChatGPT, not build LLM wrappers.
They'll use Notion, not code databases.
They'll use Spotify, not script music players.
This isn't failure. It's how civilization works.
The future isn't everyone building.
It's better builders serving everyone else.