AI amplifies programmers, not replaces them

| · @kyrylo · bluesky:@kyrylo.org · mastodon:@kyrylosilin

People who think “vibe coding” will fully replace programmers still don’t understand how software actually gets built.

Let’s be real, today’s AI is powerful, but it’s not thinking. The very fact that we have to say “AGI” proves the point. If current models were truly intelligent, we wouldn’t need a special term for the real thing.

Good software doesn’t materialize out of thin air. It emerges from requirements. Hundreds or thousands of tiny, explicit contracts the code has to honor.

Those requirements don’t come from the AI. They come from the real world: users, customers, regulators, performance constraints, security policies, legacy systems, and edge cases no training set has ever seen.

AI can’t invent human needs. It can only remix what humans have already asked for in the past. Until someone translates a messy, ambiguous, real-world problem into something precise enough to execute, nothing happens.

So yes, AI writes code faster than ever. It debugs, refactors, documents, and scaffolds like a wizard. But the moment you’re building something that actually matters and hasn’t been built a thousand times before, a competent human still has to sit in the driver’s seat.

Tools changed. The job didn’t disappear.

Programmers who understand requirements just became 10× more lethal.


Prev: