rvachev.orgEN / RU / 🤖
← Back to essays
· Essay · 1 min

Prototypes and Understanding Code in the Age of Neural Networks

Tools are evolving at an incredible pace, but a fear arises that we won’t understand how the created code works.

Interesting times: anyone can build a prototype of a decent website, mobile app, or service — and it will work more or less.

Tools are evolving at an incredible pace. The quality will only improve.

But there’s another side. Looking at the code generated by neural networks, I realize there’s a lot inside that people who vibe-code don’t understand at all. And a fear arises: will we reach a point where no one understands how it all works?

I started with C/C++, then PHP, then Python. In all these languages, I understand the basic concepts, can quickly glance at the code and understand what’s happening. I can catch it in time if something goes wrong.

But neural networks have allowed programming in languages I’ve never written myself — TypeScript, React. I started something, but didn’t have enough time to grasp the concepts. After all, new languages are created for a reason — they have conceptual changes that didn’t exist before.

The project I’m currently working on is built on React on the frontend. And I no longer understand in detail how everything is arranged there, why exactly those patterns were chosen.

There’s a feeling that as humanity, we will come to a situation where everything we create, we don’t fully understand from the inside. And that’s scary.

On the other hand — practically none of us know Assembler and don’t understand how the processor works. And that doesn’t stop us from using computers.

It’s interesting to see where all this will lead. What do you think?

Newsletter

Get new posts by email

A weekly digest of what I write about AI, software and products. Unsubscribe anytime.