tldr
1. AI Agents produce code faster
2. One developer can spin 100 agents increasing value
3. Companies should spend $100 per day per developer to run those agents
4. Senior devs don't like agents
5. Junior devs like agents
6. Juniors are cheaper to hire and are better with AI agents, therefore companies will hire them more
Why it's all a bs:
Let's imagine you have a 100 tasks of bug fixes and various features, and your lovely PM convinces you to use those AI agents, and let's even say that for some reason you didn't have to spend 2 weeks to write 100 detailed prompts, and you can just feed all your 100 user stories to the AI system, give it access to all the millions of lines of code of your project, and it will start a 100 agents that will do all the said tasks. Great, they are done in just a few minutes, very fast, now what? Do you have to review the code? Of course not! That will kill the whole purpose, reading code takes more time than writing it, everyone knows that. You got a whole new codebase in just a few minutes, it is merged and is running in prod.
Now you don't even know how your app works, but that's okay, you can just ask AI to explain it to you. So you check the app, all the features are there, you check the API, all good. There are 15 different features of which you have shallow understanding, but you sell them to your investors and everyone is hyped.
But at some point your user base goes from zero to eight people, and oh no, they find a bug! You try to fix it with your agents, but the debugging man, it just won't fix it, even after burning 50 bucks!
So the team decides to understand the code. After a considerable amount of time the find the issue (one of the agents created new types, instead of using existing ones) and fix it, but they also learn along the way that it's impossible to work with this code. Why? Because AI wrote "average" code. There's no problem with average code, no need in "great" code, it's just not the "right" code, because, you know, they were not building a "snake in python" nor a "todo list". So they tried to proompt AI to rewrite it the "right" way, no success. Out of necessity they decide to rewrite this code from scratch.
In the meantime the startup burnt all of its money and went bankrupt. True story by the way.
Moral: the market will soon get flooded with people who buy into sentiments of the article. It's never been a better time to *actually* learn to code.