I keep seeing developers complain about LLM-generated code, that it's low quality, that codebases will decay, that soon the world will be flooded with junk and applications will start to break.
I don't get it. You've always been free to write bad code, this hasn't changed. If you always cared about quality before, why would you change now? When you copied code from the internet, did you care about its quality? If yes, why wouldn't you care about the code LLM generated for you?
Who are the people who always cared about their code quality and now think "since this code was LLM generated, I can deploy it as shitty as it is"?
What I find funny is when people say things like, "I'm using agents, but I review all the code!" Why do you feel the need to clarify that? It's still your code. You don't preface a pull request with "I wrote this by hand, but don't worry, I reviewed it!"
Ask yourself: why do you write quality code in the first place? Is it because it is easier to maintain, because it breaks less? If so, why would you not care about LLM-generated code? Do you stop caring about breaking your application because the LLM generated it?
If the code you commit causes any problems in production do you think it is reasonable to say "ohh, but that was Claude's code, not mine"? It is still your responsibility for everything you commit.
One exception is coding prototypes, PoCs, etc. because in that case I have lower quality standards, but I already had lower standards for my own prototype code, so this doesn't change either.
Those who cared about their code quality, if written by themselves, copied from elsewhere, reviewing other people's code, will still care about the quality of the code generated from LLMs, they will review, refactor or reject it to meet their quality standards, and those who didn't care (or didn't have the experience yet to know the difference between good and bad code) will continue to do so with LLM-generated code.
In summary, I don't think LLMs change the quality of programmer's code that much. Maybe LLMs will lead to a greater volume of low-quality code. That's possible. They let non-programmers ship apps that would've taken months to learn how to build. But that's not a disaster, it's democratization. And honestly, the code those newcomers generate with an LLM is probably better than what they'd write after a "Learn to Code in 30 Days" crash course.
At the end of the day you control what gets merged and deployed. If you'd reject a piece of code you wrote yourself, why would you accept it just because a model wrote it? Agents are just tools and your personality is what will drive how you use them, if you care about code quality, LLMs won't change that.