Eniko is correct and eloquent and sweary in the thread.
It writes code approximately as good as Stack Overflow examples.
Which scales about as well as that would.
You only run into trouble when you want more than a few lines of code written to illustrate a point; and instead want it to be effective, scalable, secure, legible, maintainable and efficient.
It is reviewing the pull requests at work, and that is definitely catching bugs. Some of them even important ones. It is also definitely wasting time on chasing non-bugs.
It is possible the ridiculously detailed error coverage in the code is even better code, in some sense. If it matters *how* you crash and what error you report when you do so and you don't mind reams and reams of error-handling.
It may improve. In my job and profession I think I do have a responsibility to be aware of what it can and can't do. If only to fight the people who seem to want to pipe it into root on the prod servers, just for upgrades.
And to know when or if it becomes good.
Which means using it.
I've worked with worse coders. It's very fast. But those coders then improved.
We'll see how well Claude can generalize from textbook examples to effective, scalable, secure, legible, maintainable and efficient over time I guess.
I wouldn't rule it out, but there isn't nearly as much training data.
#ai #programming
It's like everyone who came on board to AI because "they can write code now!" thinks the primary argument the rest of us have is "no it can't"
So I'll be the one to say it: yes AI can write code. I believe you. They can write code. It compiles. It even passes the tests that the AI also wrote.
Unfortunately the code they write is shit. It's automated technical debt with built in landmines. You think people coding in C causes issues? Wait'll you see what LLM code produced on massive scale does