> It’s not that the agents are now producing flawless code. I spent a good 20 minutes yesterday watching one tie itself in knots trying to write a regex: first in Sed, then in Bash, and finally in Python (six times).

This sounds very strange.

I'm using Claude (Opus 4.5 via Code) every day and it's very good with regexes, sed, awk and similar bash oneliners.

We don't know what the author asked it to do, but this smells like the problem started at least several messages before that.

To author's point: code brings me joy. I'm currently learning Zig, for no reason whatsoever other than intellectual challenge and I, subjectively, like the language. I'm writing silly little programs that nobody will ever see. It's fun.

Then I switch over to a paid project, and claude[0] another task from my backlog.

There's code, and then there's code. You can find joy in some code and absolutely want to avoid coding in something else.

[0] code using Claude

> We don't know what the author asked it to do, but this smells like the problem started at least several messages before that.

Author here. It was what I assumed would be a fairly simple task, fixing some duplicate closing frontmatter delimiters.

I think the LLM took a wrong turn early on, and then just spiralled. It was morbidly fascinating watching it rabbit hole.