It's insulting to read AI-generated blog posts

Frustrated, I just throw that mess straight at claude-code and tell it to fix whatever nonsense it finds and do its best. It probably implements 80–90% of what the doc says — and invents the rest. Not that I’d know, since I never actually read the original AI-generated PRD myself.

In the end, no one’s happy. The whole creative and development process has lost that feeling of achievement, and nobody seems to care about code quality anymore.

I’ve seen exactly that. In one case, it was JPEG scans of handwriting, but most of the time, it’s a cursive font (which may obviate “handwritten”).

I can’t remember which famous author it was, that always submitted their manuscripts as cursive writing on yellow legal pads.

Must have been thrilling to edit.

For essays, honestly, I do not feel so bad, because I can see that other than some spaces like HN the quality of the average online writer has dropped so much that I prefer to have some machine-assisted text that can deliver the content.

However, my problem is with AI-generated code.

In most of the cases to create trivial apps, I think AI-generated code will be OK to good; however, the issue that I'm seeing as a code reviewer is that folks that you know their code design style are so heavily reliant on AI-generated code that you are sure that they did not write and do not understand the code.

One example: Working with some data scientists and researchers, most of them used to write things on Pandas, some trivial for loops, and some primitive imperative programming. Now, especially after Claude Code, most of the things are vectorized, with some sort of variable naming with way-compressed naming. Sometimes folks use Cython in some data pipeline tasks or even using functional programming to an extreme.

Good performance is great, and leveling up the quality of the codebase it's a net positive; however, I wonder in some scenario when things go south and/or Claude code is not available if those folks will be able to fix it.