Writing Books with Opus 4.1

I've been asking Claude Code Opus 4.1 to write books for me. Sometimes they're quite long. It takes time and energy to guide the process, and the way you guide it heavily impacts the quality and nature of the work that you get out of it.

This is not pleasing to many. I'm hearing arguments against AI (that's just a bit broad), against AI writing anything other than code, and of course the arguments that anything created by AI is "slop".

I'm curious if anyone else is writing multi-chapter books using Claude Code, Codex (Cloud or CLI), or Gemini CLI.

I have it make one chapter per markdown file, and with Claude's ability to pull web pages and github repos, I'm able to feed it back in to Claude to get very granular review.

If you're curious about the books I have published so far under the CC0 license, they're collected here: https://github.com/cloudstreet-dev

You have to click in to the markdown files to read them. An example from a popular one is https://github.com/cloudstreet-dev/React-is-Awful/blob/main/01-table-of-contents.md

There's a little button on the markdown page on GitHub that says "raw", which gives you a link you can feed directly into Claude for review: https://raw.githubusercontent.com/cloudstreet-dev/React-is-Awful/refs/heads/main/01-table-of-contents.md

I do my best to acknowledge that AI wrote the books. (Sometimes claude writes a commit message or phrase that very much sounds like I'm the author of what it made.) The real question, though, is should these works exist?

Should LLMs write books?