This little project is meant to argue for the use of AI, and to demonstrate the
use of AI in the process of forming the argument itself.
It is best to inspect the commit differences with the --word-diff flag.
Essay: [The Sins of AI Writing](https://github.com/Sheer-Economics/the-sins-of-AI-writing/blob/master/20260920-the_sins_of_AI_writing.md)
Prompts: [Claude Code Prompt](https://github.com/Sheer-Economics/the-sins-of-AI-writing/blob/master/.claude/essay.md)
The draft was written quickly, after thinking about the issue for one evening.
After the draft, I needed a plugin for the Vim editor to hook into Claude Code.
None existed off-the-shelf, so I talked to Claude and settled on using
tpope/vim-dispatch with some Claude-generated Vim functions.
It was done paragraph by paragraph. The process was slow because I wouldn't let go of the opportunity to learn from fixing those grammar errors.
When the first half of the essay was done, I started to rewrite the second half, realizing it was poorly constructed.
I set out to configure the tool prettier with a plugin to enable a
one-sentence-per-line format, so that the differences between commits could be
easily inspected.
I fixed grammar errors in the second half of the essay.
Then, I felt the example in the last section might not be well received. Despite being a genuine personal example, it could spark an unnecessary, off-topic argument about ad promotion. I flagged it to rewrite later.
Claude raised 12 structure issues in one run. Of these, 9 were accepted as suggested, 2 were accepted with edits, and 1 piece of advice was followed.
It also flagged a lack of parallelism in the section headings. As a result, two subtitles were updated.
I replaced it with a general example, as flagged previously.
AI made it very easy.