Almost Never Use AI to Write Anything Substantive
Essay argues writers should avoid AI for substantive prose because writing is thinking, AI output is subtly vague, and unlabeled AI text misleads readers.
The author argues that drafting is inseparable from thinking, citing Paul Graham and a Patrick McKenzie podcast discussion with Clara Collier. The piece claims AI-generated writing is vague and subtly wrong in hard-to-notice ways and that using it without disclosure is misleading. AI is deemed acceptable for transcription, data analysis, brainstorming and human-reviewed editing, illustrated by an LLM paragraph about AI chip smuggling amid US export controls imposed since October 2022.
- Writing process is framed as inseparable from the thinking process
- AI writing described as vague and wrong in hard-to-notice ways
- AI acceptable for editing, transcription, brainstorming and feedback
- Cites Paul Graham, Clara Collier and Patrick McKenzie
- Vendors
- Anthropic
- Products
- Claude
- Organizations
- Paul GrahamPatrick McKenzieClara Collier
- Countries
- United StatesChina
From the original article
15 min read at erichgrunewald.substack.com
I think you should almost never use AI to write -- that is, to do the thing you’re doing when you type words on a page -- whether for a blog post, a research report, a memo, a thoughtful email, a novel, or any other text aimed at conveying an idea, an argument, an analysis, or other substantive 1 thoughts.
By erwald · erichgrunewald.substack.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.