AI Helps Code, But Hollowes Prose: In Praise of Messy, Human Writing

The author values writing as a direct expression of thought and distrusts AI-generated prose. They happily use LLMs for coding and documentation, where AI feels like genuine progress. But for essays, they want intentional, messy human effort—proof-of-work that AI increasingly obscures.

Key Points

Writing reveals a person’s thinking, so outsourcing prose to AI erodes the reason to read it.
The author embraces LLMs for coding tasks (docs, tests, scaffolding) where AI brings real efficiency.
For content, intention and the struggle to articulate thoughts are essential—human proof-of-work matters.
Imperfect, less polished writing now reads as a positive signal of authenticity.
Because AI can fake roughness, authenticity signals are deteriorating, feeding the dead internet theory.

Sentiment

The community broadly agrees with the article's thesis that AI-generated prose feels hollow and that human intentionality in writing matters. However, there is strong and persistent pushback on the perceived hypocrisy of embracing AI for code while rejecting it for prose. The tone is engaged and largely constructive, with many commenters sharing personal anecdotes and nuanced takes. The em dash subthread is particularly lively, reflecting genuine frustration about how AI has contaminated previously innocent stylistic choices.

In Agreement

Writing embodies a social contract where the writer invests more effort than the reader; AI breaks this contract by making writing effortless
AI prose is detectable, hollow, and verbose — it sounds smart but communicates poorly, full of unnecessary metaphors and TED Talk-style sentence structures
Typos and grammatical imperfections have paradoxically become signals of authenticity in an era of AI-polished text
The internet is being degraded by AI-generated content that optimizes for appearing informative while adding no real value
Writing forces thinking and clarifies ideas; outsourcing it to AI means losing that cognitive benefit

Opposed

The article contradicts itself by embracing AI for code and documentation while rejecting it for prose — code is also written for humans to read and conveys intent
Everyone considers their own AI usage justified while labeling others' usage as slop, revealing a self-serving bias rather than a principled stance
Dismissing text solely because it appears AI-generated is a form of genetic fallacy; writing should be judged on its own merits regardless of origin
AI writing artifacts will normalize over time, just as autotune, digital photography, and JPG compression became accepted parts of creative production
LLMs can serve as valuable editing and thinking tools — challenging assumptions, finding gaps in arguments — without replacing the writer's voice