I Don’t Want to Read What You Didn’t Write – Why AI‑Generated Text Is Undermining Communication

AI‑Generated Writing Is Unreadable and Counterproductive

Takeaway: AI‑written design docs, PR summaries, and emails are often dense, context‑free, and punish readers, which harms collaboration.

Authors use AI to retroactively summarize work they have already built, turning design documents into exhaustive machine‑generated reports that lack perspective and consensus‑building intent.
Pull‑request descriptions generated by LLMs list every change without indicating risk, urgency, or where reviewer input is needed, leaving reviewers unable to prioritize.
Personal messages crafted with AI become impersonal and disjointed, stripping away the speaker’s voice and emotional nuance.
The author quotes Simon Sarris: “write as much as you can with your own empiricism… give the reader your own characterization of life or events.” This underscores the value of authentic, human‑authored prose.

“I think one should write as much as they can with their own empiricism, their own senses, giving the reader their own characterization of life or events.” – Simon Sarris, Resist Summary

Readers Lack the Context That Prompt‑Engineers Have

Takeaway: When a human writes with AI, they retain the prompt context, but anyone else receiving the output lacks that context and must read the entire text to understand it.

Cynthia Dunlop’s survey shows 78% of developers stop reading an article they suspect is AI‑generated, and 71% avoid the author thereafter.
Bryan Cantrill describes this as an “LLM‑triggered ejection handle” that aborts reading mid‑sentence.
Bjarne Stroustrup notes that a reader’s sense of the author’s voice is essential for engagement; AI‑rewritten text feels dry and academic.
The author’s own experience: when he prompts an LLM he can skim the output because he knows the constraints; a colleague receiving the same output must read exhaustively, often abandoning it.

When AI Is Actually Helpful for Writing

Takeaway: AI excels at low‑level tasks—verification, citation, grammar, and abstract summarization—when used as an assistant rather than a writer.

In an academic paper, the author used AI to verify technical details (e.g., database indexing, Parquet sorting) against source code and logs, catching errors that human reviewers missed.
AI auto‑filled BibTeX citations, saving time without breaking writing flow.
Grammar and style suggestions from the LLM improved readability, and the model generated a perfect abstract, a task well‑suited to summarization.
AI‑drawn TikZ diagrams accelerated figure creation.

“The AI was incredibly good—indeed ruthless—at finding spelling and grammar mistakes… It even found a very subtle mistake where I had used the incorrect notation in one paragraph, something four human reviewers missed.”

Limits of Current LLMs and Prospects for Improvement

Takeaway: LLMs are fundamentally statistical predictors lacking empathy, a mental model of the reader, and personal stake, which limits their ability to produce truly resonant prose.

Murat Demirbas argues that writing quality has plateaued because LLMs optimize next‑word probability rather than creative intent.
The essay cites the ASD‑STE100 Simplified Technical English standard as a way to enforce clarity in technical documentation, though it is only suitable for instruction manuals.
Pangram, a model fine‑tuned to detect AI‑generated text, is being deployed at Oxide to enforce a “human‑authored” contract for public writing.

“Since LLMs lack an active mental model of a specific human reader, they are just optimizing for the statistical probability of the next word… They cannot empathize with the human reader, as they don’t have the human lived experience.” – Murat Demirbas

The Human Element: Voice, Uncertainty, and Narrative

Takeaway: Authentic writing conveys uncertainty, personality, and narrative depth that AI cannot replicate.

Iain McGilchrist argues that important concepts resist reduction to language; AI‑generated “detail” creates an illusion of meaning without genuine understanding.
Simon Sarris emphasizes that “narrative resists compression” and that attention to detail is the opposite of summary.
Comments on HN echo this sentiment: many users refuse to read AI‑filled PRs, tickets, or essays, preferring raw human thought even if imperfect.

“When you write, the meaning of our words is not always direct, factual, or concise… The whole cannot be decomposed into parts without a substantial loss of meaning.” – Iain McGilchrist

Community Responses and Practical Workarounds

Takeaway: The Hacker News discussion surfaces concrete strategies to mitigate AI‑generated noise while preserving useful assistance.

Pre‑amble sections: One commenter adds an “AI;DR” preamble to PRs that is guaranteed to be human‑written, summarizing risk and reviewer focus.
Human‑only TL;DR: Others suggest keeping TL;DR sections free of AI to convey high‑level intent.
Pair review: Conducting synchronous PR reviews restores human context and catches mistakes that AI‑generated summaries miss.
ASD‑STE100 prompts: Teams have tweaked their agent prompts to output in Simplified Technical English, making technical text more digestible.
Selective prompting: Some users find that carefully crafted prompts can produce readable AI output, but they still edit to retain their voice.
Boundary setting: Several comments recommend explicitly asking colleagues for “non‑LLM” versions of documents, establishing cultural norms around AI use.

“I ask coworkers point blank to send me a non‑llm version of their slop sagas. The initial exchange can be a bit awkward, but consistently asking for human prose has worked for me.” – HN commenter

Conclusion: Preserve Human Voice, Use AI as a Tool, Not a Substitute

Takeaway: To keep communication effective, writers should let AI handle repetitive, verification, and formatting tasks while retaining authorship of narrative, context, and voice.

The essay’s final plea: “I don’t want to live in a world where you use AI to summarize something important into unreadable text, and then I use AI in an attempt to decipher it. I want to hear you, imperfections and all.”
As AI becomes more pervasive, intentional writing—writing to think, to create, and to share—will grow in value, while the flood of AI‑generated “slop” will be filtered out by readers who demand authentic human expression.

Key Quotes from the Discussion

“If it isn’t worth the time for someone to write something themselves, it isn’t worth my time to read it.” – HN commenter

“Reading a document like this isn’t just difficult—it is punishing.” – Repeated by multiple commenters as a succinct summary of the problem.

“The design document is no longer a proposal to build consensus… it is a summary made by a machine, in exhausting detail, without context or perspective.” – Author’s critique of AI‑generated design docs.

“I would much rather someone be themselves and write with their own voice, or with some passion, even if they can’t quite find the words.” – Author on personal communication.

“AI;DR” preamble can rescue PR descriptions by providing a human‑written high‑level overview. – Community suggestion.

Takeaway for Practitioners

Use LLMs for low‑level assistance: grammar checks, citation formatting, diagram generation, and factual verification.
Keep narrative, risk assessment, and reviewer guidance in human‑written sections.
Adopt standards like ASD‑STE100 for technical docs to enforce clarity.
Deploy detection tools (e.g., Pangram) to enforce a human‑authored contract where needed.
Establish team norms that request non‑LLM versions for critical communication.

By treating AI as a collaborative assistant rather than a replacement for authentic voice, engineers and writers can avoid the “unreadable slop” epidemic and preserve the human connection that makes technical communication effective.