AI;DR (AI; Didn't Read)

Original Article ↗ Hacker News Discussion ↗

Growing use of large language models to write emails, docs, blog posts and code comments is provoking a backlash from people who feel flooded with low-effort “AI slop.” Commenters argue that pasted, unedited AI output shifts cognitive burden from writer to reader, erodes trust and personality in communication, and often masks shallow understanding or even incompetence at work. Many propose emerging norms such as “AI;DR” (AI; didn’t read), sharing prompts instead of full outputs, demanding concise, human-edited summaries, or even banning AI-written text in human-facing contexts to protect readers’ time and preserve authentic voices.

What “AI;DR” Captures

  • Shorthand for “AI; didn’t read”: a reaction to walls of obvious LLM output, especially in Slack, Jira, PRs, LinkedIn, and blogs.
  • Many see it as a time-protection heuristic: if the sender didn’t invest effort or understanding, the reader won’t either.
  • Some worry it’s sliding into a cheap dismissal of opinions (“new ‘paid shill’”) rather than a critique of text quality.

TL;DR vs. AI;DR and Etiquette

  • TL;DR can be hostile when used as a reply, but valued when used as a self-summary at the top.
  • AI;DR is proposed similarly as a label or preface, but some see it as more condemnatory because it targets process (use of AI) not just length.

Prompts, Transparency, and “Meat Proxies”

  • Strong recurrent idea: if output is AI-generated, share the prompt and model, or even a link to the whole session.
  • Rationale: readers want the human’s intent and selection, not an unfiltered expansion of a one-line thought.
  • “Don’t be a meat proxy”: if you outsource thinking and just paste AI output, you’re wasting others’ time and can’t defend the content.

Workplace Friction and Misuse

  • Many report coworkers and managers pasting huge AI replies into email, Slack, Jira, PRs, or architecture docs, often unedited and poorly constrained.
  • Problems: unreadable verbosity, loss of key details in noise, made‑up metrics and roadmaps, code and tests that no one understands, juniors unable to explain “their” work.
  • Some orgs now discourage or ban AI‑generated text “for humans,” or fire staff who can’t explain AI-produced artefacts.

Quality, Style, and Detection

  • Common complaints: low information density, generic “corporatese,” smug tone, repeated AI tics (“load‑bearing,” “that’s the gap,” em‑dash habits).
  • Others note false positives: good human prose is increasingly mislabelled as AI; writers feel pressured to avoid certain words or styles.
  • Some argue the real issue is laziness and lack of editing, not AI itself; carefully curated AI‑assisted text can be valuable.

Broader Concerns

  • Fear of an internet dominated by homogeneous “slop,” making genuine human voices harder to find.
  • Worry that people stop thinking deeply—delegating not just wording but reasoning to LLMs.
  • Counterpoint: AI can be a useful editor, summarizer, or accessibility aid (non‑native speakers, cognitive impairments) when humans remain responsible for the final output.