AI;DR (AI; Didn't Read)
TL;DR
Rick Manelius adopts "AI;DR", short for "AI; didn't read", as a personal rule. If someone sends him AI output they didn't bother to review and edit, he won't bother to read it. He accepts fully generated text in some settings, such as customer support.
The article
- The phrase isn't his. He credits an August 15, 2026 tweet by @seclilc, which drew about 16.6K likes. He says he has thought about it ever since.
- He describes a physical reaction, a flinch or an eye twitch, when someone he respects sends him unfiltered and unedited AI output.
- He accepts that by Q3 2026 nearly everyone uses AI somewhere in their process, for sourcing ideas, outlining or refining prose. His line falls at publishing text you haven't reviewed and edited yourself.
- Fully generated text is fine in some places, such as customer support, where nobody wants artisanal "did you reset your phone" dialogue. A colleague who pastes a wall of Claude output into Slack sends a different message than they meant to.
- The same applies to newsletters and social posts. Your name is on them. If the text is raw AI output, a reader could have asked Claude directly. He frames AI;DR as the AI-era counterpart to TL;DR.
Why it matters
The post takes two minutes, and the summary covers its whole argument. The later updates are the main reason to click. They link dontpastetheai.com, which a reader suggested, and the shareable AI;DR graphics he made after the post reached #1 on HN.
The discussion
A huge, mostly sympathetic thread whose biggest subthread had little to do with essays. Engineers described codebases and PRs buried in AI-written comments and argued over whether pushing back costs you your job. Elsewhere people fought over whether provenance matters if the content is good. One commenter posted his own raw prompt next to Claude's essay and was told, nearly unanimously, that the prompt read better.
AI comments and docs flooding code review
The top thread opened with LPisGood describing coworkers whose PRs carry hundreds of lines of AI documentation and comments about "load bearing paths". Others added that models narrate the diff instead of the code, and cite plan items and scratch files nobody can see. They keep notes about removed code and write tests that check whether a Dockerfile contains the lines just added. Fixes ranged from CLAUDE.md rules and hooks that block comments over two lines to stripping comments out entirely. Several argued the comments hurt the next AI pass too, because a note about what not to do primes the model to do it.
The ability of people to write things has exploded tenfold. You can’t out review the slop.
The problem is the comments only make sense if you knew the original prompt in the first place.
Push back, or approve and stay employed
BeetleB and tcmart14 said to reject sloppy PRs and become the bottleneck so the team has to talk about it. Review was always the constraint, and LLMs only made that obvious. Many replies called this naive, citing top-down AI mandates, managers who measure velocity, and colleagues who rubber-stamp. Some had been PIPed, threatened or retaliated against for raising quality concerns, and one named Cisco. 27183 concluded that caring was his biggest career mistake. Others answered that a neutral question in a retro is low risk, and that a workplace where it isn't was already broken before AI.
They have a machine that produces a lot of code, but don't have a machine to review that code.
If you hold up the line by flagging quality issues, you will be seen as the root cause of the problem.
Just send me the prompt
A popular suggestion was to share the prompt instead of the output, since the prompt holds all of the sender's actual thinking. supermatt tested it by posting his rambling prompt beside Claude's essay, and nearly every reply preferred the prompt's voice and visible reasoning. Others suggested asking the AI only for spelling and paragraph fixes, and supermatt, whose cognitive impairment slows his writing, found that readable. Skeptics noted that real work often spans dozens of prompts or an agent run over logs and infrastructure, which leaves nothing short to share.
The LLM response has none of that character, it doesn't give me that little sliver of insight into the writer.
Your thoughts are entirely contained within your prompts. Just send me those.
Does it matter who wrote it
Defenders of the post said writing is communication between people. You can't ask follow-up questions of an author who didn't write the text, and generated prose moves the work of thinking onto the reader. Critics said only quality matters, that ghostwriters and copywriters long predate LLMs, and that heavy AI editing helps non-native speakers. Two commenters who work with non-native writers answered that they would rather read broken English. The restaurant analogies didn't settle it. One side said the chef owns the outcome. The other said reheated food is still reheated.
It shifts the cognitive burden from writer to reader.
No. It very much isn't about outcome. It's about interpersonal relationships.
False positives and lost phrases
Writers complained that em-dashes, "load bearing" and tidy markdown now get them accused of using AI, sometimes wrongly. One commenter apologized to an intern he had accused, unsure whether he had been right. Others said tells like "It's not X, it's Y" are real, and false positives are a fair price for filtering so much slop. A twenty-reply subthread of mock AI phrases, like "You're absolutely right", showed how familiar the style has become.
I loved using "load bearing" to describe weird code that you think you can delete but is actually holding everything together.
the alarming trend, at least here on HN, is crying wolf at the first em-dash.
Consensus
Pasting unreviewed AI output at a person is rude, because it hands the reading and checking to the recipient. Using AI as an editor or spell checker on your own writing is a different thing and mostly fine.
Dissent
A sizable minority said judging text by its origin is lazy and that quality is what counts. They warned that AI;DR is turning into a cheap way to dismiss opinions people dislike. It also penalizes non-native speakers and people who struggle to write, they said.