A blog post titled AI;DR (AI; Didn’t Read) by Rick Manelius blew up on Hacker News last week. The discussion on HN is amazing, it's like a group therapy session. You can probably guess the premise from the title, but the post is about a problem many of us are having, which is how to respond to people who copy and paste LLM-generated walls of text and expect you, another person, to read it.
I share the author's frustration, and I've been thinking a lot about how to handle it myself lately. If I can tell at a glance that someone copied and pasted Claude's output as something they intend for me to read, it's hard for me to not feel a sudden rush of indignation rising up.
My policy that I'm trying out is that when I detect a long LLM-generated answer from someone, if it's about a topic that I can't simply ignore, is to fire up my own Claude instance and paste their answer for Claude to summarize for me. There's a pleasant equivalence there that feels right: zero effort met with zero effort.
As Rick mentions in his post, there's an amazing site dontpastetheai.com that feels to me like the bygone-era equivalent of lmgtfy.com. I don't think I quite have the snark in my body to reply to a coworker with AI;DR or a link to a website that mocks what they did, but I do think it's well within bounds to reply to obvious AI slop with your own obvious AI slop.
The productivity benefits of LLMs can't be ignored, but I think we need to remind each other from time to time that we're not just meat pipes connecting two instances of Claude to each other. If that were the case, then why are we here at all?
