The Roast Everyone Needed

The folks behind dontquotetheai.com just dropped what might be the most relatable piece of web copy to hit Hacker News this year. Titled 'Don't quote the AI at me,' the site is a scathing indictment of developers who take questions from colleagues, friends, or strangers, dump those questions into an LLM prompt box, and paste the unedited result back as if they actually thought about it. The site's central argument cuts clean: if you're going to reply with 800 words of raw ChatGPT output starting with 'Here's what Claude said,' you've just proven there's no difference between asking you or asking the AI directly. The question wasn't for the model—it was for you, specifically. You made everyone lose time and respect in one lazy copy-paste.

What Makes This Different

Unlike typical productivity essays that tiptoe around the issue, dontquotetheai.com doesn't waste words. It breaks down exactly what those 'great question!' responses actually communicate: 'I couldn't be bothered to read your question carefully enough to reply, so here's what a chatbot guessed for me.' The site calls this behavior 'the conversational equivalent of forwarding the email,' and honestly, that's generous. The manifesto also takes direct aim at the self-congratulatory tone many LLM-pasters adopt. It mocks responses that start with industry-recommended best practices and mild Google search results wrapped in confident-but-empty hedging language. 'Wow... Enlightening. Thanks for this piece of wisdom I could've gotten myself,' the site deadpans, mimicking the recipient's internal reaction.

The Actual Point About AI Tooling

Here's where it gets interesting: dontquotetheai.com isn't anti-AI. It explicitly encourages using these tools daily—to draft faster, think rougher, unstick yourself on hard problems. The site's actual beef is with treating LLM output as a finished deliverable rather than a starting point. 'Treat it like a junior intern's first draft,' the site advises. 'Edit it. Cut it. Disagree with it. Make it yours, or don't send it.' That's practical advice most developers could use hearing. The manifesto even suggests marking when you do quote a model: 'I asked Claude and this part actually checks out:' is fine. Actually adding value to that information is required.

Key Takeaways

  • If your answer starts with 'Here's what [model] said,' you've already lost the conversation
  • AI output is a starting point, not a deliverable—read it, edit it, or don't send it
  • Models are confident in ways that have nothing to do with being right; half of it is probably wrong
  • If you have nothing to add beyond what an LLM would say, say nothing instead

The Bottom Line

This site is the spiritual successor nohello.net and dontasktoask.com always deserved—written by a human, on purpose, with feelings. The developer community needed someone to say out loud what everyone was thinking: pasting AI responses isn't helping, it's just proving you can be replaced by the thing you're quoting. Use the tools. Think for yourself. Or stay silent.

Spread the Word

The site even includes a shareable call-to-action for when you receive one of these dreaded unedited LLM dumps in your DMs or code review comments. No explanation needed—recipients will know exactly why they got it.