Don't Just Paste the AI — Write Your Own Take

The site dontquotetheai.com makes one point bluntly: pasting raw AI output into a reply is lazy and unhelpful. If someone asked you a question, they want your take — not a generic answer they could've fetched themselves in four seconds.
What's the Problem?
Someone asks a real question. You pop it into a chatbot, copy the answer, and send it back. It feels fast and helpful. But the recipient has the same tools you do. If they wanted the generic answer, they'd have gotten it themselves. They asked you because they wanted your context, taste, and judgment.
What to Do Instead
- Use the AI as a drafting partner. Read what it generates, then write your own take.
- Extract only the bit that actually answers the question. Drop the rest. Three sentences from you is plenty.
- If a piece of the AI's answer is genuinely useful, quote it and say why. For example: "I checked with Claude and this part lines up: ..."
- If you have nothing to add, say so. "No strong opinion here" is a real, helpful reply.
Sending This to Someone
If a colleague drops a wall of model output in your DMs, Slack, or PR review, you can send them dontquotetheai.com. No lecture required. There's also an 'angry version' for stronger feelings (not safe for sending to your manager).
The site is a spiritual cousin of nohello.net and dontasktoask.com. Written by a human, on purpose. Free to share, remix, and translate — PRs welcome on GitHub.
📖 Read the full source: HN AI Agents
👀 See Also

How to Fix Claude Code's CSS Guesswork with a Design System
A developer found Claude Code repeatedly regenerated misaligned HTML/CSS because it designs blind without visual feedback. The solution: provide a complete design system with spacing, colors, and type variables, then separate HTML and CSS prompts.

Agent-Ready Codebases: Negative Rules, Precise Names, Directory READMEs
A developer shares how CLAUDE.md rules, negative instructions, and precise naming cut token waste and prevented Claude Code from bloating classes like UserManager.

Token Usage Tips for Claude Code
Practical advice from a Reddit post on reducing token burn: start fresh chats, group questions, keep CLAUDE.md lean, be precise with file references, summarize and restart threads, and use lighter models for simpler tasks.

OpenClaw AGENTS.md template for automated sales call prep
A Reddit user shares an AGENTS.md instruction for OpenClaw that automates lead research before sales calls, investigating company details and pain points to send a briefing 10 minutes before meetings.