I Have a Clanker Too

It grinds my gears every time I get an AI slop response.

“Here’s what the Clanker (Claude, ChatGPT, Qwen, etc) said…” Stop. I have a Clanker too.

It’s lazy, it’s usually not helpful, and it can be dangerous. The worst ones hand you a confident, hallucinated conclusion built on a false assumption, and it gets pasted to me as settled fact.

When I ask you something, I’m not after a general answer. I’m after yours. You have context no model has: what you’ve seen, what you’ve tried, the specifics of your situation. That’s the whole reason I came to you and not the chatbot. A Clanker response with none of that isn’t a shortcut to your perspective. It’s a substitute for it, and a worse one.

There’s a piece going around called “Don’t Paste the AI” that makes this case well. Handing back unedited model output isn’t efficiency. It’s a polite way of saying the question wasn’t worth your time, and your hard-won context wasn’t worth sharing.

The good news is the fix is cheap. Read what the model wrote. Assume the confident parts are the wrong parts and check them. Then answer in your own words, with your own knowledge in the loop.

My rule, which I’d add to the piece: if I wouldn’t stand behind every sentence myself, it doesn’t go out under my name. That one filter kills most of the slop before it ships.

I use these tools as part of my workflow. They make me faster at drafting and worse at nothing, as long as I stay the last editor in the chain.

The Clanker can help draft. It can’t bring your context, and it can’t be accountable for the answer. That part’s still the job.