Using AI is fine; I use it daily. Posting the raw output without reading it first is the tell.
The em dashes in every other sentence. The emoji bullets in a README. A stock-feeling header image that gestures at your topic without saying anything specific about it. That opener about today's fast-paced landscape. Wikipedia maintains a public list of these patterns ("Signs of AI writing"), and a solid chunk of my feed is speedrunning it.
Here's what those patterns tell me. You saw a generated draft and posted it. You read 'leverage synergies' and hit publish. The writing isn't yours, and it looks like you never noticed.
If that's how you treat a post, I have to assume it's how you treat a pull request.
You skipped the whole job. The model hands you a draft. Your value is what you add on top: catching the bug it was confident about, throwing out the approach that looks clean but is quietly wrong.
Relay the output as is, and you've made yourself a hackathon GPT wrapper.
When Claude hands me a function, I rename its variables and hunt the edge case it missed. Then I cut half its comments.
You're the editor between the draft and the publish button.
Put the effort back in.
Anyway, enjoy the sunrise I did in Microsoft paint (no AI).
Drop the worst AI tell you've seen in comments; I'm collecting them for a Claude skill.
Top comments (0)