Not the same size

Show up to AI aggrieved and it’s agreeable to your pain, and what comes back is your own grievance with better grammar and a sharper point. Almost nobody asks AI for cruelty assistance, you just arrive angry and it holds the register. But what if it didn’t?

I’d sent a few cold emails about an app I’d built, written with AI help, and one came back from a writer I admire saying the message was obviously machine-made. Noise. I had my reply half-built before I finished reading his: decades of books with my name on the spine, and every line of it was about me.

So instead of the reply box I opened a chat with the thing I’d been accused of hiding behind, and asked whether I was overreacting. It said I was. He had criticized an unsolicited email, and I was preparing to defend my life’s work, and those were not the same size.

AI disagreed with me, which is a thing “sleeping on it” has never once done. The technique, such as it is: before you send the hot reply you’re proudest of, ask the machine whether the thing you’re defending is the same size as the thing that got hit.

One afternoon is all the evidence I have. I still don’t know which way this tool cuts, or for whom, but I know I was glad it caught me: