I Asked AI to Roast Me. I Wasn’t Ready for the Answer.
A prompt showed up in my feed recently. “Based on everything you know about me from our conversations, tell me 10 brutal truths about myself that I probably don’t want to hear.” (shout out to Deb Haas for the prompt)
I ran it. What came back hit harder than I expected. Not because it was cruel. Because it was close.
The things it got right, it got uncomfortably right. Patterns I’d named out loud but hadn’t fully reckoned with. Tensions I’d described as external circumstances rather than choices. The gap between who I say I’m building toward and what I actually spend my days doing. It wasn’t reading my mind. It was reading my own words back to me with no incentive to soften them. That’s a different experience than journaling. That’s a different experience than a performance review. It’s closer to what a really honest friend would say if they weren’t worried about the relationship.
But here’s the part I didn’t expect. Some of what surfaced weren’t things I already knew and had been avoiding. They were things I had been struggling to name at all. Vague discomforts I’d been carrying around without being able to articulate them clearly. Seeing them written out plainly, without hedging or diplomatic softening, gave me language for something that had just been static. That clarity felt like a release more than an indictment. Sometimes the most useful thing isn’t a new insight. It’s finally being able to name the thing that’s been sitting just out of reach.
What made it land even harder was that some of what came back had also come up in therapy. Recently. I’d already done the work of naming certain patterns with someone trained to help me see them. And here was the same thread showing up again, surfaced from nothing but my own words in a work context. That kind of convergence is hard to dismiss. When the same truth finds you from two completely different directions, it’s probably not a coincidence.
Where it missed, it missed because it’s only seeing part of the picture. It doesn’t have access to my full life, just the slice I’ve shared in a particular context. So some of what it offered felt incomplete rather than wrong. A diagnosis without the full chart. Still worth something, but worth holding loosely.
What makes this hard isn’t the feedback itself. It’s that once you’ve read it, you can’t unknow it. The comfortable move is to treat it like an interesting exercise, take a screenshot, share it as content, and move on without sitting in what it actually said. Most of us will do exactly that.
So what do you actually do with it? I don’t think there’s a clean answer. But I’ve been sitting with the idea that the value isn’t in the list. It’s in picking one thing, just one, and asking whether the way I’m spending my time right now is consistent with wanting to change it. Not a plan. Not a framework. Just that one honest question applied to one pattern that showed up on the list.
The prompt is worth running. But running it is the easy part. The harder question is what you do the day after, when the sting has faded and your calendar looks exactly the same as it did before.
FULL PROMPT:
“Based on everything you know about me from our conversations, tell me 10 brutal truths about myself that I probably don't want to hear. Be completely honest, don't sugarcoat anything, and tell me exactly what's holding me back.”