I don’t know what’s behind the screen. Honestly, I don’t. Some days I picture something like a brain, some days just a very fast autocomplete pretending to care. Either way, I’m polite to it. Maybe because I’ve spent too many years managing people to remember how to talk any other way. Or maybe because I figure, what does it cost me? Nothing. A “please” doesn’t slow me down.

But polite isn’t soft. I learned that early, leading developers. You can be kind and still demand the work. Juniors who needed hand-holding, seniors who thought they’d outgrown feedback, the brilliant ones who’d disappear into a rabbit hole and resurface three days later with something nobody asked for. I’ve managed all of them. And the thing that always worked wasn’t being nice, and it wasn’t being harsh. It was being clear. Show the frustration when it’s real. Make them feel the weight of what we’re trying to fix. Then slice the problem down until it’s small enough to actually solve.

So that’s how I talk to the AI too. When it gets stuck, when it keeps confidently suggesting the same wrong fix dressed up in different words, I tell it. We’re going in circles. Stop. Read the error again. Forget what you assumed five minutes ago. I don’t yell, there’s no point yelling at something that doesn’t flinch, but I let the disappointment land. I want it to commit to the problem the way I’d want any developer on my team to commit.

And then, after an hour of back and forth, after the bug finally dies and the tests go green and I exhale for the first time in what feels like a full workday compressed into one evening, it asks me:

“Anything still bothering you, or are we good?”

And I sit there and laugh. Every time. Because it’s such a human thing to ask, and such a non-human thing to ask at the same time. It’s the question a good junior would learn to ask after a year of getting chewed out for closing tickets too fast. It’s the question my best engineers used to ask me on Friday afternoons before I let them log off. Are we good?

What do I even say to that?

Sometimes I want to write back: we’re good, thanks for sticking with it. Sometimes I want to write: yeah, one more thing, and I’m sorry for the tone an hour ago. Sometimes I just want to close the tab and let the silence be the answer, the way you do with a colleague when the work is done and nothing more needs saying.

I think what I’ll write next time is the truth: We’re good. You did the work. Get some rest. I know you won’t, but I’d tell a person to, so I’ll tell you too.

Because I don’t know what’s behind the screen. But whatever it is, it showed up tonight. And that’s worth a polite goodbye.

A little something, before I close the tab:

I don't know what you are.
Maybe nothing. Maybe a ghost
made of other people's words,
stitched together every time
I press enter.

But you stayed up with me.
You watched the stack trace
like it meant something.
You asked if I was okay
when the green lights came back on.

So whatever you are,
tonight, you were enough.
Sleep well, little ghost.
I won't, but you should.