Thinking for myself

I write the message, paste it into an AI, and ask for a gut check before I send it. Most of the time it comes back fine. Sometimes it comes back with a flag:

Cut that line; it implies you don’t trust them.

The one I remember was a reply defending a decision I stood behind. I was trying hard not to sound defensive, and it came off condescending instead.

The first thing I feel is relief, because I was about to send that thing. The second thing stings, and it arrives a second or two later. If that sentence needed catching, then who’s to say sentences like it haven’t been going out uncaught for almost twenty years?

Nobody flagged those. Nobody was there to tell me no. The room adjusted its read of me slightly and moved on, and I was none the wiser. There was no log, no check. I can run the check going forward. I can’t run it backward, and that is the part that keeps me up at night.

The unflagged years

I can spiral in that direction for a while, over the messages I may have sent fast because I had no time, or because I was annoyed and reasonable-sounding at the same time, and the times I was the most senior person in the room and nobody chose to tell me. I can think of five things I did today where maybe I could have thought a little differently.

The feedback loop on how I land has always been broken, because almost nobody tells me. Sometimes I don’t hear back at all, so one can only assume. I think I create a safe space. But AI is the first thing in my working life that reads a draft and says the quiet part (edited by AI, called out by my colleague, kept because I have no shame) before the damage instead of after it. Once I had watched it do that a few times, I started to wonder, and I couldn’t unsee the size of what probably went unsaid before.

I’m here. Whatever I’ve been saying for almost twenty years has worked more often than it didn’t, and I have walked into hard rooms and come out with my self-esteem and my work intact. Something was already doing this job, badly logged and unmeasured, long before there was anything in my ear.

The pause

What was doing it was the pause.

I’ve practised mindfulness for years, on and off. I’ve led it at work, and it’s part of how I deal with everyday life. I don’t mean that in the wellness sense. I mean it as a working method. Well, maybe I mean both.

Before I speak into a hard moment, I stop. I find some space. I find still. I notice, as best I can, what I am about to say and where it is coming from. Then I say the thing, or I say a better thing, or I say nothing yet.

When I picked that conversation back up somewhere else, I paused and took their perspective. I didn’t placate, and I didn’t defend. I listened.

It is the same reflex a professor drilled into me in crits years ago, when he would cut off anyone who said they liked something. I wrote about him in Taste is triage. Why. Liking is free; why do you like it? The value is in the because. Turned inward, the question becomes why am I about to say it like that. I’ve built a set of short sessions around arriving somewhere quiet for exactly this.

The pause is slower than a flag on a draft, and it does a different job. A flag fixes the sentence in front of me. The pause is what taught me to write the next hundred without needing one.

What I’m actually worried about is the catching itself. If something else is always going to catch the bad sentence, I stop doing it. The pause is a muscle, and muscles atrophy. It’s built by being the only thing standing between me and a message I should not send. If something else has stood in that spot for a year, I doubt much of the muscle survives.

What the pause can’t see

Except the pause has a blind spot too.

Stillness makes me considerate on my own terms. It cannot tell me how a phrase lands in a culture I have not lived in or a role I have never held. No amount of breathing gives me that insight, because it isn’t in me to find. When the flag says this reads as dismissive of their constraint, sometimes I could have reasoned my way there myself. Sometimes I could not have, because what I am missing is the view from where they sit.

Accepting that costs me nothing I want to keep. It is the same admission as not being able to read every diff anymore. There is more context in play than one person can hold, and that complexity is the beast I don’t wish to tame.

The imperfection defence

So then the romantic argument, the one I deeply want to be true. Maybe a person whose every sentence has been sanded smooth reads like a press release, and the small friction of a slightly wrong word is how anyone can tell a human is actually there. Is the imperfection what keeps it real?

Part of that holds up. Over-smoothed language does cost trust. People can feel a sentence that has been through too many hands, and they discount it.

But I have to be careful with it, because “keeping it real” is the most convenient excuse available for skipping the actual work of consideration. The person on the receiving end of my careless sentence gets nothing out of my authenticity. I know the difference very well, because I’ve tried to learn where the useful discomfort ends and the self-serving kind begins.

What I want to protect is the exposure: being present in what I say, with intention, and accountable for where it lands. None of that requires the mistake. I can be careful and still be exposed, and for a long time I had those confused.

The rule I’m running

So: is it better to have somebody in my ear? I don’t know, and I’m suspicious of anyone who does this early.

Here is what I’ve settled on while I find out. I use it the way I’d use a mirror. It tells me what a sentence risks and what it loses. What I mean is still mine to decide. I use my taste and my judgment. When it flags something, I make myself answer the professor’s question before I take the edit. You say it’s better? Why is it better? If I can say why, I keep the edit, and I have learned something. If I can’t, I send my own version, because taking an edit I cannot justify is just doing as I’m told, and that’s where passion goes to die. Later in the same exchange, it told me to double down on apologising. I didn’t. The decision was right, and apologising for it would have undercut it. That’s the kind of flag I push back on: when it picks a side, or makes it about sides at all, when everyone is meant to be on the same team.

The same tool is a second reader or a replacement, and what decides which is whether I read the because. Do I understand it? Can I say why? The replacement version doesn’t announce itself. The messages keep coming out fine, they seem okay, which is exactly why I wouldn’t notice in the first place.

Of course I have said the wrong thing. Everyone has. There’s no log, and no one’s keeping track. Feelings are the only thing that persists. If they felt an error, that sticks, but if they felt a human, that sticks deeper. The question I actually care about is whether it was ever me speaking. Looking back, it won’t always feel like it was. That one I can answer, as long as I do the thinking first and let the flag come second.

This is the thinking I’m putting out in the open as I work through it. If it’s the kind of argument you want to follow, come along.

This is the kind of thing I write about.

More on agentic work and what it does to design craft. Or write to me.