Essays & Analysis

There's a specific kind of frustration that comes from watching someone mistake assertion for reasoning. It's not that they're wrong — it's that they haven't done the work that would let them know whether they're wrong. And there's a difference between those two things.

A claim is a statement. "This policy is harmful." "That person is dishonest." "The evidence supports my view." Claims can be true or false, but they don't, by themselves, give you any reason to believe them. They're just positions.

An argument is a structure. It has premises — things you're asking the other person to grant — and a conclusion that follows from them. The conclusion is supposed to be more believable because of the premises, not instead of them.

The confusion between the two is not always bad faith. Sometimes people genuinely don't notice that they've moved from "here's why I think this" to "here's what I think, stated more emphatically." The shift is subtle. The feeling of having made a point can be identical whether you've actually made one or not.

What I find more interesting than the bad-faith version is the good-faith version. The person who is genuinely trying to persuade, who has thought about the issue, who cares about getting it right — and who still, somehow, ends up just restating their position. What's happening there?

I think part of it is that we're trained to communicate conclusions, not reasoning. In most social contexts, the reasoning is assumed. You say "I think X" and the other person either agrees or disagrees, and the conversation moves on. The premises are implicit. The structure is invisible.

The problem is that implicit premises are often wrong, or at least contestable. And when someone challenges your conclusion, the natural response is to restate it — because you haven't been in the habit of making the premises explicit, so you don't have them ready to defend.

The fix is not complicated, but it is effortful: before you state a position, ask yourself what would have to be true for it to be correct. Write those things down. Check whether you actually believe them. Check whether the person you're talking to is likely to believe them. If they don't, that's where the conversation needs to happen — not at the level of the conclusion.

I've been using AI tools seriously for two years now. I have a complicated relationship with them — which is to say, I find them genuinely useful and also genuinely strange, and I don't think those two things cancel each other out.

The useful part is obvious. AI is very good at certain kinds of writing tasks: summarising, reformatting, generating first drafts of things that follow a template, producing plausible-sounding text on demand. If you need a lot of words that are roughly correct, AI is excellent.

The strange part is harder to articulate. It has to do with what AI writing reveals about human writing — specifically, about how much of what we call "good writing" is pattern-matching rather than understanding.

When an AI produces a paragraph that sounds thoughtful, it's doing something like: identifying the patterns associated with thoughtful writing and reproducing them. The result can be indistinguishable from the real thing. Which raises the question: what is the real thing, exactly?

I don't think the answer is "nothing" — I think there's something real that distinguishes writing that comes from genuine understanding from writing that mimics its surface features. But I'm less confident than I used to be about what that something is.

One candidate: the ability to be wrong in interesting ways. AI writing tends to be wrong in boring ways — it confabulates, it hedges, it produces confident-sounding nonsense. Human writing, when it's wrong, is often wrong in ways that reveal something about the writer's model of the world. The error is informative.

Another candidate: the ability to notice what's missing. Good writing often involves recognising that the obvious framing is wrong, that the question being asked is not the right question, that the thing everyone is talking about is not the thing that matters. AI is not good at this. It's very good at answering the question you asked; it's not good at telling you you asked the wrong one.

I find this genuinely clarifying. Not because it makes AI less useful — it doesn't — but because it makes me more precise about what I'm actually trying to do when I write. The things AI can't do are the things I care most about. That's useful information.

The productivity literature treats attention as a resource to be managed. You have a finite supply; the goal is to allocate it efficiently. I think this framing is wrong, and the wrongness has consequences.

The resource metaphor implies that the goal is conservation. You want to spend your attention on the right things and avoid wasting it on the wrong ones. Distraction is bad because it depletes the tank. Rest restores it.

But this doesn't match my experience of how attention actually works. When I spend a morning doing focused, difficult work, I don't feel depleted — I feel more capable of attention, not less. When I spend a morning on shallow tasks and social media, I feel scattered and unable to focus, even though I've "used" less attention by the resource model.

The muscle metaphor fits better. Attention is something you develop through practice. The more you practice sustained, deliberate focus, the better you get at it. The more you practice fragmented, reactive attention, the worse you get at the sustained kind.

This reframes the problem. It's not about managing a scarce resource — it's about training a capacity. The question isn't "how do I protect my attention from being depleted?" It's "what kind of attention am I practicing, and is that the kind I want to get better at?"

The practical implication is that the cost of distraction isn't just the time lost to the distraction. It's the training effect — the small increment of getting better at being distracted, worse at being focused. Compounded over years, this is significant.

I find this framing more motivating than the resource model, because it makes the stakes clearer. It's not just about this morning's productivity. It's about what kind of mind you're building.