My body of work is more coherent than I am

Something is happening to the way I work, and I want to write it down before it becomes invisible to me.

I think what might be happening is this.

I’ve had these ideas for twenty years and nowhere to put them. Now they’re flooding out, in a form I can actually read back — and when I read them, I think: yes, that’s how I think.

Which is an odd thing to be able to say.

Two explanations, and one of them is testable

The first is a backlog. Two decades of observations that never had anywhere to go, now draining. Things I noticed in interviews, in newsrooms, in campaigns that didn’t work, in research that said something nobody wanted to hear. None of it was written down as thinking. It went into whatever piece I was producing at the time, or nowhere.

The second is generation. The accumulated context and the habit of publishing are producing connections that wouldn’t otherwise have existed — links between a customer research finding from 2019 and something I noticed last week.

Both are probably true in some proportion. But they predict different futures, which means I’ll find out.

If it’s mostly backlog, the rate drops. The good material gets spent, and each new piece takes more effort than the last. If it’s mostly generative, it doesn’t drop, because each published argument creates new surface for the next observation to attach to.

I’d rather say now which one I expect, so I can’t reinterpret it later: I think it’s mostly backlog, with a generative component that’s real but smaller than it currently feels.

There is at least a mechanism I can describe, even if I can’t weigh it. The published pieces are in front of me while I work, so a new observation doesn’t arrive into an empty room. It arrives next to everything I’ve already argued, and the first thing I do with it now is check whether it contradicts something, extends something, or has already been said better a fortnight ago. That is not how I used to think, and it changes what a new idea is for — less a thing to write up, more a thing to test against a position that already exists.

It also means I’m reading my own back catalogue constantly, which is an odd habit and a useful one. Whether it amounts to generation or just to better-organised backlog, I genuinely don’t know. It would produce the same feeling either way.

And a draining backlog is a success rather than a failure. Twenty years of unpublished observation finally getting out is exactly what should happen. It just shouldn’t be mistaken for a permanent rate.

The strange part is the recognition

You don’t normally get to see your own thinking from outside.

Most people’s ideas stay in the form they occurred in — half-articulated, dependent on the moment, never fixed in place long enough to be examined. They get expressed in conversation, or in a piece of work with somebody else’s brief on it, and then they’re gone.

What’s accumulated here is something different: a set of my own accounts, dated, in my own voice, that I can now read the way a stranger would.

Which is the essay I wrote about interrogating an unreliable narrator, from the other side. I’ve been writing about the problem of assessing somebody’s testimony against the record.

I’m now the narrator.

Why I don’t entirely trust the recognition

Here’s the part I’d rather not write, which is usually a sign it’s the useful part.

Reading your thinking back in finished prose makes it look more coherent and more considered than it was when you had it.

The edges got smoothed. The arguments acquired their concessions. Contradictions between pieces got noticed and resolved — sometimes by me, frequently by the AI I use to develop this material, which is very good at spotting that an essay published on Tuesday cuts against something published on Sunday, and asking which one I actually mean.

That process is genuinely valuable. It is also a manufacturing process.

So when I read the body of work and think yes, that’s how I think, some of that is recognition and some of it is the flattery of good editing. The corpus is more consistent than the head it came from. It has fewer loose ends, because loose ends get tidied at publication rather than in the mind that produced them.

None of which makes the thinking not mine. The observations are mine, the arguments are ones I’d defend, and several of them changed while being written. But the coherence is partly an artefact of the process, and it would be convenient to forget that.

Why this matters beyond me

Because I keep arguing that accumulated context is where the advantage sits, and this is the failure mode of that argument.

A body of work that reads as a settled position is more persuasive than one that reads as somebody still working it out. Which means the process quietly rewards the appearance of having arrived — smoother arguments, fewer visible contradictions, a system that looks like it was designed rather than accreted.

The methodology is meant to be a set of propositions rather than a doctrine. Publishing well makes propositions look like doctrine. That’s not a reason to publish worse. It’s a reason to keep saying which parts are still unsettled, and to treat the moment when it all starts feeling finished as a warning rather than an achievement.

If this ever reads like a completed system, something has gone wrong — and the most likely culprit is the editing, not the thinking.