I think I accidentally built my own Substack

I didn’t build it because I needed another content channel. I needed somewhere to put my brain.

I think I accidentally built my own Substack.

I call it Field Notes. It sits on this website, and I’m increasingly realising it isn’t a publication in the way I assumed it was when I started.

It’s a public notepad.

Somewhere I own, where I can scribble down whatever I’m noticing about AI, editorial work, customer research, organisational memory, careers, publishing, games, or anything else that happens to connect with what I’m thinking about.

I don’t particularly need thousands of people to read it.

I hope people do, and I hope some of it is useful. But it would still have value to me if hardly anyone did — because part of the point is being able to see what’s actually in my head.

Publishing as a way of finding out what you think

I’ve been writing professionally for nearly twenty years. I know I can write.

The problem was never whether I could turn an idea into sentences.

It was whether I had the time and headspace to take every half-formed observation, work out what was underneath it, connect it to everything else I’d been thinking about, structure the argument, and turn it into something coherent.

That’s the part AI has changed for me — and I’ve written about the mechanics of it elsewhere, so I won’t repeat them here.

What I hadn’t expected is the second-order effect.

I capture the thought. Give it the context I’ve already built. See an interpretation of it. Push back. Notice what’s missing. Realise I believe something more strongly — or considerably less strongly — than I thought I did. Then edit.

Which means Field Notes isn’t only a way of publishing my thinking.

It’s a way of discovering it.

Several times now I’ve started a piece believing one thing and finished it believing something adjacent but different. That used to happen too, over weeks. It now happens in an afternoon, which means it happens far more often.

The double standard nobody quite says out loud

There’s an asymmetry here worth naming.

Nobody finds it strange that a developer uses AI to turn an intention into working code. We’ve adjusted quickly. The developer is judged on whether they understand the system, make good architectural decisions, and can tell whether the output actually works.

Writing hasn’t made the same transition.

If you’re “a writer”, there’s still a widespread assumption that generating the sentences yourself is the thing that proves the expertise. That the typing is the skill.

For some kinds of writing, I think that’s right. Voice, rhythm and precision are real, and there’s work where the sentence is the product.

But it’s an increasingly odd way to describe what I do.

Writing is part of it. So are editorial judgement, research, synthesis, argument, context, systems, publishing, and increasingly the ability to work intelligently with AI.

The scarce thing isn’t producing another paragraph.

It’s having something worth saying, and knowing whether the resulting argument actually says it.

That second half is the bit people skip. Recognising that an argument has come out weaker than the idea behind it is a specific skill, and no amount of fluent prose substitutes for it.

Why it turns out to be useful professionally

At Sage, and with people outside Sage, I’m often thinking about themes that aren’t directly relevant to somebody’s role.

They don’t need the whole Editorial Intelligence system. They don’t need the repository, the research, or months of accumulated context.

They might just find one observation about customer research useful. Or something about organisational memory. Or what AI might do to junior careers.

Field Notes lets me surface one piece without making anybody navigate everything underneath it.

That’s a more modest function than “publication”, and I suspect it’s the more valuable one.

The risk, which isn’t hypothetical

There’s an obvious failure mode in building something like this: everything starts to look like material.

A holiday becomes a series of posts. A conversation becomes evidence. An experience gets assessed, while it’s still happening, for whether there’s an angle in it.

I’d like to say I’ve avoided that. I’ve written from a family holiday in Thailand about the local cost of living, built on my wife’s account of her own town and our electricity bill, and that fortnight produced several thousand words. So the risk isn’t theoretical.

The distinction I’m trying to hold is narrow but I think real. I’m not doing things in order to have something to publish. I’m using the writing to understand things I was doing anyway — the electricity bill arrived whether or not I wrote about it.

The test, if I ever wondered: has anything been done differently because it would make a better piece? So far I don’t think so. It’s exactly the sort of thing that would be invisible from the inside, though, which is why it’s written down here rather than left as an assurance.

The format decides what kind of argument you’re allowed to have

This is the reason I’d give now, and I only worked it out by trying to do both.

A LinkedIn post has to arrive somewhere. It can concede things — the good ones do — but a concession there is a move inside an argument that still lands. What a post cannot do is stay genuinely unresolved, because an ending without a position reads as somebody who hasn’t finished thinking, and on that platform that is indistinguishable from having nothing to say.

Most of what I actually want to write about is unresolved. Whether AI is raising the floor or removing the ladder. Whether the coherence I read back in my own work is real or a product of good editing. Whether my comfort with all of this is judgement or just proximity. In each case the interesting part is precisely the part I can’t settle, and an argument I could settle would usually be one not worth making at this length.

So the pieces here carry limits sections, concessions I don’t recover from, and endings that name what I still don’t know. That isn’t a stylistic preference or an affectation of rigour. It’s the only form in which those arguments are honest, and there is nowhere else I could put them.

Which leaves something slightly uncomfortable about my own publishing model, and I’d rather say it than let somebody else notice. Every LinkedIn post I write is an extraction from one of these pieces, and extracting means compressing something that stays open into something that closes. The post is structurally the less honest version. I don’t think that makes it dishonest — a post is a door rather than a room, and nobody expects a room behind a door they haven’t opened. But anybody who only reads the posts is getting the settled version of arguments that aren’t settled, and the remedy isn’t writing better posts.

What it actually is

Part publication. Part notebook. Part thinking tool.

I should be honest that this arrangement suits somebody with twenty years of material to draw on and a system already built to hold it. Starting from nothing, a public notepad is a much harder thing to make useful — you need something to put in it.

But the version of me who had observations and no time to develop them wasn’t producing better thinking. He was producing less of it, and forgetting most of what he did produce.

AI helps me get what’s in my head somewhere I can inspect it.

Field Notes gives those thoughts somewhere to live.

And occasionally somebody else finds one useful too.