A diary that answers back — Editorial Intelligence

I think I’ve accidentally turned my holiday into an Editorial Intelligence experiment.

I’ve been having a ridiculous number of ideas, partly because Sage isn’t occupying most of my head and London life isn’t taking up the rest of it. A smartphone, time to think, and enough distance from the usual routine for things to start connecting.

So I’ve been letting them flow. A thought about AI. Something about getting older. A film. RimWorld. Thailand. Work. Writing. A comment somebody made.

Most start as a few scrappy lines typed into my phone. But instead of going into Notes and being forgotten, they go into my Editorial Intelligence system.

That changes what a note can become.

The system answers back

It asks questions. It connects the thought to something I was exploring two weeks ago. It pulls on an argument I hadn’t quite articulated. Sometimes I disagree with where it takes the idea and push it somewhere else.

Eventually some of those conversations become posts. Others become frameworks, field notes, or things I decide should stay private.

Which is a strange thing for me to be describing, because I could never keep a diary. I tried more than once, and the mechanics defeated me every time. The sitting down. The deciding what counted as worth recording. The upkeep of a habit whose only reward was a stack of pages I was never going to reopen.

I liked the idea of the apps built for it, too. Day One is a well-made piece of software and I wanted it to work. But what it produces is still a passive record, dead the moment you’ve written it. Searchable, timestamped, neatly presented, and entirely uninterested in what you happen to be thinking about today. That isn’t a flaw in the app. It’s what the format is.

Even when I did write something properly, it stayed where it was. It didn’t change what I wrote next. A finished piece in a folder isn’t a living record, and old writing is only worth anything if it’s contextual and interlinked, which is this whole website’s argument applied to something much smaller than an organisation.

This one reaches back.

So it feels surprisingly cathartic. Almost like a modern diary, except that a traditional diary is mostly:

experience → write it down → archive it

This is closer to:

experience → fragment → conversation → interrogation → connection → evolving record

Which makes it an Editorial Intelligence experiment in its own right.

One of the ideas behind Editorial Intelligence is that knowledge shouldn’t sit in documents waiting to be rediscovered. Research, interviews, observations, previous arguments and unfinished thoughts can stay active, and participate in whatever you’re thinking about next. I’ve argued that a knowledge system stops being an archive once it can work across enough accumulated context.

Apparently that applies to one person’s thinking as readily as to an organisation’s. The surprise isn’t that it works. It’s that the unit is small enough to be a single person on holiday with a phone.

I was missing a layer: Daily OS

Coming back to this after the holiday, I realised I had described the interesting part of the system without naming the boring part that makes it work.

The conversations themselves are not the shared memory.

They are the intake.

A lot happens in a day that is useful for a few hours and irrelevant afterwards. Some of it matters: a decision, a changed opinion, a useful observation, an application I decide not to pursue, a problem that keeps recurring, an experiment that worked. If every interaction becomes permanent memory, the system fills with noise. If none of it is captured, the useful context disappears with the conversation.

I’ve started thinking about a Daily OS as the layer between those two states.

Its job is not to preserve a transcript of my life. It is to capture the consequential parts of the day cheaply enough that I will actually do it: what happened, what I decided, what changed, what remains open.

Then the durable parts can be promoted into the longer-lived system I think of as shared memory: the career file, the project record, the argument, the evidence base, the preference or constraint that should affect what happens next.

So the architecture is becoming clearer:

conversation → Daily OS → shared memory → future conversation

Or, in slightly more practical terms:

activity → daily capture → promotion → reusable context

That distinction matters. A daily log is chronological. Shared memory is editorial.

The first records enough of the stream that important things are not lost. The second applies judgement about what deserves to survive and where it belongs. GitHub is useful here not because Git itself is magical, but because the record is inspectable, portable and not trapped inside one AI provider’s memory system.

This also makes the idea of a “save point” more concrete. The save point is not just a periodic export of finished knowledge. It can be continuously rebuilt from a low-friction record of the decisions and observations that created that knowledge in the first place.

The failure mode is obvious: logging everything would turn this into administrative theatre. The Daily OS only works if compression is built in. Most conversations should vanish. The system should preserve the few things that would make tomorrow’s reasoning better.

That is probably the connection I hadn’t made explicit before. The diary is not the memory. It is the feedstock for it.

The published pieces are inside the record, not outside it

There’s a stage I left off that sequence, because I hadn’t noticed it yet.

I assumed publishing was where a thought stopped. You work it out, it goes up, you move on to the next one. That is not what has been happening. A large share of what I now do is going back into pieces that are already live — adding an argument that only arrived afterwards, naming a mechanism I had asserted without explaining, folding in a draft that turned out to be a restatement of something I’d already published.

Which is odd, and worth being honest about rather than presenting as a method. I appear to be in a continuous argument with things I wrote a week ago.

The explanation I find most convincing is dull. These pieces are the only place the thinking is properly worked out — not the notes, not the fragments, the published version. So going back to one is not maintenance. It is the same activity as having the thought in the first place, resumed. Nothing was ever finished. It was left at a reasonable stopping point.

That reframes what the archive is for. Not a record of what I once thought, which is what a diary produces and part of why I could never keep one. Something closer to a working position that happens to be legible in public while it is still being argued about.

I’d flag the obvious failure mode, because it applies to me directly. Endless revision feels productive and frequently isn’t — polishing is much more comfortable than starting, and a body of work you keep returning to can become a way of never risking anything new. The test I’d want to hold myself to is whether a return adds an argument or just improves some sentences. The first is the system working. The second is avoidance with good grammar.

Twitter was a diary too

It was a short-form public one. Thoughts recorded at roughly the speed they arrived, in front of anybody who happened to be looking.

What’s changed is what comes out of the other end. The cadence is the same rapid fire, but instead of a fragment I get a finished argument, because the thing that used to make speed and completeness mutually exclusive has gone.

It’s also entirely mine to judge. My website, no editor sitting between the thought and the publish button. That’s the freedom, and it’s obviously also the risk. I’ve already written that a body of work built mainly from one practitioner’s experience can become a very convincing account of how that practitioner thinks the world works. Producing it faster doesn’t fix that. It’s an argument for the research, not against the pace.

I’m not pretending I’ll keep this pace. As I’ve said elsewhere this week, the conditions are temporary and normal work resumes. So I’m treating it as a sprint. Capture while the ideas are coming, build the archive, and see what survives.

And it’s an experiment, in the same way Editorial Intelligence is an experiment. It may work and it may not. I’m going on what I feel like doing at the moment, and I’ll write about how it turns out either way, including if the answer is that a fortnight of this produced very little worth keeping.

Which is perhaps another thing AI changes about writing.

The notebook doesn’t have to be where an idea goes to sleep. It can be where the idea keeps thinking with you.