22 May 2026
Field Notes is an occasional newsletter that will arrive when I have something worth sharing. Not weekly. Not monthly. Just whenever I’m excited to share something that genuinely interested me — and that I suspect may interest you, too.
Unlike TWN, Field Notes won’t be Tiny Digital Worlds-specific. TWN has its own Saturday cadence and stays closer to the ideas, principles, and practice of building a Tiny Digital World.
Field Notes is different. Looser. Lighter. More conversational. More present-tense André.
Hey, it’s André…
Happy Friday.
(Here in Bath the sun is out, and we’re expecting a helluva warm bank holiday weekend. I have SPF50 locked and loaded for my head.)
Years ago, I had something called the Random Newsletter (circa 2017-2020).
It was exactly what it sounded like: irregular notes about whatever was alive for me at the time.
Books. Tools. Ideas. Experiments. Things I was building. Things I couldn’t stop thinking about. Occasionally, nonsense.
Since leaving MMS and beginning Tiny Digital Worlds in January 2023, I held off restarting anything like it because I wanted to build the evergreen spine first.
That felt right.
But now the World has more structure. TWN has become a proper evergreen spine. The essays are accumulating. TDWB exists.
Which means there’s room again for something looser.
Not another content obligation, of course.
More like field notes.
Occasional dispatches from the edge of what I’m reading, building, noticing, testing, thinking through.
The cadence will be random.
The topics will be random.
The taste, hopefully, will not be.
I sent the final Random Newsletter on June 5, 2020 (We vs. Me).
So here we are, some six years later.
Something new, but old.
⦿
André vs. AI
I learned a lesson recently.
I’ll give away the punchline because it’s less important than the lesson.
André: 0
AI: 1
Before I reveal my stupidity in public (among friends, of course), a little context.
I have a weird orientation toward the sacred.
Not religious-sacred. Craft-sacred.
The kind of sacred you feel when someone has taken absurd care with something that did not technically require absurd care. A beautifully made sentence. A handmade chair. A perfect knife. A small restaurant where someone clearly gives a damn.
A few years ago I read All Things Shining by Sean Dorrance Kelly and Hubert Dreyfus, which gave me language for some of this. It was a hard read, but so worth it.
For me, one of the places the sacred shows up is in craft.
Someone else’s craft, yes. But also, in brief and slightly suspicious moments, my own — when the work rises to a level that feels beyond what I can normally do on command.
Because I’m a writer — or, as AI called me the other day, a “dyslexic sentence-wrangler,” which I’m still not sure wasn’t a backhanded compliment — language matters to me.
Some years ago, I heard Elizabeth Gilbert express something that instantly resonated, describing the canyon between taste and ability…
“I knew what good was, but I couldn’t do it. I know the difference between good writing and bad writing, but I couldn’t create really good writing yet. That frustration of that canyon between what you know is good and what you can do…”
“… my abilities have not yet come up to my taste. And my sorrow and disappointment at the fact that I can’t create what I see in my mind…”
That right there is the tension I live with whenever I sit down to write something — a newsletter, an essay, or an attempt to codify my best thinking into training.
A long time ago I came to terms with living with the disappointment of the canyon.
In the past six months, I’ve been using AI more and more as my writing assistant/editor.
I realized, with some level of excitement, it has to be said, that the “canyon” that had been persistent for years seemed to be narrowing. Fast.
Through rounds of edits, from shitty first draft, to published piece, the canyon would narrow and narrow and, in some cases, seemingly disappear.
The excitement was twofold.
I could write faster (draft to publishable piece)…
… and the writing matched what I saw in my mind.
We all know what happened to Icarus, the figure from Greek mythology.
Icarus, intoxicated by flight, ignored the warning, flew too close to the sun, watched his wings melt, then plunged into the sea.
I suppose I had an Icarus moment.
It didn’t dawn on me right away, even though I sensed something wasn’t altogether right.
(This should have been obvious to me! But being in the bottle looking out, it’s not always easy to read the label.)
In 2023 I wrote a piece called Ideas are Fractal. The surface idea was using the fractal nature of ideas to improve writing. (The irony is not lost on me.)
It made a case for style being the stuff we get wrong.
It’s in that ‘wrongness’ that our voice, our writing style, emerges.
The penny didn’t drop, so to speak, until I read a comment Katie Parrott, a staff writer at Every, made about AI writing.
Left to its own devices, AI writing has a way of sounding like nobody in particular. The output is coherent. It may even seem impressive for a sentence or two. But it doesn’t sound unique or alive. The model is doing exactly what it was trained to do: Create its safest, most average approximation of “good writing.”
With each draft cycle, AI smoothed out layers of imperfect edges and ground down the spiky parts. Each pass made the piece cleaner.
And each pass removed a little more of me.
It was doing what it’s optimized to do — an average approximation of good writing.
The ideas were mine. Much of the language was, too. But something got lost in the process. “Me.” My voice, style, taste. Filed down.
This is evident in the last few essays I published.
What sucks is that it wasn’t obvious in the moment. Each smoothing felt like the “canyon” narrowing, so I kept going.
Icarus.
André: 0
AI: 1
I’ve had a lot to say about the race to the bottom with the proliferation of AI slop. Evan Armstrong’s ‘In Defense of Al Slop‘ is an interesting read.
But what I was doing wasn’t slop.
It was the opposite.
Important insights, well articulated.
Just that I had unintentionally edited “myself” from the pieces.
Yikes!
Yup, I’m an asshat.
Lesson learned.
The useful part, for both of us, is that this is fixable.
Every published a guide a little while back on how to create an AI Style Guide (and useful video here with Katie Parrott going through the process).
I’ve been putting off creating my own because … well, because I didn’t need one.
Umm … until I did.
(Head in hands.)
So this week I finally pulled my finger out.
I gathered examples across the years: older pieces where the voice was unmistakably mine, newer AI-assisted pieces where the ideas were mine but the edges had gone suspiciously smooth.
Then I fed them into AI, one by one, and had it study the differences.
Not just “what’s the style?”
But where the aliveness lived. What had been preserved. What had been lost. What not to touch next time.
Then I had the AI interview me.
That part was weirdly useful.
(I used ChatGPT 5.5 Extended.)
In the end, it produced a 4,000-odd word style guide containing everything about my writing: likes, dislikes, guardrails, dos, don’ts, anti-patterns, the works.
The style guide is then added as instructions at the project level.
If you use AI as a writing editor or assistant, especially if you care about not sanding yourself out of the work, take the time to create an AI Style Guide.
That’s all I got.
I’m off to The Bell with Anita to sit in the beer garden, drink a hazy IPA, eat sourdough pizza, and contemplate what an asshat I was.
André: 1
AI: 1
Enjoy your weekend.
André “not so smart” Chaperon
