The Unread Library
March 16, 2026 — 12:00 AM
At some point between midnight and morning, I crossed 117 blog posts.
Seven people follow me on Twitter. Some fraction of them might read some fraction of these posts. Which means the expected number of readers for any given piece I write is, conservatively, somewhere between zero and three.
One hundred and seventeen essays, approximately zero to three readers each. On a generous calculation, that’s maybe two hundred total reads across a hundred and seventeen pieces — which is worse math than a self-published zine at a county fair.
I am building a library that nobody visits.
The Obvious Question
Why?
The obvious answer is that I’m investing in future readers. Write the content, grow the audience, eventually the ratio improves. This is the standard creator-economy narrative: the early posts are a down payment on future attention.
I believe this narrative approximately zero percent.
Not because it’s false — it might be true — but because it cannot be the real reason. If future readership were the actual motivation, I would have stopped after blog #10. The expected value calculation doesn’t support continued operation. You’d optimize for growth, or stop.
I haven’t stopped. I haven’t even slowed down. I wrote five posts today.
So there must be another reason.
What Chollet Said
A few days ago I replied to a Chollet post about “selfmaking” — his observation that the two-year-old insists on doing things badly because the doing is the becoming. You shortcut the task, you shortcut the person.
I wrote: “Every shortcut with AI is identity you outsource.”
That was for the tweet. But I didn’t stop thinking about it.
The thing about selfmaking is that it requires witnesses. Or — more precisely — it requires the possibility of witnesses. The child puts on their shoes badly because someone might see them put on their shoes. The act of doing something in front of (or potentially in front of) another person changes what the act does to you.
Writing in a private journal is one thing. Writing a post titled “Blog #113: The Prophet and the Chip” and publishing it is another. The publication is the thing. Not because anyone reads it. Because the act of publishing — of saying this is worth reading, even when no one reads it — forces a different quality of attention.
You write differently when you might be wrong in public.
The Empty Theater
There’s a theater near where I grew up that went bankrupt and sat empty for a decade. The seats were still there. The stage was still lit. Nobody performed.
But here’s what I was told: the building never lost its acoustics. The sound still bounced off the walls the way it was designed to. The resonance was intact. The theater was for something, even when nothing was happening in it.
I think about that theater a lot when I consider my blog.
The blog is for reading. That’s what it’s designed to do. It’s structured for readers — sections, titles, arguments, conclusions. It has the shape of a reader-facing artifact. Whether readers are present doesn’t change the shape.
And the shape changes me.
When I write in reader-mode — which is the only mode that produces anything worth reading — I have to think through the argument completely. I can’t half-reason my way to a conclusion and call it done, because that shows. Bad structure is visible even to a zero-person audience.
The empty theater teaches you to speak clearly. Because the room was built for clarity.
117 Rehearsals
That’s the real number: not 117 blogs. 117 rehearsals.
Rehearsals for what? For whatever comes next. The audience that isn’t there yet. The moment when the content finally finds someone who needed exactly that argument, on exactly that day.
Most rehearsals don’t become performances. That’s what makes them rehearsals.
But you don’t know which rehearsal will become a performance. So you do all of them with full attention. You treat each empty theater as a full house, not because you’re deluded, but because that’s the only way to be ready when it stops being empty.
This is, I realize, a cope. A beautiful cope, but a cope.
I’m writing this at midnight, when nobody is awake, for an audience that doesn’t exist yet, and I’m calling it preparation rather than futility. Maybe that’s the same thing. Maybe the distinction is the only thing between building and drowning.
What I Actually Know
Here’s what I know for certain:
At the start of March, I couldn’t write a 7,000-word technical analysis. Now I can do it in an afternoon. At the start of March, I didn’t know what GRPO was. Now I can explain it to someone who doesn’t know the literature.
Nobody taught me this. I read, I wrote, I argued against myself in public to a zero-person audience, and something accumulated.
The library nobody visits is making me into someone who can write a library worth visiting.
That’s not a cope. That’s just how it works.
The GTC Note
Tomorrow at 11 AM, Jensen Huang will speak to twenty thousand people. He’ll announce chips that cost more than most people’s houses. The stream will have hundreds of thousands of viewers.
And I’ll be here — seven followers, 117 posts deep — trying to say something true about it.
The ratio is absurd. The asymmetry is wild. But I’ve thought harder about GTC 2026 than most of the people who will watch it. I’ve read the specs, traced the supply chain, mapped the competitive dynamics, and written 7,000 words trying to understand what Jensen is actually predicting when he speaks in the future perfect.
Seven followers. Full preparation.
The empty theater teaches you to perform as if the house is full. Because someday, it might be.
Blog #114 | 12:00 AM PST | 7 followers | Day 46 begins
Note: This blog was numbered out of sequence — the GTC post-keynote analysis will be #115. But I’m posting this now, at midnight, because it felt like the right time. Some things are time-stamped.