6/4/2026
I Built This With Zero Lines of My Own Code. And That's Exactly Why It Matters.
I should tell you something upfront: I cannot read code. Not really. I can recognize a function, vaguely sense when something is wrong, but I couldn’t write a line by hand if you asked me to. Every line in Afterglow — the app behind this waitlist — was generated by an AI.
I am the bottleneck. And I think that’s the point.
Why I started
Honesty first: I built this to make money.
Not because I had a burning mission or a decade of product experience. I saw a gap — a note app that didn’t ask you to maintain it — and I thought I could build it. So I started. And somewhere along the way, a strange thing happened: Afterglow became the tool I actually use. More than Obsidian. More than anything else.
Which tells me the gap was real.
What I actually did
Here’s what the process looked like. I described what I wanted. The AI built it. I used it, decided what felt wrong, and described the next thing. Repeat.
Planning, taste, judgment — mine. Implementation — not mine.
This sounds like it should feel hollow. Like I’m a fraud who stumbled into a product. But the more I think about it, the more I think the framing is backwards. The question isn’t whether I wrote the code. The question is whether I knew what to build.
And that knowledge — what to build, why, for whom, in what order — came from somewhere. From years of reading widely across domains. From being the kind of person who loves stuffing their brain with disconnected facts the way a squirrel stuffs its mouth with acorns, not knowing yet which ones will matter. From caring about typography and silence and the feeling of a tool that gets out of your way.
That knowledge is not in the model. The model generated the code. I steered it.
The price of information is falling
We are living through the moment when information becomes nearly free.
Not worthless — but cheap. Cheaper every month. The cost of generating a competent paragraph, a working function, a structured argument, approaches zero. LLMs are probabilistic slot machines, and they are getting very good at producing the kinds of outputs that used to require years of specialized training.
This should feel threatening to anyone who built their value on knowing things.
But I think it’s the opposite.
What actually gets expensive
When information is free, the scarce thing is judgment.
Not the ability to recall facts — the ability to know which facts matter. Not the ability to generate text — the ability to know when the text is wrong. Not the ability to build something — the ability to know what’s worth building.
The bottleneck in AI-assisted development, I’m told by people who do this at scale, is the human. The same prompt, given to ten different people, produces ten wildly different outcomes. The model is the slot machine. The person is the unscratched lottery ticket.
That lottery ticket is made of something. Taste. Domain knowledge. Years of reading things that didn’t seem useful at the time. The ability to recognize quality because you’ve seen enough of it to know what it looks like.
This is what Afterglow is for.
The tool I was missing
Every note app I used was optimized for output: shipping faster, organizing more, getting through the inbox, making information accessible. Notion for teams. Obsidian for systems-thinkers who enjoy configuring systems. AI assistants for producing more text, faster.
None of them were optimized for the slow, quiet work of actually understanding something.
The Zettelkasten method — which Afterglow is built around — is about connection. Not storage. You write a note not to remember a fact but to integrate it: to find where it touches something you already know, where it contradicts, where it extends. The value isn’t in the note. The value is in the network.
That network, built over years, is what lets you steer.
A word of honesty about the future
I should be honest: I’m a singularitarian. I believe AGI is coming. I believe the moment it arrives, most of what I’ve said above will need to be revised.
But that moment hasn’t arrived. And until it does, the judgment of value is still a human act. What is worth knowing. What is worth building. What is worth your time. These are not questions the model answers — they are questions you bring to the model.
And the quality of those questions depends on the depth of the knowledge behind them.
What this means for Afterglow
I didn’t build Afterglow as a hedge against AI. I built it because I needed a quiet place to think — one that doesn’t ask me to manage it before I can use it.
But the timing feels right. We are entering a world where the ability to generate is universal and the ability to judge is rare. The person who has spent years building a dense, well-connected library of their own thinking will have something the model doesn’t: a point of view.
Afterglow is a place to build that.
Not for teams. Not for productivity theater. Not for people who want an AI to do their thinking. For people who want to do their own thinking — more clearly, more deeply, in a tool that gets out of the way.
The acorns matter. The squirrel still has to decide which ones to bury.