Think you can dump your scraps into the trough of an LLM and generate something like this? Good fucking luck, pal.
That’s my posture right now. I just finished writing a book. For now, I’m proud of it. There’s a little afterglow, a little bravura.
But…
Just as I’m hitting my stride with writing, I can’t shake the feeling that I’m doing something that will soon seem archaic. Like writing in cursive, or getting good at ground-level cartography the year before we start flying planes and taking pictures from the sky.
As I put a bow on this project, I keep thinking about The Sopranos. In the pilot, Tony says, “Lately I’ve been getting a feeling that I came at the end. The best is over.” We meet him at the top of a dying empire, a world vanishing. I’m nowhere near the top of anything. But like Tony, I feel like I’m arriving at the end of the party, trying to make sense of what’s being lost.
The book was a ratcheting down of notes, highlights, ideas, and philosophy I’ve been curating for decades. I squeezed an orchard of fruit, boiled the juice, and now here’s this piece of hard candy. Hope you like it.
But that process of reading voraciously, highlighting, note-taking, synthesizing, iterating, etc., feels like exactly what LLMs are on their way to doing better than me. Not just faster. Better.
Maybe I’m just nostalgic for my journey of wandering through books with a pen, of journaling to converse with my past self, of experiencing aha moments late at night… Because those are happening less often for me. Every time I have an idea, my reflex is to bounce it off an LLM.
I use AI to proof my work, and that sped up the book editing process. Having an instant sparring partner and proofreader, even if clumsy, is incredible. LLMs are a backstop for so much of the tedium of writing, but they’re creeping into the fore. Right now, I tell myself I use AI “just for proofing and catching typos.” But every interaction is training data for my replacement. And the question is, when will it take over? When do I stop being the thinker, the assembler?
It’s always been easier to broadcast a thought than to guarantee it would be understood. Now, with AI suggesting what we should say and how we should respond, there’s even less of an expectation for us to pursue nuance, to understand what we’re thinking, what we really mean.
This didn’t start with AI. There was the like button, autocomplete, emojis, memes, prewritten greeting cards, canned responses, on and yawn. We’ve been slouching toward convenience over contemplation for so long that as AI offers to do our thinking entirely, it feels like a natural progression.
There’s something irreplaceable about the struggle and discovery that happens when you force cloudy thoughts onto a page. Lo, that feeling when the paragraph finally sings after the tenth rewrite. But I can’t think of reasons why people will still read and write “manually,” especially long-form content, without leaning hard on AI. Why struggle? I have my reasons for writing, but they’re personal and unlikely to appeal broadly when machines can do heavier work faster.
Maybe this is just the vertigo that comes with finishing a difficult project. Maybe I’m lining up excuses for this book to pass like a fart into a windy world. I did something that felt challenging and worthwhile, and now I’m scrambling to justify the time spent. I hope the book (and the medium of personal writing and blogging) helps others grow faster than I did. That’s all that really matters.
I’ll always write. Humans will always write. But the incentives are thinning out.
It just feels like we’re there… at the end of something.
