Thought Lag
Three years into the AI era and here I am, sitting at my iPad, no agent running, no LLM prodding me and …boom, I hit a lag. A noticeable delay between the thought and the sentence that didn't used to be there.
Turns out, no surprise here, there's a debt to pay when you outsource the first draft of the thinking. Atrophy sets in. The part of the brain that used to just start — messy, imprecise, no other option — that part has gotten quieter. It still works. But I have to wake it up now.
BLLM (before LLMs), I just started writing. Never outlined. Never waited for the conditions to be right. The first sentence was always a little wrong and that was fine because the second sentence corrected it. That's how it worked.
Now I pitch ideas to Claude. A prompt or two. Sometimes I share a snippet that inspired me. I throw thought vomit into the chat box knowing my trusty LLM can handle whatever mess I give it.
Then I shape it — with Claude — from there.
And dammit if I'm being honest, the output is better. Thoughts come out clearer, word choice a little snappier. It's better — but is it me?
That's the question many folks are asking these days.
So I went looking for a way to protect my voice.
I ran a hundred-question interview on myself to build a voice profile doc. The document captures things I couldn't have articulated if you'd just asked me — patterns I didn't know I had until I saw them named.
But a voice profile is, by definition, backward-looking. It's who you were when you answered the questions. But we are all impermanent. We do not stay still. The things I care about this year aren't identical to the things I cared about two years ago. The edges that felt important then might feel inconsequential now.
If I let the profile do too much of the protecting, I'm not preserving my voice. I'm taxidermying it.
A taxidermied voice is still recognizable. It just can't grow.
There's a version of this that worries me — the AI becoming the arbiter of what sounds like you, the feedback loop running long enough that the "you" it's protecting starts to calcify, until the voice profile stops being a snapshot and turns into a ceiling instead.
Because the humanity the system can't replicate is writing on your own when you don't know what you're going to say. The discovery version where the third paragraph tells you what the first paragraph was actually about. That process doesn't just produce writing, it produces the writer. Every time you do it, you find out something about where you are right now that you couldn't have gotten any other way.
I'm less concerned with AI making me sound like someone else. I've built a system to protect that. But I am afraid that I'll stop doing the thing that keeps me sounding like myself.
So I make a point to keyboard away — no Claude, no prompt, just the blank page and my self doubt. The writer's block so many have feared but now I long for. I don't do this nearly enough, but it's as intentional a practice for me as phone-free moments are for others.
Because I know what I lose when I don't.