AI cold showers
Sometimes I’m AI-pilled and want to AI all the things, all the time, and sometimes I need to take a cold shower and remind myself that these are powerful tools that can wreak havoc in a multitude of ways, perhaps most importantly on what makes humans, human. Here’s an evergreen list of articles—and one quote from each—that ground my enthusiasm around AI and help me be a responsible user of AI.
The current generation of agentic systems is built around the premise that the human is the bottleneck — that the loop runs faster and cleaner without the awkward delay of someone reading what is about to happen and deciding whether it should. This is, in a great many cases, exactly backwards. The human in the loop is not a vestige of an earlier era; the human is the only part of the loop with skin in the game. Removing the H from HITL is not an efficiency. It is the abandonment of the only mechanism the system has for catching itself.
What’s at risk is nothing less than the ability to think for oneself. If people become overreliant on AI to write for them, they could lose the capacity to interrogate or even develop their own views. These are quintessentially human capacities.
I proposed a design, and a teammate prompted an AI to critique it. The teammate sent an AI document to me, with the disclaimer: “I didn’t read this, so it might not be entirely accurate”. My thought was, if reading this wasn’t worth your time, why is it worth mine?"
A “centaur” is a human being who is assisted by a machine (a human head on a strong and tireless body). A reverse centaur is a machine that uses a human being as its assistant (a frail and vulnerable person being puppeteered by an uncaring, relentless machine).
LLM-generated writing undermines the authenticity of not just one’s writing but of the thinking behind it as well. If the prose is automatically generated, might the ideas be too? The reader can’t be sure — and increasingly, the hallmarks of LLM generation cause readers to turn off (or worse).
Lines in the sand
The common theme across these posts is about losing something that makes us human. They tease possible futures that make me feel like I’m watching myself or my peers descend into an uncanny valley in 4K slow motion. I need to draw the line somewhere.
Maybe that makes me a Luddite, maybe I’ll revisit someday when Fable 6 becomes self-aware, or maybe I’ll give up and embrace our robot overlords. I’m truly not sure, but I think that’s why it’s important for me to write these in the first place.
These are personal choices that I’d never dare impose on others, but if you do catch me drifting from these and especially if you catch me lobbing slop your way: call me out on it and link this article. I promise I’ll be receptive to it. Until then, here are the lines I’ve drawn on a shore of AI tides and hype itching to wash them away.
✍️ I won’t let AI write for me
I’m unafraid to say that AI is at the point where it’s a far better coder than I am or ever will be. It’s probably a better writer as well, but… that’s not the point.
Maybe I’m clutching my pearls, but I find it deeply unsettling to relinquish my voice and all the thinking I put in when I write. I often write to clear and synthesize my thoughts, and giving that up for the sake of short-term productivity seems like a long-term net-negative.
So here’s a public pledge: I won’t ever send you any written words that were written by AI, unless I explicitly quote or attribute it.
All the content on this site, all my emails, all my pull request prose, all my typos and grammatical errors and incorrect emdash usages—in all their wabi-sabi glory—you can be certain genuinely came from me.
📖 I won’t let AI read for me
Or, perhaps more narrowly: I won’t read AI summaries as a substitute for reading the source material with my own damn eyes. (Strong caveat on the latter: provided the source is not AI slop itself, in which case I just won’t read it. See the article on human effort above…)
To be clear, there’s nothing here that changed from before LLMs. I treated human-written summaries the same way. I hope you treat my summaries the same way, too. I think it’s fine to use summaries and synopses to see if something might be worth reading, but I believe it’s a mistake to let them, AI-written or otherwise, tell me what to think. Reading a conclusion and arriving at a conclusion couldn’t be more different.
In the immortal words from Brandon Sanderson:
Journey before destination
📝 I won’t let AI take notes for me
I really can’t recall a time that I was more productive from AI-written notes than if I had written those notes myself. It felt cool at first. It felt like I had a personal archivist or assistant. But I don’t think I’ve captured more value than I lost.
I think the acts of both writing and then later retrieving those notes provides intrinsic value for learning, growth, and thinking. (Also see: Learning How to Learn.)
I, perhaps selfishly, write notes for me and it’s critical to my thinking process. I’m not ready to let that go.
I also think it’s more true than ever that:
We are drowning in information but starved for knowledge. —John Naisbitt
Things like AI meeting notetakers might be capturing everything, but are they really surfacing the right things?
To be continued…
This article isn’t done. I’m not sure if it will ever be. I’ll update it as my beliefs change. They surely will. Until then, thanks for coming to my TED talk. ✌️