LLMs are a helluva drug I guess
Published by marco on
The article AI Doesn’t Reduce Work—It Intensifies It by Simon Willison demonstrates the Overton Window of addiction pretty well. The author writes,
“This captures an effect I’ve been observing in my own work with LLMs: the productivity boost these things can provide is exhausting.
“[…] I’m frequently finding myself with work on two or three projects running parallel. I can get so much done, but after just an hour or two my mental energy for the day feels almost entirely depleted.”
Is it marketing?
If I didn’t trust the author, I would immediately suspect that he’s applying a classic marketing tactic: “OMG this tool is so powerful that I can’t even control it!”[1]
This is a classic tactic of trying to sell a product by arguing that you probably shouldn’t use it, not because it doesn’t work, but because you probably couldn’t handle it. This is the kind of marketing that appeals to children, teenagers, and Jackass fans.
Instead, I think he might be seriously not noticing that his argument amounts to, “All of this cocaine I’m doing has doubled my productivity but I can only work a quarter of the day. Also, I feel like shit.”
Maybe you’re using it wrong?
A snarky response would be: “Hey! Here’s an idea I’ve heard somewhere: maybe you’re not prompting it correctly.”[2]
But I’m not going to be snarky.
You’re not special
The author continues to describe what he clearly seems to think is a unique phenomenon. It’s only unique when you’re trapped in an information bubble where you start to attribute every detail of existence to the thing that you have grown to unreservedly love.
“I’ve had conversations with people recently who are losing sleep because they’re finding building yet another feature with “just one more prompt” irresistible.”
My friend, you are describing addictive behavior.
It was no different before LLMs. This is how it has always been with programming (or for any interesting task, really). I spent a lot of my early 20s programming day and night.
As you get older, though, you learn that just leaving it be is just as efficient. That is, instead of staying up two or three more hours growing increasingly frustrated, you can finish whatever you’re working on in the morning—and probably in five minutes. That’s almost always the more efficient and sustainable solution.
But, sure, let’s pretend that this behavior is unique not only to programming, but to programming with LLMs.
“I think we’ve just disrupted decades of existing intuition about sustainable working practices. It’s going to take a while and some discipline to find a good new balance.”
“We’re all doing too much cocaine, right guys?”
Sure we are, Simon. Sure we are.
[1] While I trust that the author himself is more likely to have been duped than being deliberately misleading, I don’t extend the same generosity to others making the same argument, most especially any of those who actually work for companies where their stock options increase whenever people believe that the thing they’re working on is revolutionary. I will continue to discount the obvious grifters, even while I extend Simon the benefit of the doubt that he isn’t grifting, even when he ends up sounding very much like a grifter.↩
[2]
For those not steeped in the lore of gaslighting skeptical programmers, that’s the answer that LLM fanboys inevitably have for anyone who asks why they don’t feel more efficient when using LLMs, why the tools seem to deliver so many incorrect answers that make more work not less.
You’re not prompting it correctly.
↩