I read an essay today: I Wanted to Own the Harness. Then Codex Desktop Won. It is seemingly about someone who has fully integrated AI into their workflow — and their life — and finds it indispensable. Cory Doctorow would probably call him a “centaur”.
Cory Doctorow has written extensively about two ways that people interact with AI. Some people are “centaurs” and use AI to enhance and extend their own abilities. Others, who are not so lucky, are “reverse centaurs.”
Reverse centaurs are people that live in a Taylorized hellspace where the AI is using the person to extend its abilities.
In automation theory, a “centaur” is a person who is assisted by a machine. You’re a human head being carried around on a tireless robot body. Driving a car makes you a centaur, and so does using autocomplete.
And obviously, a reverse centaur is a machine head on a human body, a person who is serving as a squishy meat appendage for an uncaring machine.
Like an Amazon delivery driver, who sits in a cabin surrounded by AI cameras, that monitor the driver’s eyes and take points off if the driver looks in a proscribed direction, and monitors the driver’s mouth because singing isn’t allowed on the job, and rats the driver out to the boss if they don’t make quota.
The driver is in that van because the van can’t drive itself and can’t get a parcel from the curb to your porch. The driver is a peripheral for a van, and the van drives the driver, at superhuman speed, demanding superhuman endurance. But the driver is human, so the van doesn’t just use the driver. The van uses the driver up.
Obviously, it’s nice to be a centaur, and it’s horrible to be a reverse centaur.
I say above that the article is “seemingly about someone who has fully integrated AI into their workflow” because I can barely understand what he’s saying.
As a long-time programmer and technical user, I’m generally familiar with technology. However, as an author and educator, I’ve personally avoided using generative AI. I’ve written a handful of posts explaining why. Which isn’t to say I haven’t tried to generally keep up with the field. That said, I found reading this post almost totally incoherent.
I get that he started out looking for a way to maintain control of his his front end and workflow and still be able to switch among back-end models, in order to not get locked into a single vendor. That makes sense. But then he quit worrying about that as a goal because this setup required less of his attention? I think?
I do understand that he has a little dog (a Phalène) on the screen that sounds a lot like the Talking Moose. Except maybe it actually listens to you? I installed an early version of the Talking Moose. It did not measurably improve my productivity.
He goes on to say things like this:
In its own agentic benchmark, Ponytail cut source code by 54 percent across twelve Haiku 4.5 feature tasks and passed all twenty adversarial safety runs. It also used 22 percent fewer tokens, cost 20 percent less, and finished 27 percent faster. Its large wins came when a native HTML input replaced hundreds of custom lines. On irreducible backend work, it barely changed anything. The mechanism is visible in the diff.
What?
My brother said, “Cutting it by 50% would leave you with 102 syllables. Cutting it by 54% would leave you with about 94 syllables.” I guess I’m only familiar with Haiku 1.0.
Anyway, this excerpt is selected more or less randomly. I could have picked any of another dozen paragraphs that are equally orthogonal to anything in my experience and that use a bunch of terms I would have to go off and research to really understand what they mean. (Years ago, I might have used “grok” here, but not any more. Thanks, asshole trillionaire.)
I’m glad it’s working for him, but I’m not sure I’ll ever be able to speak centaur. Honestly, I’m not sure I really want to either.