When coordination breaks down, the fantasy of self-sufficiency rushes in to fill the gap. GenAI, after all, is built as an individual technology, whether it is expressed as a one-on-one chat, the fantasy of the one-person startup, or the sycophantic assistant set up to glaze you. It says “just get it done, no skills needed.”

What surprised me wasn’t the AI hype, though, but the lack of solidarity that came with it. Faced with the story of AI labor displacement, our first instinct as technology workers wasn’t to protect one another, but to search for ways to use the tools to replace our collaborators.

In other words, instruments can surprise you with what they offer, but they are not automatic. In the end, they require a touch. You use a tool, but you play an instrument. It’s a more expansive way of doing, and the doing of it all is important, because that’s where you develop the instincts for excellence. There is no purpose to better machines... See more

As the tools evolve, the metaphors we use to understand them must also be updated. Herndon and Dryhurst describe this next phase as the move from sampling to spawning. Sampling was the logic of the 20th century. You took a slice of a record—a James Brown breakbeat, a horn stab from a jazz LP—and folded it into a new track. It was transformative,... See more

In other words, every new technology promises better clarity, yet its essence is determined by the noise it produces. The friction of limits is what gives a technology its character, so when a system becomes too smooth, too all-encompassing, or too accommodating, it stops having a signature at all. Here’s Eno again from earlier this year:

“I can see... See more

I wouldn’t want an irregular AI in my bank app, but in a creative workflow, hallucinating feels like the point of it all.

The internet challenged copyright by creating infinite distribution of perfect copies. With AI, what happens when infinite distribution is hooked up to infinite imitation?

Another answer is not to cultivate the machine’s output but to compose through it, treating the model as material, choosing the inputs and shaping the rules. In other words, not working beside the machine, but stepping inside it.

A good prompt doesn’t need to function like a blueprint. They can also behave like a horoscope or a fortune.