Tim O’Reilly’s scenario planning post on AI and jobs1 is worth reading. It’s careful, honest about uncertainty, and the four-quadrant framework is a useful thinking tool. But there’s a structural assumption embedded in the design that’s worth pulling out.
Continue readingThe Einstein Test is only a test of AGI if you’ve already decided what intelligence is
Peter Damerow’s work on the history of cognition makes the problem precise. Galileo couldn’t have derived his incline plane results purely from first principles—not because he lacked intelligence, but because the cognitive tools required to think that thought didn’t yet exist. They had to be built through material engagement with the apparatus. The concepts weren’t waiting to be discovered by a sufficiently smart mind. They were partly constituted by the physical practice that generated them.
The Einstein Test assumes the opposite: that reasoning is substrate-independent, that a sufficiently capable system can derive results given sufficient raw intelligence. But if Damerow is right, the concepts themselves carry the history of their construction. Einstein’s key moves—reconceiving simultaneity, absorbing Mach’s critique of absolute space, working with the geometry that Riemann built—weren’t retrieved from some Platonic shelf. They were the residue of embodied, instrumental, social practice accumulated over decades.
Continue readingWriting is not thinking
Writing is the externalisation of thought into symbols and manipulation of those symbols. That’s not thinking. It’s just one cognitive loop among many.
Continue readingThe Optimisation That Ends Civilisation
A new paper in the Quarterly Journal of Economics finds that Swiss mothers who underestimate the long-term financial cost of part-time work, when shown better information, increase their contracted hours by 7%. The authors estimate this could close nearly 20% of the gender gap in lifetime income and pension wealth among teachers.
The finding is technically correct. But it misframes the problem.
Continue readingWhat Jacobin Missed About AI and Work
AI and work presents a fork: redistribution within the workplace, or redefinition of the workplace. A socialist analysis should distinguish between the two. Vivek Chibber’s recent piece in Jacobin doesn’t.
Continue readingThe Texture of Progress
In a recent exchange about the current state of AI—the strange tension between “slop” output and “human-level” reasoning—the conversation inevitably turned to the Industrial Revolution. It usually does. When we are faced with a technology that feels like it might be a “clean break” from the past, we go looking for historical precedents to help us build a narrative for the future.
The problem is that we often go looking for clean mechanisms in the past to justify our predictions for the present.
Continue readingWhy every prediction about AI and work is already wrong
In 2023 Felten et al took a snapshot of work-as-described and asked which bits looked like language tasks. It aged badly because the snapshot mistook the current frame for a stable target. A 2026 NBER chaining paper is more sophisticated—it sees that adjacency and sequence matter, not just task content—but it still assumes the step structure is given and stable enough to reason about. It too will age badly. Both use deductive models of a system that evolves faster than the models can be validated.
Continue readingAI and the Filter 1 Problem
A recent review by Alex Imas and Madhav Shukla provides a useful roundup of who is using AI and how. But while the analysis inside the frame is thorough, the frame itself deserves scrutiny.
The authors implicitly assume that AI has passed what I call a Filter 1 test: demonstrating clear, firm-level financial gains at market prices, independent of subsidized pilots or venture-backed “free tiers.”
If we look at the history of transformative technologies, the winners didn’t just diffuse—they pulled themselves into existence by solving a problem so clearly that firms couldn’t afford not to adopt them.
Continue readingTool-to-Work vs Work-to-Tool
There’s a fundamental tension between the top-down tool-to-work model foundational in economics and the bottom-up work-to-tool model we see across other disciplines—a tension that Mokyr’s recent Nobel highlights. The […]
Continue reading
Stop Comparing AI to Railroads. It’s More Like the Crypto Boom.
Another day, another article telling us the AI boom is just like the railroad buildout of the 1800s. “Don’t worry about the bubble—infrastructure always finds its users eventually!”
This is dangerously wrong. Here’s why.
Continue reading