For Every AI Tool, There is an Equal and Opposite AI Tool
The hope (the pitch? the grift?) was that AI would make us more productive. New tools would enter the world and, by automating a stack of previously laborious tasks, would liberate human energy for higher purposes. It's a fantasy that still circulates. It's in every VC blog post about GPT copilots. It's in the marketing copy of a hundred startups that promise to take tasks off your plate.
But despite an explosion of AI tools over the past two years, people do not seem especially more productive. They are busier, certainly. But the busyness is frenetic, almost hunted. Digital static.
My theory: it comes down to a simple law.
For every AI tool, there is an equal and opposite tool that cancels out your productivity gains.
There is a kind of Lagrangian symmetry to AI tooling. Every time an AI app accelerates a task, it creates a shadow tool somewhere else that reinflates the work. AI makes generating slide decks faster; so your boss asks for more decks. AI makes writing cold emails easier; your inbox fills with cold emails. AI can now write usable code; your codebase bloats with features no one needed. The net effect resembles a treadmill rather than a lever.
AI summarizers are the obvious example. These tools promise to condense meetings, papers, and long articles into crisp digestible takeaways. But within weeks of adopting one, most users discover that they're now receiving three times as many documents, all of them "summarized," each demanding review. The speed of intake increases, but so does the volume of what must be intaken. AI helps you skim more papers per hour, but also ensures you receive far more papers to skim.
The economist William Stanley Jevons noticed in 1865 that technological improvements in coal efficiency led to more, not less, coal consumption. Known now as Jevons Paradox, the dynamic is straightforward: make a resource easier to use, and people will find more ways to use it. AI is triggering the same pattern. Its gift is abundance, and abundance invites overuse.
Restraint, policy, human discretion are the levers to rein in the chaos. But the very appeal of AI is that it disintermediates. Who gets to decide what work is "useful"? Who will tell the eager PM not to ask for that extra feature now that it's "just one prompt away"? As in the Sorcerer's Apprentice, the brooms multiply faster than anyone can intervene.
AI is meant to reduce bureaucracy, but it gives us something similar in spirit and wholelly worse in execution: an information system that makes more information, which must be processed by more systems, which then generate new tasks to manage the output of the old ones. There is no ground floor. You can go faster, but not deeper.
Is it possible that the great productivity revolution will look, from a distance, like spinning in place? That the promised acceleration was real, but that it pointed us in a circle?