The Atlantic's profile of economist Erik Brynjolfsson, the man who saw AI coming, lands in a week when the concrete downstream effects of AI's economic impact are visible in three simultaneous and wildly different registers: Mars reformulating M&Ms for MAHA compliance, newly minted SF tech millionaires speccing robot closets, and Samsung committing half a trillion dollars to AI chips. Brynjolfsson spent decades arguing that technology's productivity gains are real but distributed with grotesque inequality. The evidence this week is extensive.
The Productivity Paradox in Consumer Form
Brynjolfsson's original insight was the productivity paradox: computers were everywhere but productivity gains were hard to measure. His updated argument for the AI era is that the gains are coming but the adjustment costs, displacement, deskilling, wage polarization, will be concentrated in the people least able to absorb them. Fast Company's piece on San Francisco home design captures the beneficiary class in its purest form: cold plunges, circadian lighting, robot closets. The once-in-a-generation wealth event referenced is the IPO and acquisition cycle driven largely by AI infrastructure spending. Mars changing its candy ingredients in response to MAHA-style regulatory pressure is the other end of the same economy: mass-market consumer goods reformulating under political and health pressure while the premium tier builds biohacking kitchens.
The AI Persuasion Layer
The academic dimension sharpens this. A new arXiv paper, "AI Persuasive Framing in Collective Dilemmas" by Moller, Galdeman, Pera, and Aiello, finds that AI agents can act as behavioral nudges to enhance human cooperation in collective action problems. Brynjolfsson would recognize the irony: the same systems generating the wealth inequality he predicted are now being positioned as tools for persuading humans to cooperate on the collective problems that inequality creates. The nihilism and enshittification framework Kyle Raymond Fitzpatrick articulated feels newly precise here. The systems are not broken. They are working exactly as designed, generating surplus for a narrow class and circulating AI-powered behavioral nudges to everyone else.