Open-weight AI labs face a 'Kubernetes moment' as US developers race to match Chinese models on price and openness; Anthropic tells Claude 5 developers to delete 80% of their system prompts; and a mathematician mourns what AI-generated proofs take from the human experience of discovery

Marginal Revolution

What People Actually Want When It Comes to Human Enhancement — and How They Want It Governed

A new NBER working paper surveying 5,556 US respondents on human enhancement technologies finds a sharp split between what people would choose for themselves and what they think should be regulated. About 53% say they'd adopt some enhancement; roughly 28% would refuse at any benefit level, a refusal driven far more by an 'enhancing' framing than by actual risk — describing the same technology as restorative rather than enhancing lifts adoption by about as much as removing a severe side-effect profile would. The more striking finding is the mismatch between personal choice and preferred policy: most people who'd adopt an enhancement still want it strictly regulated, and most people who'd never adopt one don't want to stop others from doing so. Productivity-boosting enhancements draw the most ethical unease of any category studied, yet the least appetite for regulating them — respondents lean toward subsidizing access rather than taxing or restricting it, suggesting the underlying worry is inequality of access rather than the technology's safety. The authors' takeaway is that private demand is a poor guide to the governance citizens actually want, which cuts against any policy approach that infers 'what people want' purely from adoption rates.

Marginal Revolution

Why Democracies Are Unpopular Even When Their Economies Aren't Failing

Tyler Cowen highlights a Fareed Zakaria op-ed tackling a puzzle that pure economic explanations keep failing to resolve: democratic governments are broadly unpopular right now even in countries doing reasonably well by conventional metrics — France has cut unemployment and led Europe in foreign investment for seven straight years, South Korea and Japan both show real economic strength — yet voter discontent is nearly universal. Zakaria's answer looks past economics to technological disruption itself: societies are living through one of history's great technological revolutions, and such transformations reliably produce anxiety and upheaval independent of growth numbers. He draws the parallel to 1880-1920, when electricity, the telephone, the automobile, and mass urbanization uprooted entire ways of life and channeled the resulting fear into new, often extreme political movements on both the left and right. The implication for the current moment is unsettling precisely because it removes the easy fix — better GDP numbers or lower unemployment won't dissolve anxiety whose root cause is the pace of technological change itself, and how societies metabolize that anxiety politically is, per Zakaria, a matter of choice rather than economic destiny.

Construction Physics

Clean Energy Projects Face 10x More Regulatory Review Than Fossil Fuel Ones — and Other Infrastructure Notes

Brian Potter's weekly infrastructure roundup surfaces a striking regulatory asymmetry: an analysis of nearly 100 recent US energy projects found that clean energy and transmission proposals face review areas roughly 10 times larger than fossil energy projects under the National Historic Preservation Act, and are more than 8 times as likely to receive an 'adverse effect' finding that triggers further delay and mitigation obligations — largely because wind and solar's physical footprint and height create more potential sightline conflicts with historic sites than compact oil and gas infrastructure does. Elsewhere in the roundup, Fervo Energy's CEO describes the geothermal startup deliberately borrowing SpaceX's iterative philosophy — rather than attempting an ambitious deep or super-hot well on the first try, Fervo located its pilot project adjacent to an existing producing geothermal site to get a higher temperature gradient with lower technical risk, building drilling capability incrementally instead of solving every hard problem simultaneously. And as China's EV industry matures, Beijing is beginning to roll back the subsidies that built the sector, doubling a lithium-ion battery consumption tax from 2% to 4% over the next two years — a rare case of an industrial policy winding itself down once it judges the domestic industry self-sufficient.