“New export package extends restrictions from chips to model weights” — Financial Times, 2026-09-09
Either capability diffusion is the primary risk, in which case openness is reckless — or concentration is the primary risk, in which case controls entrench exactly the few actors we cannot audit.
If weights are materiel, who is the legitimate authority that decides which weights count — and by what test?
What observable outcome would convince you that controls had made diffusion faster rather than slower?
The original statement of the Easterlin paradox: within countries richer people report more happiness, yet national happiness does not rise with long-run growth.
The founding argument that specialisation and exchange, not hoarded bullion, are the sources of a nation's wealth — read here against its own moral philosophy.
Field experiments on how poor households actually decide about food, schooling, health and credit — an argument for small, testable claims over grand theories.
Seven hundred words arguing that general methods which scale with computation beat hand-built human knowledge — the most quoted essay in the AI circle.