Thinking in Decades

Most of what gets published in a week is about that week. The best of what appeared this week is about decades — a roadmap to 2040, a theory of why growth took centuries to start, and a 2,800-year-old poem that turns out to be a set list.

The plan

The AI 2027 team — the people behind the internet's most-discussed doom forecast — published Plan A, presented by Scott Alexander: their first full account of how AI development could actually go well. The linchpin is a trustless US–China regime that tracks an estimated 98.5% of existing AI chips and relocates them to mutually audited "whitesites," then deliberately paces the 2030s — accelerate inside regulated data centers, pause at genius-level AI, and spend roughly a decade using contained systems to solve alignment. You can doubt the chip-tracking math (many will) and still find this valuable: the most prominent pessimists in AI finally saying what they're for is more falsifiable, and more useful, than another warning.

Why growth happened

Tyler Cowen's conversation with Joel Mokyr, recorded after Mokyr's economics Nobel, is the rare victory-lap interview where the laureate has to defend the thesis. Mokyr's story: medieval Europe built non-kinship institutions — guilds, universities, monasteries — while China organized around clans, and essentially none of the innovations Needham catalogued in China appear after 1400. The best moment is the eyeglasses puzzle: Rome had glass, literacy, and demand, yet no spectacles — because elites held manual work in contempt and lacked any concept of progress. Growth, in Mokyr's telling, is a cultural achievement before it is an economic one.

The oldest set list

Pegged to Christopher Nolan's forthcoming adaptation, Ted Gioia's What You Won't Learn About the Odyssey from a Movie argues the poem was performed music for a mass audience — Homer closer to Stan Lee than to a Harvard syllabus. Drawing on Parry and Lord's fieldwork, Gioia maps formula-singing onto jazz: "rosy-fingered Dawn" works exactly like a musician's stock lick, a memorized unit recombined live. His star exhibit is Avdo Međedović, the Balkan bard documented performing 12,000-line epics across seven days. The freshest angle on an ancient text you'll read this year, precisely because it comes from music history rather than classics.

Where did the rule-breakers go

Adam Mastroianni's The Decline of Deviance 2 extends his argument that rule-breaking of every kind is evaporating — high-school drinking down from ~50% in 1995, smoking, teen pregnancy, and crime all falling, marijuana use declining despite legalization in 24 states. His mechanism: prosperity raises the opportunity cost of deviance, accelerated by "cultural carcinization" — fast communication drives everyone toward the same optimal strategies, like a group that has played Secret Hitler too many times. It reframes a dozen disconnected trend lines as one phenomenon, which is exactly what a good essay is for.

The ultimate power plant

Dwarkesh Patel's episode with Adam Brown — Stanford physicist, now at Google DeepMind — is a multi-hour introduction to general relativity that actually earns the length. The detail that sticks: black holes permit extraction of up to 100% of rest-mass energy, against roughly one part in ten billion for chemical reactions and about 1% for fusion. Brown's closing provocation is that a sufficiently capable AI could rediscover general relativity from almost nothing but the finiteness of light speed and mathematical consistency. The best non-AI episode this podcast has run in months, even counting that kicker.

Also worth your time

Quanta reports that a four-person team resolved a 20-year-old question about quantum proofs — quantum information in a proof is fundamentally more powerful than classical. And Dynomight's Life With Hazard Ratios teaches you to convert any scary health headline into actual years of life: that fiber study's hazard ratio of 0.75 is worth about 3.7 years; eradicating cancer entirely buys only ~2.3.

The thread

A policy roadmap that runs to 2040, an explanation that runs back to Rome, a poem that survived twenty-eight centuries of live performance. The week's real argument: the interesting timescales are the ones nobody posts on.