Scott Alexander, curated
← Back to curation

H5N1: Much More Than You Wanted To Know

Quality
74
Strong
Claude Shift
38
Slight
RWI
3
of 10

Summary

A calibrated risk explainer on H5N1 bird flu. Covers flu basics (influenza A, the H/N antigens, the pandemic-every-few-decades pattern, and the antibody-imprinting that explains the 1918 flu's weird age-mortality curve); the modern pandemic history (1918/1957/1968/1977/2009); how bird flus cross over via reassortment in pigs ('viral Tinder'); and H5N1's march through species (chickens 1959, geese 1996, minks 2022, cows 2024, pigs Oct 2024) with the 'biocomputational surface' framing. On probability, he trusts Metaculus (5% for 2025) over Manifold (40%), quoting expert forecaster comments, and gives ~25-50% by 2030. On fatality, he debunks the '50% mortality' scare (it only counts hospitalized cases; 0/61 US cases died; Metaculus central ~1.25%) and synthesizes four scenarios into 30% normal-flu / 63% 2-10x (Hong Kong) / 6% Spanish-flu / <1% worse, then an expected-value framing (~166k deaths/year in expectation, 'about as many as the Ukraine War'). Closes on raw milk, vaccine stockpiles, the mink-farm-buyout idea, and ACX grantees.

Why this score

Quality 74 · Strong. Strong band, upper. Clear, fair, well-calibrated risk communication that debunks the scare statistics and disciplines a messy question with base rates and forecaster synthesis; the antibody-imprinting explanation of 1918 and the Metaculus-vs-Manifold epistemics are highlights. Upper-Strong rather than higher because it synthesizes others' forecasts and flu background rather than offering an original thesis, and is lighter than his standout MMTYWTKs.

Claude’s paradigm shift 38 · Slight. Slight-to-moderate. Explanatory synthesis of known flu biology and current forecaster estimates; no new idea of its own.

Real-world impact 3 · Moderate. A calibrated risk explainer on H5N1 bird flu that debunks the scare statistics and disciplines a messy public-health question with base rates and forecaster synthesis (the antibody-imprinting account of 1918; Metaculus-vs-Manifold epistemics). Real topical relevance, but synthesizes others' forecasts and informs understanding — modest RWI.