website
by Dan Luu
Dan Luu's long, data-heavy essays on how software actually gets built and run: normalization of deviance, why files are hard, computer latency from 1977 to 2017, hiring as a market for lemons, and lately how well coding agents really test their own work. Contrarian in the useful way — he measures the thing everyone else just argues about.
July 20, 2026
Dynomight on the coming end of pseudonymity, Brian Potter scores 239 laws against their promises, and a Sears-catalog rewrite of American growth.
April 6, 2026
AI Snake Oil on scaling myths, Import AI on 100K training runs, and Tyler Cowen finds AI in the productivity statistics for real this time.
March 9, 2026
Lex Fridman's DeepSeek deep dive, Import AI on DeepMind's AGI timeline, and a Noahpinion post on what AI companies are actually building now.
