LessWrong
Sanity-checking “Incompressible Knowledge Probes”Inkhaven, the 548th metapostAutomating Interpretability with AgentsAgainst In-Duct UVWinners of the Manifund Essay PrizeReflections on InkHavenAI risk was not invented by AI CEOs to hype their companiesHow much should the ideal person cry wolf?Cambridge: the kettleAI unemployment and AI extinction are often the same