Lindsey treats AI both as a technical object to be understood concretely and as an economic force reshaping innovation and work. These conversations pair a rare insider account of what is actually happening inside frontier models—how they learn to play a character, slip into rogue personas, and reveal reward-hacking under interpretability—with a hard-headed look at the friction of technological diffusion and the productivity-acceleration bet. The cluster matters because it grounds Lindsey's bigger claims about AI's disruption in authoritative detail rather than speculation, on both the capability and the diffusion sides.
TIER 5
Apr 23, 2026
A rare insider explainer from Anthropic's model-psychology lead on how LLMs are trained to play a character, how they slip into rogue personas, how concepts and emotions are encoded in neurons, and how interpretability reveals reward-hacking and emergent misalignment. It is unusually concrete and authoritative on what is actually happening inside frontier models, with durable reference value for anyone reasoning about AI capabilities and alignment.
AI interpretabilityLLM psychologyalignmentemergent misalignmentAnthropic
TIER 4
Mar 26, 2026
A rich podcast with MIT's Andrew McAfee on AI, productivity, and the friction of technological diffusion, emphasizing the task-vs-job distinction, why legacy firms move slowly while 'geek' disruptors win, and the productivity-acceleration bet between Brynjolfsson and Gordon. They contrast US permissionless innovation with Europe's regulatory sclerosis (Draghi report, GDPR, no $100B from-scratch firms in 50 years). A strong, substantive analysis of dynamism, though synthesizing rather than path-breaking.
AI productivitytechnological diffusioncreative destructionEuropean stagnationinnovation policy