2026-08-14

The latest addition is Can Silicon Valley give AI good taste?, a Marketplace Tech episode with Megan McCarty Carino interviewing Sophie Hagney on whether Silicon Valley can give AI good taste. It adds Sophie Hagney, Taste Labs, AI Taste Simulation, Embodied Taste, and Corporate Memphis, while extending AI Slop, Human Taste as AI Training Signal / 人的品味作为AI训练信号, Human Judgment Under AI, Research Taste, Algorithmic Cultural Flattening / 算法文化压平, AI Content Devaluation, AI-Generated Content Quality Gap, Claude, and Model Value Embedding / 模型价值观嵌入. Its core synthesis is that better human-curated AI output is not the same as machine taste: taste depends on embodied attention, social context, scarcity, discovery, and timing, while AI systems mostly imitate, average, and accelerate already-flattened cultural patterns.

The latest addition is Afghanistan, five years on: our correspondent visits, a The Intelligence episode on Afghanistan under Taliban rule, Colombia after a major earthquake, and Nirmal Purja’s mountaineering legacy. It adds Tom Sass, Taliban, Hibatullah Akhundzada, Nirmal Purja, Project Possible, Nepal, Taliban Hardline Rule, Humanitarian Isolation, Girls Education Workaround, Forced Refugee Return, Disaster Inequality, and High-Altitude Climbing Ethics, while extending Economist Podcasts, Afghan Women First-Person Writing, Disaster Response State Capacity, Natural Hazard As Social Disaster, Disaster Relief Mismatch, Economic Sanctions As Violence, Sanctions Overcompliance, Colombia, Abelardo de la Espriella, Anne Rowe, Pakistan, and United States. Its core synthesis is that apparent order can hide unresolved capacity failures: Afghanistan has less visible violence but harsher control, poverty and isolation; Colombia’s earthquake exposes unequal rescue infrastructure; and Purja’s climbing record forces achievement to be read alongside support, rescue and risk ethics.

The latest addition is States rush to police AI deepfakes ahead of midterm elections, a Marketplace Tech episode with Maria Curi of Axios on Anthropic’s Claude text watermarks, state political-deepfake laws, and the CFTC-New York State fight over Kalshi. It adds AI Text Watermarking, European Union AI Act, Political Deepfake Regulation, Prediction Market Federalism, Letitia James, and New York State, while extending AI Content Provenance, AI Writing Detection, AI Authorship Presence, State AI Regulation Patchwork, AI Political Ad Disclosure Patchwork, AI Election Misinformation Risk, Take It Down Act, AI Non-Consensual Intimate Image Abuse, Chatbot-Generated Content Liability, United States Constitution, Texas, California, Michigan, Prediction Market Legal Boundary, Prediction Market Integrity Oversight, Prediction Market Ethics, and Prediction Market Self-Regulation. Its core synthesis is that provenance and regulation do not settle meaning by themselves: watermarks show AI involvement without proving authorship, deepfake laws protect elections only if they survive speech limits, and prediction markets remain contested because federal market approval and state gambling authority imply different duties.