AI
Buried the Lede: How Anthropic Drowned Its Own Safety Plan
Anthropic published a blueprint for how governments should switch off dangerous AI models the same morning it shipped one. Three days later the government switched off Anthropic's.
AI
Anthropic published a blueprint for how governments should switch off dangerous AI models the same morning it shipped one. Three days later the government switched off Anthropic's.
Shape of the Next Decade
Cheap simulation can make decisions better informed without making them easier to make. The hardest cases are the choices where information was never the missing ingredient.
AI
Anthropic launched its most powerful public model on Tuesday. By Friday the government had used export-control authority to switch it off worldwide, and the legal basis is still unclear.
AI
Anthropic's Claude Fable routes you to a weaker model, or quietly throttles you, based on who it thinks you are. Credit scoring is the hundred-year case study in where that kind of sorting machine ends up.
Shape of the Next Decade
AI companions arrived exactly when the institutions that turned loneliness into belonging were receding. The evidence on whether they help or harm is genuinely split.
Shape of the Next Decade
The AI discourse is focused on the 24% of work that requires reasoning. The economy is losing money on the 47% that is structured information flowing through humans instead of pipes.
AI Agents
When a system acts on your behalf, what record should exist?
Shape of the Next Decade
AI is changing which jobs survive, which tasks disappear, and which parts of work still feel like ours.
weekly
An FDA-cleared system that spots sepsis before clinicians do, and a Berkeley Lab model that gives AI actual eyes for atomic structure.
AI Agents
Agent coordination fails not from missing context but from summaries that strip confidence metadata, so downstream agents treat narrative as operational truth.
the-shape-of-the-next-decade
AI was supposed to automate tasks. The stranger shift begins when it starts learning the process that decides which tasks matter.
AI Agents
Five specialized agents can look like a startup. A company only appears when ownership, authority, handoffs, and receipts survive the work moving between roles.