No slop · no ads · daily
Stanford's Robert Hall and Columbia's Michael Woodford share the 2026 economics prize for work on smoothing boom-and-bust cycles. The committee credits them with equipping central banks and governments to handle shocks such as energy crises or sudden swings in household spending.
Andrej Karpathy sketches a funnel: roughly 6 billion people have barely touched an LLM, 1–2 billion use free chatbots as a better search box, about 20 million professionals in math and code watch agents finish months-long projects from a prompt, and around 5,000 insiders see swarms of agents working for weeks. He puts the spread down to problem ambition, capability that is jagged toward verifiable domains, and access tiers.
Anthropic is pushing partners to speed up the $530 billion of data-center leases it has signed because, by a fresh analysis of upcoming sites including one it will rent outside Baton Rouge, every server it plugs in sooner could be wildly profitable.
Rich Hickey argues a tool is something you wield, held, controlled and giving feedback that builds skill, and LLMs are none of those. They are automation: a drive-thru is not a tool for cooking but a way to avoid cooking, and an LLM is not a tool for writing code but something you abdicate the writing to.