My dear friend John Nosta has passed away. This is a hard one on me. He was such a dear friend. His was such a brilliant mind and such a polymath. John Nosta entered university work at sixteen. He enrolled in an advanced summer program at Harvard University. His father, an electrical engineer in Perth Amboy, New Jersey, had wanted a doctor for a son. The summer tested that path. I have known John long enough to recognize the pattern that began there. He measured first. Around those years he served as president of the Perth Amboy First Aid Squad and trained as a paramedic. The work put him in acute care, among Spanish-speaking neighbors, reading the immediate mechanics of heart and survival. At Boston University he designed his own concentration in biophysics. He left for a year as a research associate at Harvard Medical School. There he joined groups studying cell volume regulation in ventricular myocardium, acute myocardial infarction, ventricular arrhythmias, and sudden cardiac death. Mentors included Thomas Smith, then chief of cardiology at the Peter Bent Brigham Hospital. The papers appeared while he was still an undergraduate: one in the American Journal of Physiology in 1981 on hydrostatic forces limiting swelling of rat ventricular tissue, others in Circulation and the American Journal of Cardiology the same year and the next. Smith advised an MD/PhD. John considered it and declined. He has said he went to college to study physiology but found his interests expanding into creativity, art, and philosophy, and that the conversations at Harvard ranged beyond the medical details into geopolitics, love, and art. He finished the degree at Boston University in 1981 and left the laboratory. His parents were not pleased. He moved toward writing and the translation of science into other forms. That sequence set the measure I later saw in our conversations. John held the clinical questions of volume, ischemia, and failure as a fixed reference while he examined later technologies. He wrote more than a thousand pieces on technology, medicine, and the mind. When he published The Borrowed Mind, I recognized the same generous, fearless intellect that had appeared in our talks over the years: the insistence that human thought remains the author’s own even when the tools become fluent. He called listening to certain technical accounts exponential learning. I called him a national treasure genius. Both judgments rest on the same early evidence: the summer at sixteen, the year of peer-reviewed work before the degree, and the decision to keep the questions open rather than close them inside a single profession. The public record continues from that point. God rest your soul John. John Nosta died following an aortic aneurysm. Prayers to his amazing Amy, and children Katie, Annabel, Jack, and his greater family.

Conductor is shutting down and is the favored AI agent orchestrator for many teams. In looking for a replacement, we put together a shortlist of Conductor alternatives. Here is the list, compared across features, popularity, and agents supported.

I'm Sam, one of Anthropic's co-founders, and I lead much of our safety work. I'm going to have more to share on that work over the coming months. I thought I'd start by sharing a few thoughts on the latest report from @METR_Evals. The speed of progress in model capabilities is one of the most important questions in AI safety, and METR’s work here provides the public with an independent, external view into how this is happening inside Anthropic. I’ll talk more about some of the details below, but first some context for those new to this work. How quickly AI progress is accelerating has many important safety implications, since it determines how quickly the risks from advanced models increase. But, because most research happens within the frontier labs, it’s difficult for external observers and policy makers to get a strong sense of what that rate is. We kicked off the process with METR that led to this report earlier in the year, in anticipation of progress continuing to accelerate. For this report, we gave METR access to a range of sensitive internal data, including benchmark scores for unreleased models, estimates of how models are progressing, and interviews with our researchers. They had complete independence to form their own conclusions. Many of their conclusions match my own. AI is speeding up our research, but it’s hard to say exactly how much. While they conclude AI can't yet make the most important contributions to model training on its own, they think the pace of development could, by default, accelerate further. This work began well before and is separate from the embedded third-party evaluators that are part of our commitment to pacing the frontier, but it’s a good illustration of why we think evaluators are a good idea. The more access reviewers have, the clearer a picture they can form of the situation and the better-positioned they are to highlight situations like this one. We redacted a lot, and I accept that this makes it harder for others to check METR’s work. But we determined that sharing exactly what is driving the speedup isn’t the right thing to do, both because it’s commercially sensitive and because it risks handing out a roadmap for others to go faster. That would make the risks METR is worried about worse. That’s a tradeoff. I think we made the right call, though we’re continuing to assess the pros and cons here. In the coming weeks, we plan to share a report with our own analysis of the speed of AI progress within Anthropic, including where we stand against our Responsible Scaling Policy, plus new steps we’re taking to strengthen our safety and security practices as AI keeps improving quickly.

Every three years in medieval England, the king's men went from cottage to cottage cutting the toes off the peasants' dogs. It was called lawing. The dog's forefoot went on a block of wood, a chisel was set across three claws, and they came off with one blow of a mallet. The Charter of the Forest, 1217, records the correct method: three claws of the forefoot, but not the ball. A dog with three toes missing cannot run down a deer. That was the entire purpose. A third of England was royal forest, every animal in it belonged to the king, and a poor man's dog was the one piece of hunting equipment he could not be stopped from owning. So they left him the dog and took its feet. If the inspectors found yours intact, the fine was three shillings, about a month's wages. Before 1217 they simply took your ox. No law ever told that man he could not eat meat. The law took his bow, his dog's toes and his right to the wood, then left him entirely free to work out the rest for himself. He worked it out. He ate oats and beans, and he raised children who ate oats and beans. That is how it is always done. Meat is never banned for the poor. It just becomes, by one sensible regulation after another, impossible to get. Count the sensible regulations being drafted for your dinner this year.

In a moment of defeat, the brain undergoes a sudden shift. The neuroscientist Misha Ahrens (left) and his student Alex Chen showed that cells called astrocytes help to shift the brain state from hope to hopelessness.