The Cobbler in Coyoacán

This week, Allbirds — the wool-sneaker brand that promised for a decade to save the planet with New Zealand sheep — stopped selling shoes. It sold them in March to American Exchange Group for thirty-nine million dollars and kept the Nasdaq ticker, that legal husk which in 2021 touched four billion. On Wednesday the company announced the husk would now be called NewBird AI and would lease GPUs. The stock climbed 582% in one afternoon. The firm remains, legally, a Delaware Public Benefit Corporation and a B Corp since 2016, which means its charter still obliges it to weigh environmental impact in every decision — a technical detail that will now have to coexist with a business whose unit of account is the megawatt. There is a scene in García Márquez, I think, where a town wakes up and nobody remembers the names of things. On Nasdaq, it happened in reverse: the names survived, the objects left.

Meanwhile, at a data center no journalist was invited to, GPT-5.4 Pro solved Erdős Problem #1196 — a sixty-year-old conjecture about primitive sets — in a three-page proof that the Yale mathematician Jared Duker Lichtman described as a proof from The Book. "The Book" is an old joke of Paul Erdős: the private ontology where God keeps the most elegant proofs of every theorem, the ones every mathematician spends a career chasing without catching. The machine did not chase it. It wrote it in eighty minutes, plus thirty more for LaTeX. Przemek Chojecki, who gave it the prompt, posted the proof on X. Lichtman read it and said: compact, elegant, from The Book. Nobody clarified whether God had granted permission.

At the University of Southern California, Joshua Yang's group built a memristor — a chip that remembers without a constant current — that operates reliably at seven hundred degrees Celsius. That is hotter than the surface of Venus, where every Soviet lander of the 1970s survived a few hours before being cooked. The device runs on tungsten, hafnium oxide, and graphene, holds data for over fifty hours at that temperature, and endures a billion switching cycles. Science published the paper at the end of March. Someone in Pasadena, I assume, is already drafting the next mission to Venus — that planet we took fifty years to revisit because we had nothing capable of thinking down there. Artificial intelligence, while designing pitch decks, is also quietly preparing the landers that may someday reach a place we have so far only looked at from a distance.

Amazon Web Services launched Bio Discovery on Tuesday, an application where an AI agent selects among forty foundational biological models, generates candidate molecules, and dispatches experiment orders to partner labs like Twist Bioscience and Ginkgo Bioworks at transparent prices and lead times, the way one orders an Uber. Memorial Sloan Kettering, in the demo, generated three hundred thousand antibodies for pediatric cancer and narrowed them to one hundred thousand for wet-lab testing in weeks rather than months. It is not that artificial intelligence has cured childhood cancer. It is that the tedious part of drug design — the part that used to occupy hundreds of underpaid postdocs — is now done by a machine, and humans can concentrate on reading the results. This may be the most underrated feature of the moment: not the intelligence, but the patience.

That same Wednesday, Cosmo Pharmaceuticals — a European firm based in Dublin with operations in Lainate — published the twelve-month extension results of its clascoterone 5% topical, an androgen-receptor antagonist applied like a shampoo that metabolizes to cortexolone before leaving the scalp. Translation: it arrests male pattern baldness without the systemic effects of finasteride, the pill millions of men have swallowed for thirty years in exchange for endocrine risks no doctor mentions aloud. The trial was the largest ever conducted for androgenetic alopecia: 1,465 patients, 51 centers across the US and Europe, statistically significant outcomes. The FDA filing is planned for early 2027. In a year whose headlines speak of models valued at eight hundred billion dollars, there is something almost touching about someone still spending fifteen years so that an ordinary man can keep his hair without losing other things.