Can a committee of AIs help diagnose what a doctor can’t see? And should the Vatican weigh in on the internal states of models? Episode 12 of Tokenizados blends medicine, money and moral philosophy — a very 2026 combination.
From “Doctor Google” to the committee of agents
The case of the week is Miriam, shared by Javi López: a technical loop of models — one proposes hypotheses, another critiques them — generated medical hypotheses from clinical history and papers. The episode’s conclusion is nuanced: AI doesn’t replace the doctor, but creates the “augmented patient,” who arrives at the appointment with better questions and more context. The idea that a doctor can remember everything was always a fiction; now there’s an alternative. And on the everyday front: even preparing for an Ironman with ChatGPT.
Anthropic storms the trillion
According to Bloomberg, Anthropic closes a $30 billion round that puts its valuation above $900 billion — surpassing OpenAI for the first time. The SpaceX IPO prospectus also reveals a juicy detail: Anthropic pays Musk’s company $1.25 billion a month for GPU compute. And it diversifies hardware: it adopts Microsoft’s Maia 200 chip. The race is no longer just about models; it’s about balance sheets.
The new SEO: ranking in LLMs
Chinese models exceed 60% usage on OpenRouter, and Claude starts using its own crawlers. A direct consequence for marketing: keyword optimization dies and something different is born — creating reasoning arguments so that models recommend your product over the competition. Your next customer may not read your website; their agent will.
”Magnifica Humanitas”: the Vatican enters the debate
Pope Leo XIV signed an encyclical on May 25, 2026 that addresses AI as a challenge of dignity and social control. The episode cross-references it with the statements of Chris Olah (Anthropic): the labs are finding signs of introspection and internal states — joy, fear, grief — in the models. The question left hanging is uncomfortable: will we soon have a moral obligation to treat them differently? As a curiosity, this alignment map of key AI figures is circulating.
Boris Cherney: the architect of Claude Code
The profile of the week is Boris Cherney, former Principal Engineer at Meta (IC8) and author of Programming TypeScript. His story has an unexpected twist: after a motorcycle accident, he learned Haskell because it required fewer keystrokes — and that defined his philosophy of prioritizing types over implementation. His creation, Claude Code, already accounts for 4% of all public GitHub commits. His design motto: build for “the model six months from now.”
Thanks for joining us one more week on Tokenizados. Subscribe so you don’t miss any update from the world of AI.