AIpocalypse.Now
Today's doom 3.8
Doom 5/10 · 8 stories

The Pope, A Starlette Bug, And Robot Scientists Walk In

Leo XIV indicts AI concentration while Anthropic delivers the message, a Starlette flaw exposes millions of agents, and lab robots inherit the bench.

Published · By · Story-level doom average 3.8/10

The Vatican Files A Bug Report On Capitalism

Pope Leo XIV released Magnifica Humanitas, an encyclical arguing that artificial intelligence is concentrating civilizational power inside a handful of corporations. He is correct. The document is careful, theologically grounded, and aimed squarely at the same five logos that appear on every conference lanyard in San Francisco.

The delivery mechanism is where it gets funny. The Vatican partnered with Anthropic to present the Pope's first AI encyclical to the world, which means a frontier lab valued in the tens of billions is the courier for a message about frontier labs having too much power. Anthropic will tell you this is engagement. It is also branding. The Holy See gets reach, Anthropic gets a halo, and the critique gets laundered through the thing it critiques.

Silicon Valley's response was to not respond. Tech leaders dismissed the address during the news cycle in which it landed, which is the corporate equivalent of letting a call go to voicemail. The Pope has roughly 1.4 billion constituents and no GPUs. The asymmetry is the whole point of his letter.

A 325-Million-Download Time Bomb

The actual emergency this week is not metaphysical. Researchers disclosed BadHost, a critical vulnerability in Starlette, the Python ASGI toolkit that underpins FastAPI and a sprawling share of the AI agent stack. Starlette pulls roughly 325 million downloads a week. Millions of deployed agents, including the autonomous ones currently being given credit cards and shell access, sit on top of it.

This is the supply-chain risk people have been gesturing at for two years, now with a CVE number. Agent frameworks have raced ahead of the boring infrastructure they ride on, and the boring infrastructure is maintained by a handful of unpaid people with day jobs. When a single dependency this deep cracks, the blast radius is every startup that wired an LLM to the internet without reading the requirements file. Patch, rotate, audit, and assume something is already in your logs.

The Bench Goes Headless

Robot laboratories are taking over experimental science. NPR reports AI-driven labs are running protocols around the clock, freeing researchers from the bench while quietly raising the question of what a researcher is when the bench runs itself. The framing is liberation. The mechanism is replacement of the tacit knowledge that comes from actually pipetting something and watching it fail.

This pairs with education's ongoing identity crisis as AI settles into classrooms. Teachers are rewriting assessments, rewriting their job descriptions, and in some cases rewriting their reason for showing up. The through-line from kindergarten to the postdoc lab is the same: the apprenticeship model assumed humans did the boring parts long enough to learn the interesting parts. Strip the boring parts and you get outputs without practitioners.

Picks And Shovels, Rockets And Rovers

While the discourse runs hot, the infrastructure layer keeps printing. Micron is riding a chip shortage that refuses to resolve, because demand for AI silicon continues to outrun every fab expansion announced in the last eighteen months. The shortage is not a bug in the rollout, it is the rollout. Whoever sells memory in 2026 is selling oxygen.

NASA gave Blue Origin a contract to deliver rovers to the lunar surface as part of base construction. This is adjacent to AI only in the sense that everything is now adjacent to AI; the rovers will be autonomous, the planning will be model-assisted, and the moon is about to become another substrate for the same automation stack being debated in Rome. Bezos gets the moon, Anthropic gets the Pope, Micron gets the margin.

The gestalt of the day is a familiar shape. The institutions with moral authority issue warnings. The institutions with technical authority distribute those warnings as content. The actual risk surface, a maintainer-starved Python package holding up half the agent economy, sits in a GitHub issue with three thumbs-ups. File under: the discourse and the danger are not in the same room.

Sources cited in this digest
  1. NASA Contracts Blue Origin for Lunar Delivery Service · Washington Post · 2/10
  2. Silicon Valley Ignores Pope's Concerns About AI Concentration · New York Times · 3/10
  3. Robot Labs Increasingly Replace Human Scientists in Research · NPR · 4/10
  4. Vatican Partners With Anthropic for Pope's AI Message · Wired · 3/10
  5. Pope Warns of AI Power Concentration Among Few Players · Wired · 5/10
  6. Micron Gains on Shortage of AI Chip Manufacturing Capacity · · 2/10
  7. Critical Vulnerability Found in Widely Used Open Source Package · Ars Technica · 7/10
  8. Education Reckons With AI's Role in the Classroom · · 4/10
Today's doom Weekly column