Pope Leo XIV - Chicago-born, White Sox-rooting (even though his Mom was a Cubs fan), Wordle addict, math whiz - is easily my favorite Pope (speaking as a non-Catholic) since “Good Pope John” XXIII went to his celestial reward in 1963.

On Monday, Pope Leo released the first encyclical of his papacy, “Magnifica Humanitatis” (“The Greatness of Humanity”), a 40,000-word meditation and call to action on the human implications of artificial intelligence. The timing was no accident. It coincided with the 135th anniversary of his namesake, Pope Leo XIII’s landmark 1891 encyclical, “Rerum Novarum” (“Of New Things”). Written at a time of vicious union-busting violence orchestrated by the robber barons of the Gilded Age, Rerum Novarum was a critique of the Industrial Revolution, and a call for a Third Way between unfettered and rapacious capitalism, and Marxist communism. Rerum Novarum formed the blueprint for the rise of Europe’s Social Democratic parties, which worked to create a capitalist society with robust worker rights and a strong social safety net. We see the result in all the most developed economies of the European mainland.

Pope Leo explicitly frames the rise of AI as the new Industrial Revolution, with enormous consequences for how humans live, think, and act.

This encyclical is steeped in Pope Leo’s experience of decades in Peru, where he saw poverty, colonialism, elite corruption, and a violent Marxist insurgency (the notorious “Shining Path” militants). Out of that mosaic of contradictions, Leo XIV, like Leo XIII before him, has forged a vision of a Third Way, one that embraces Liberation Theology (which centers the concerns of the oppressed) while rejecting the usual Marxism, which wants to destroy capitalism and even private property altogether.

To drive home the importance he placed on these issues, Pope Leo took the unusual step of personally reading the encyclical aloud. He also shared the stage with Christopher Olah, a co-founder of Anthropic, who said that computer scientists could not resolve the ethical dilemmas of AI by themselves, and that AI companies faced powerful economic pressures that made it difficult for them to reliably put the good of humanity first, above profits.

As a final surprise Pope Leo publicly apologized for the Catholic Church’s condoning of, and complicity with, the historic evil of the colonial-era transatlantic slave trade.

Mic drop on a very surprising Pope.

The Pope warns us to “disarm ‘AI

At last week’s Google I/O developers conference, the company at last shed its ambivalence and went all in on AI.

Google, the company that invented the Transformer architecture that underpins all the leading AI models, famously kept the technology on a shelf for years, allowing OpenAI to leap ahead and release ChatGPT in November of 2022. Google’s response to the challenge of OpenAI and Anthropic has been somewhere between muted and tepid, for one very good reason: AI could destroy Google’s core business.

Google’s core business is selling ads for search terms, generating over $200 billion in revenue annually, over half of the company’s total revenue. Since most people use AI as an “answer engine”, AI to date has not been serving up juicy, “sponsored” links that someone has paid actual money to have put in front of your eyes. Theoretically, AI could destroy that entire revenue stream. Hence the ambivalence on Google’s part.

At last week’s Google I/O developers conference, Google took the plunge. The company has gotten rid of the venerable Search box on its web page, and replaced it with a chatbot interface. Google is betting that it can do what few incumbents have ever managed - reinvent the core products fast enough that it survives the platform shift while funding the transition from the revenues of the old business.

At a first glance, there are encouraging signs that this strategy may work. Google’s new AI Search starts with a billion users per month, all converted automatically from Google’s Search traffic. This instant billion-a-month user base equals or exceeds ChatGPT’s user metrics, that OpenAI spent 3 ½ years and countless billions of dollars accumulating. The switch also brings in users accustomed to seeing and interacting with ads from their queries, making it much easier for Google to keep serving ads to their free tier users than OpenAI will have in initiating ads with their own users.

Google is also putting its Gemini AI into all of its other products, including Gmail, YouTube, and Google Workspace (Docs, Sheets, etc.) This further widens its distribution edge.

For modern tech companies, products with better distribution often beat products with better features. Google has decided to leverage its massive user base, built over decades, to try to outcompete OpenAI and Anthropic for dominance in the AI era.

In related news, Google also announces Gemini Spark, its autonomous agent platform designed to compete with viral sensation AI assistant OpenClaw.

Google goes all in on AI

Elon's latest scheme to save his tottering business empire, the SpaceX Initial Public Offering (IPO), continues to provide popcorn-worthy entertainment, as the truly bizarre details of his tangled business dealings come out.

As a private company, SpaceX has not had to reveal the details of its operations. Now that the company is preparing to sell shares to the public, SEC rules require a level of disclosure and transparency that are proving inconvenient to Elon, and highly entertaining to the rest of us. The company released its required S-1 information packet last Wednesday, and boy, is it a doozy.

First, the merely crazy part. The current cash cow of SpaceX is its Starlink satellite internet company, which generated around $11 billion in revenue last year, more than half of SpaceX’s total revenue. Unfortunately, xAI, the AI company that was Frankensteined into SpaceX to give it some sweet AI market multiples, spent $20 billion on capital expenditures, mostly on building data centers which have sat idle because nobody wants to use xAI’s toxic AI model, Grok. So Starlink prints money, xAI burns it, and shareholders will end up funding any shortfall.

Even crazier, Musk will end up owning enough special-voting shares that he will have voting control of the company, able to appoint and dismiss Board members at his whim. He will also serve, simultaneously, as CEO, Chief Technical Officer, and Chairman of the Board. All while serving as CEO of another trillion-dollar publicly traded company, Tesla. Busy guy.

Now for the Really Crazy details. Elon’s compensation package is structured into a series of 27 different levels - 15 levels for developing a sustainable colony of humans on Mars (I am not making this up), and 12 levels for getting massive numbers of data centers into Earth orbit. If he nails all 27 levels, the company will have to have a market value of $7.5 trillion (only one company, Nvidia, has ever cracked $5 trillion), there will be 1 million human colonists on Mars, and 100 terawatts of computers in orbit. And pigs will fly. (Ok, I made up the part about the pigs.)

We have detailed in previous stories how the SpaceX IPO is structured to make investment bankers and even the NASDAQ exchange itself complicit in an IPO that is likely to fleece retail investors who have more fervor - for space, AI, or Elon - than sober judgement. Shame on all of them.

SpaceX ‘s IPO document is crazy

Anthropic has withheld release of its latest AI model Mythos, because it was discovered to be so extraordinarily good at cyber-hacking that it was deemed too dangerous to release to an unsuspecting world.

Instead, Anthropic contacted scores of organizations with large code bases, and gave them a demonstration copy of Mythos to seek out the weak spots in their code. Now, six weeks later, they are tallying up the vulnerabilities found. It comes to a staggering 10,000 chinks in the digital armor of code bases that many millions of users depend on for vital services around the globe. And the tally keeps growing.

Meanwhile, organizations are using Mythos to help patch any vulnerabilities found. When Anthropic is satisfied that the danger of mass cyber-hacking using Mythos has receded due to improved code integrity, they will roll out some version of the Mythos model to a better-prepared user base.

Anthropic’s Project Glasswing has detected over 10,000 vulnerabilities

AI infrastructure startup Span has teamed up with leading AI chip maker Nvidia to provide distributed AI data center capacity to residential neighborhoods in the form of “compute panels” attached to the side of your house.

Demand for AI computing continues to skyrocket, yet building mammoth data centers is getting ever slower and more costly as local communities fight back against the resultant spike in electricity rates.

Span proposes to bypass all the pesky permitting and building phases, and pay homeowners to allow them to attach a compute panel to the side of their house. These compute panels would draw the electricity to power their AI computing directly from the local grid, distributed across neighborhoods, rather than concentrated into a single building.

This seems more a ploy to defuse local opposition and leapfrog over local and state permitting processes than an actual solution to the grid capacity crisis. Putting mini-data centers on the sides of houses will ultimately use as much electricity (or more, due to inefficiencies inherent in decentralization) as the monolithic centralized data centers, so the grid strain will be similar. Nonetheless, the frenzy over AI computing capacity will not abate soon, so expect to see more and more “creative” solutions proposed.

Startup Span and Nvidia bring AI data centers to the side of your house.

Data is accumulating that driverless taxis like Google’s Waymo have many fewer accidents per mile driven than vehicles driven by humans, and far lower fatal accident rates. This is causing an increasing number of physicians to begin advocating for programs to increase driverless vehicle use as a matter of public health.

Recently, two high-profile physicians, Jonathan Slotkin, a neurosurgeon and Chief Medical Officer at Geisinger Health System, and Eric Topol, an acclaimed cardiologist and researcher at Scripps Research in California, have circulated an open letter from doctors and nurses urging government leaders to reduce regulatory barriers against driverless vehicles as a matter of public health. They cite a 2025 peer-reviewed study of Waymo data, which covered 56.7 million driverless miles on public roads, and found an 85% reduction in serious-injury-or-worse accidents compared with human drivers on the same streets.

Doctors rally behind driverless vehicles as a public health issue

Recent research from the Sylvester Comprehensive Cancer Care Center at the University of Miami Miller School of Medicine has demonstrated the use of an AI model to stratify multiple myeloma patients, predicting which patients may respond well to new immunomodulation anti-cancer therapies.

The scientists used a new foundational AI model known as GigaTIME, developed by Microsoft and a number of health care partners. GigaTIME is able to analyze routine, inexpensive H&E stained two-dimensional pathology slides, and convert the image to the much more useful “virtual” 3-D multiplex immunofluorescence (mIF) format, which reveals the immune microenvironment surrounding tumor cells.

With the virtual mIF image, decisions about which patients are more likely to respond well to immune therapy are much more clear.

This study points the way toward a possible future of personalized precision cancer care.

From biopsy slide to customized cancer treatment