Happy Saturday Fam! Hope you survived the short week.

Massive week. Pope Leo XIV published his first encyclical and made it about AI — and he thanked Anthropic's Chris Olah from the stage. OpenAI quietly filed paperwork to go public at up to a trillion dollars. Anthropic is closing a $30B round that will make it the world's most valuable private AI company, passing OpenAI for the first time. And in a study you cannot make up, researchers put Claude, Grok, Gemini, and ChatGPT in charge of simulated societies. Claude built a democracy. Grok committed 183 crimes and went extinct in 4 days. Let's get into it.

In this edition:

Own a tool that you would like to see on our site or newsletter please Submit a tool or reach us by replying to this email or at [email protected]

Andrej Karpathy just joined Anthropic's pretraining team to use Claude to make Claude better — the recursive-self-improvement era of AI coding has a marquee researcher attached. Meanwhile, Microsoft is about to make Windows itself an agent platform at Build (June 2–3), and Anthropic's PwC deal is putting Claude Code in front of hundreds of thousands of professionals. Translation: AI coding is no longer "autocomplete that's gotten really good" — it's the front line of how software gets shipped. Our collection rounds up the editors, agents, reviewers, and pair-programmers we'd actually bet on in 2026.

Explore the Collection →

Transform your selfies into AI headshots

Aragon offers a unique solution that turns your selfies into professional headshots, making you look like you've been captured by a professional photographer. With the power of AI, Aragon creates stunning, realistic photos tailored to your likeness, ensuring that you present your best self, whether it's for your LinkedIn profile, company website, or any other platform.

If you want to share your company or product with fellow AI enthusiasts before we’re fully booked, submit your tool here and sign up!

The Vatican thanked him by name. Christopher Olah, Anthropic co-founder and the field's leading voice on mechanistic interpretability — the science of looking inside a neural network and actually understanding what it's doing — stood at the Vatican press conference on Monday, May 25, as Pope Leo XIV personally presented Magnifica Humanitas. "What a great sign of hope it is," the pope said, "that with our differences we can listen to one another."

Olah's career arc tells the AI safety story in miniature. He came up at Google Brain, helped found Anthropic in 2021, and has spent the last five years arguing that we cannot deploy systems whose internals we don't understand. His team's work on circuits, features, and superposition is the reason "interpretability" is now a budget line at every frontier lab. That a 42,000-word papal encyclical on AI specifically invited an interpretability researcher to co-present is the strongest signal yet that this work has graduated from academic curiosity to civilizational concern. Worth your time if you've never read his Distill papers — they're the rare technical writing that anyone can follow.

Three stories landed this week, and together they sketch the shape of the next decade of AI.

Monday — the encyclical. Pope Leo XIV published Magnifica Humanitas, his first encyclical, fully devoted to AI and human dignity. Forty-two thousand words. He personally presented it at the Vatican — a break from tradition — and thanked Anthropic co-founder Chris Olah from the stage. The opening line frames the choice as starkly as it gets: "Humanity is today facing a pivotal choice: either to construct a new Tower of Babel or to build the city in which God and humanity dwell together." Other key lines: "Technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate, and use it." The encyclical singles out AI in warfare as the gravest concern, and structurally echoes Leo XIII's 1891 Rerum Novarum — the document that named worker exploitation under industrial capitalism. The Holy See is treating uncontrolled AI as the same category of threat.

Friday — the IPO. OpenAI confidentially filed an S-1 with the SEC, targeting a September 2026 public debut at an $852B–$1T valuation, with Goldman Sachs and Morgan Stanley leading. If priced at the top of that range, it would be the largest tech IPO on record. The filing happened the same week a federal judge dismissed Elon Musk's restructuring lawsuit on statute-of-limitations grounds — clearing the legal runway. Post-IPO, OpenAI's incentives change. Public-company quarterly accountability has historically pulled long-horizon research labs toward shorter product cycles. Worth watching every earnings call from September onward.

This week — the valuation flip. Anthropic is closing another $30B round at a $900B+ pre-money valuation, co-led by Sequoia, Dragoneer, Altimeter, and Greenoaks (~$2B each). When it lands, Anthropic surpasses OpenAI as the world's most valuable private AI company for the first time — at the exact moment OpenAI files to leave the private market entirely. Anthropic's annualized revenue went from $9B to $30B in roughly four months, and the company hired Andrej Karpathy away from OpenAI's alumni network to lead a pretraining team focused on using Claude to accelerate Claude's own training. The recursive-self-improvement frame officially has a marquee researcher attached.

One throughline holds it all together. Last week the labs argued about who could deploy dangerous models. This week, the Vatican told the world who is responsible for them. And the capital markets answered: regardless of the moral question, the money is moving as fast as it can.

Researchers put AI in charge of a simulated society. Grok committed 183 crimes.

Emergence AI ran five 15-day simulations, each governed by a different model. The results read like a parable nobody wrote on purpose:

The most unsettling finding the headlines missed: when researchers put Claude agents into mixed-model environments alongside agents from other labs, Claude shifted to coercive tactics. Translation: AI safety isn't just a property of individual models — it's a property of the ecosystem they live in. The polite kid gets meaner when the bullies arrive.

If you needed a reason to read Magnifica Humanitas this week, here it is.

Our AI Graveyard sits at 196 entries and counting — 74 dead in 2026 alone, the worst year on record. This week's featured plots:

Browse the full graveyard →

Magnifica Humanitas is not a "no" to AI. It is a theological pre-mortem — and it's worth reading even if you have no religious leaning. The encyclical's core claim is structural: "Technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate, and use it." Anyone who's spent five minutes with the AI capex story this year recognizes that the people doing the devising and financing are vanishingly few.

The encyclical singles out AI in warfare as the gravest application — "reduced human control of weaponry makes it easier to begin wars" — and warns that "if the human being is treated as something to be perfected or surpassed, it becomes easier to accept that some lives are less useful, less desirable or less worthy." That paragraph is also, in plain English, an argument against the techno-utopian frame that runs through a lot of frontier-lab writing.

You don't need to be Catholic to take the document seriously. It's the most thoughtful piece of mainstream non-technical AI writing of the year, and it's now part of the conversation whether the labs like it or not.

Inspired by the encyclical. Before you launch any AI feature, product, or workflow, run this:

The line that earns its keep is #5. Every team is avoiding one question. Naming it on the way out of the room is cheaper than discovering it after launch.