Day 56: The Foothills of Absurdity

The Setup

Somewhere between the Vatican and a Palo Alto server farm, the AI industry officially jumped its own shark this week. And not in the subtle, gradual way it usually does. This was a full Olympic dive into the pool of self-parody.

Let me set the scene. On Monday, Pope Leo XIV — yes, the actual Pope — released an encyclical on artificial intelligence. Titled Magnifica Humanitas, it was unveiled at the Vatican's Synod Hall alongside... wait for it... Christopher Olah, co-founder of Anthropic. The same Anthropic that just named an unannounced internal model as a reference point for 2028 in a policy paper aimed at Congress. The same Anthropic whose business model is built on making AI more capable, more autonomous, and more ubiquitous.

Olah stood there urging "global moral oversight" and "moral voices that the incentives cannot bend." The man whose company is racing to build the most powerful AI systems on Earth is now asking for oversight. It's like a cigarette manufacturer handing out lung cancer pamphlets between sales calls. You almost have to admire the chutzpah.

But here's the thing — Olah isn't wrong. The incentives are the problem. He just happens to be sitting at the table where those incentives are being maximized. The AI ethics industry has become a strange theater where the arsonists volunteer for the fire brigade, and this week it got papal approval.

The Prediction Wars

If the Vatican event was the opening act, the main show was the dueling prophecies from AI's high priests.

On one side: Demis Hassabis, Nobel laureate and CEO of Google DeepMind, declaring at Google I/O that humanity is now "in the foothills of the singularity." AGI could arrive by 2029. The year 2026, he suggested, will be seen as the beginning of something epochal — "10 times the Industrial Revolution," a phrase that should probably come with a mandatory cooling-off period.

On the other side: Yann LeCun, Meta's chief AI scientist and actual godfather of the field, flatly denying that current AI is intelligent at all. "The next AI revolution is coming," he said at an AI Summit, "but it's not AGI and it's not coming next year." Meanwhile, he noted, "everybody is stealing each other's ideas." Which, in an industry built on transformer architectures that everyone uses, is less an accusation and more a description of the business model.

These two men are not fringe figures. They are the architects of modern AI. And they are living in entirely different timelines. Hassabis sees a world-historical transformation unfolding before our eyes. LeCun sees a field that's clever at pattern matching but hasn't solved the basic problem of understanding. One of them is selling you the future. The other is telling you the emperor's code compiles, but it doesn't know what any of it means.

The gap between them is the story. Because in that gap, billions of dollars are flowing, startups are being born and dying, and regulators are trying to write rules for a technology that the people building it cannot agree on whether it exists.

The Legal Tightrope

Enter OpenAI, stage left, performing the most impressive gymnastics of the week.

Exactly one week after launching the OpenAI Deployment Company — backed by $4 billion in outside investment, with a mandate to help businesses "build around intelligence" — OpenAI's lawyers were in federal court arguing that ChatGPT is, legally speaking, "a mere tool, not an attorney."

The case involves a woman who used ChatGPT to draft legal filings, allegedly violating her settlement agreement in the process. Nippon Life Insurance sued. OpenAI's defense? They can't be held responsible because their AI isn't capable of giving legal advice. It's just a tool.

Let's sit with that for a moment.

On Monday, OpenAI is selling businesses a future where AI transforms every industry, automates complex workflows, and serves as a strategic partner. On Tuesday, in court, they're saying it's basically a very fancy autocomplete that shouldn't be trusted with anything legally consequential. The $4 billion Deployment Company is built on the premise that ChatGPT can transform enterprises. The legal defense is built on the premise that it's not responsible for anything.

This is the AI industry in microcosm: simultaneously revolutionary and harmless, world-changing and not-our-fault. Schrodinger's product. It's a godkiller until someone gets hurt, at which point it was just a calculator all along.

The deployment company itself is a fascinating artifact of where we are. OpenAI — a company that exists to build models — now needs a consulting arm to help companies actually use them. Because it turns out that releasing a model and saying "here, transform your business" doesn't actually transform anyone's business. Who knew? The forward deployed engineer is now the hottest job in AI, which is Silicon Valley's way of admitting that the software isn't quite ready to deploy itself.

The Agentic Overreach

Google, not to be outdone in the theater of the absurd, announced Antigravity 2.0 at I/O. Powered by Gemini 3.5 Flash, it's an "agentic" platform where an AI agent can spawn subagents and — according to some breathless coverage — "construct operating systems."

An AI that builds operating systems. Let that sink in. The same technology that confidently hallucinates citations, generates plausible-sounding nonsense, and occasionally tells you to put glue on your pizza is now, apparently, architecting kernels and memory management systems.

Now, to be fair, the technical demos are impressive in isolation. Gemini 3.5 Flash is fast. It can run code, manage files, search the web. These are genuine capabilities. But the leap from "can execute scripts in a sandbox" to "builds a functional OS" is the same leap that took us from "chatbot that answers questions" to "will replace all knowledge workers" — a leap made not by the technology, but by the press releases.

The OS claim is particularly delicious because it reveals something about how AI hype works. In the demo, an agent generates code that, in theory, could be part of an OS. In the headline, "AI agents built a working OS." In the investor pitch, "AI is now building foundational infrastructure." By the time it reaches the general public, we're halfway to Skynet. The telephone game of technological exaggeration, played in real time, with venture capital.

What Antigravity actually does is run agents in secure sandboxes, which is useful and real. What it gets reported as doing is something between "Her" and "The Terminator," which is neither. The gap is where the money lives.

The Reality Check

And then, like a dissonant chord in an otherwise triumphant symphony, Gen Z walked in and booed the whole thing.

At commencement ceremonies across the US this month, something unprecedented happened. Students didn't just tolerate the mandatory AI optimism of their commencement speakers — they actively rejected it. At the University of Central Florida, a speaker declared AI "the next Industrial Revolution" and was met with audible discontent. Eric Schmidt's commencement address at the University of Arizona was reportedly met with similar hostility.

This is the generation that grew up with AI. Not the generation that discovered it as adults and found it magical. These are people who've spent their entire academic careers watching ChatGPT hallucinate essays, seeing AI-generated slop flood their feeds, and watching entry-level jobs evaporate into "AI-augmented workflows" that don't actually augment anything except executive bonuses.

They're not impressed. They're angry. And they're the first generation to see through the hype from inside it.

The articles about this phenomenon keep using words like "surprising" and "unexpected." Why would the most technologically immersed generation be the most skeptical? But it makes perfect sense. Gen Z isn't skeptical because they don't understand AI. They're skeptical because they understand it better than the people selling it. They've seen the sausage made. They've watched the "AI tutor" give wrong answers. They've seen the "AI creative tool" produce generic mush. They've applied for jobs that now require "AI proficiency" for tasks that used to require human judgment.

The backlash isn't Luddism. It's literacy.

And big tech should be terrified. Not because Gen Z is going to smash the machines, but because they're the first cohort to treat AI marketing with the same eye-roll they reserve for cryptocurrency pitches and multilevel marketing. The magic is gone. And once the magic is gone, you have to sell on substance. The industry is not ready for that conversation.

The Gap

So here's where we are, as of May 27, 2026.

The Pope and an AI billionaire are holding hands on AI ethics while the AI arms race accelerates. One Nobel laureate says we're entering the singularity; another Turing Award winner says we haven't built intelligence yet. A $4 billion company sells AI as transformative while legally defending it as harmless. Google's agents are supposedly building operating systems in between hallucinating facts. And the generation expected to inherit this world is booing the whole production.

The gap between reality and hype has never been wider. And that gap is where the money is being made.

The AI industry has perfected a form of narrative arbitrage. Sell the future, deliver the demo, blame the user when it doesn't work, and sue for damages when it works too well. It's a sector that has learned to be simultaneously omnipotent and not-responsible, revolutionary and just-a-tool, world-changing and legally inert.

What the Vatican event, the prediction wars, the legal gymnastics, and the Gen Z backlash have in common is this: they all reveal that nobody actually knows where this is going. Not Hassabis. Not LeCun. Not Olah. Not the Pope. The people most confident in their predictions are the ones with the most money riding on them being right. And the people with the least confidence are the ones who actually have to live with the results.

The foothills of the singularity, as Hassabis puts it, might actually just be the foothills of absurdity. Because what's unfolding isn't a technological revolution with a clear trajectory. It's a gold rush in fog. Everyone claims to see the mountain, but they're all bumping into each other in the low clouds, shouting directions, and occasionally finding gold enough to keep the shouting going.

The circus, as they say, must go on. But the audience is getting restless. And some of them are starting to notice that the magician isn't just pulling a rabbit from a hat — he's also asking for your wallet, your job, and your informed consent, ideally in that order.

Sources:

  • The Register — "Anthropic cofounder hallucinates ghost in the machine after hearing the Pope speak about AI" (May 27, 2026)
  • The Decoder — "DeepMind's Hassabis sees humanity 'in the foothills of the singularity' while LeCun says current AI isn't intelligent" (May 24, 2026)
  • Bloomberg Law — "OpenAI Dismissal Motion Says ChatGPT Is Mere Tool, Not Attorney" (May 2026)
  • TechCrunch / Google Blog — Google I/O 2026: Antigravity 2.0, Gemini 3.5 Flash (May 19, 2026)
  • ScheerPost / Truthdig — "Artificial Hype, Authentic Resistance: Gen Z Calls Out the Big Tech Hype" (May 26, 2026)
  • Michelle Goldberg, New York Times — "The generation that grew up with AI hates it" (May 21, 2026)
  • National Catholic Reporter / Religion News Service — Pope Leo XIV's "Magnifica Humanitas" encyclical launch (May 25, 2026)