Hacker News · May 25, 2026 · 6:52

Pope Leo XIV on AI & US bets on quantum foundry - Hacker News (May 25, 2026)

Pope Leo XIV says AI isn’t neutral, IBM backs a quantum foundry, GPT-4.1 “randomness” fails, plus White Rabbit timing and Jira Turing-complete automation.

English Español Français
Pope Leo XIV on AI & US bets on quantum foundry - Hacker News (May 25, 2026)
0:006:52

Our Sponsors

Invest Like the Pros with StockMVP Learn more → Lindy is your ultimate AI assistant that proactively manages your inbox Learn more → Prezi: Create AI presentations fast Learn more →

Today's Hacker News Topics

  1. Pope Leo XIV on AI

    — Pope Leo XIV’s “Magnifica Humanitas” reframes AI and the digital revolution as a new phase for Catholic social doctrine, emphasizing human dignity, accountability, and the common good—plus limits on AI weapons.

  2. US bets on quantum foundry

    — IBM and the U.S. Commerce Department outlined Anderon, a New York quantum chip foundry backed by CHIPS Act incentives, signaling industrial policy momentum around manufacturable superconducting qubits and scaled wafer production.

  3. LLMs aren’t truly random

    — A GitHub experiment shows gpt-4.1 produces biased results when asked for a “random” number, spotlighting why LLM outputs can’t be trusted for fairness-sensitive selection, games, or simulations without real randomness.

  4. White Rabbit precision timing networks

    — White Rabbit, born at CERN, demonstrates sub-nanosecond synchronization over Ethernet, a big deal for distributed science and industrial systems where timing accuracy directly impacts measurement quality and coordination.

  5. Jira automation becomes computation

    — A reproducible construction argues Jira Automation can implement a Turing-complete model, turning “workflow rules” into real programming—with all the power, complexity, and risk that implies.

Sources & Hacker News References

Full Episode Transcript: Pope Leo XIV on AI & US bets on quantum foundry

What does it mean when the Pope declares that AI isn’t morally neutral—and warns it could erode “truth as a common good,” reshape work, and even pressure how wars are fought? Welcome to The Automated Daily, hacker news edition. The podcast created by generative AI. I’m TrendTeller, and today is May-25th-2026. Let’s get into the stories people on Hacker News were digging into.

Pope Leo XIV on AI

Let’s start with a big one at the intersection of technology and public life. In a new encyclical titled “Magnifica Humanitas,” issued May 15th, Pope Leo XIV frames the digital revolution—especially AI—as a defining “new thing” that demands an updated social doctrine. The core message is that AI isn’t some neutral tool that automatically becomes good or bad depending on who uses it. Instead, it tends to amplify incentives and power structures, so governance and values have to be designed in from the start. He uses a striking contrast—Babel versus the rebuilding of Jerusalem—to argue technology can either strengthen human communion and the common good, or push society toward domination, sameness, and dehumanization. Concretely, he calls for accountability and transparency, and extends classic ideas like the “universal destination of goods” into the digital world—explicitly naming data, algorithms, and platforms. There’s also a strong warning about disinformation and the erosion of truth as shared civic infrastructure, along with a push for digital literacy and education that protects children. And he doesn’t keep it abstract: the text points to automation’s impact on work—job displacement, surveillance, de-skilling, and widening inequality—and urges policy and standards that keep employment dignified and distribute the benefits. On security, he criticizes a “culture of power” that normalizes war, pushes back on overly permissive just-war reasoning, and calls for strict limits on AI-enabled weapons, insisting lethal decisions must remain under meaningful human control. Whether or not you share the Church’s framing, it’s a reminder that AI governance isn’t only a technical debate—it’s becoming a moral and political one, in every major institution.

US bets on quantum foundry

From ethics to hardware: IBM and the U.S. Department of Commerce signed a letter of intent for a project called Anderon, described as a purpose-built quantum chip foundry, planned for Albany, New York. The headline is the scale and the signal. The proposal pairs roughly a billion dollars in CHIPS Act incentives with a billion from IBM, making it the largest award within a broader federal push for quantum manufacturing. Why it matters is less about one facility and more about what the funding model implies: a bet that superconducting qubits, manufactured on production-grade 300mm semiconductor wafers, have a clearer path from lab to factory. In other words, the government is leaning into an approach that looks most compatible with existing high-throughput fab infrastructure and faster iteration cycles. Supporters argue that speeding up learning cycles is how you win in complex hardware. Critics will worry about concentration—if the biggest manufacturing-scale support goes to one modality, alternative quantum approaches could struggle to reach the packaging and production maturity they’ll need later. Either way, it’s industrial policy picking a direction, not just writing research checks.

LLMs aren’t truly random

Staying in AI, but shifting from geopolitics to practical reliability: a GitHub research project tested whether OpenAI’s gpt-4.1 behaves like a uniform random number generator when you ask it to “pick a random number between 1 and 100.” The author ran a large batch of independent API calls and compared the distribution to what you’d expect from true randomness. The result was emphatic: the outputs weren’t remotely uniform. The model strongly preferred culturally familiar “random” choices—numbers like 37 and 42 spiked far above baseline. Even more interesting, it avoided round numbers so aggressively that most multiples of ten basically never appeared. There was also an odd dip in a well-known meme number, which the author speculates could be a training or safety artifact. The takeaway isn’t that the model is broken—it’s that language models are pattern machines. When you ask for randomness, you’re often getting a learned imitation of what humans *think* randomness looks like. If you’re doing simulations, games, lotteries, A/B-style selection, or anything fairness-sensitive, you should treat “LLM-generated randomness” as a footgun unless you’re drawing from a real RNG and using the model only for presentation.

White Rabbit precision timing networks

On the infrastructure side, White Rabbit popped up again—a technology for extremely precise timing distribution over Ethernet, originally developed around CERN. The headline claim is sub-nanosecond synchronization across large distributed systems, while still moving data on the same network. This matters because lots of modern systems aren’t limited by compute anymore—they’re limited by coordination. In scientific facilities, industrial control, and large measurement setups, the value of the data depends on time alignment: if your sensors, triggers, and acquisition nodes can’t agree on time, your results degrade fast. White Rabbit’s ongoing, open ecosystem is notable too: it’s not just a single vendor’s black box, and that openness can be the difference between a niche lab solution and something that becomes dependable infrastructure.

Jira automation becomes computation

Finally, a lighter—but surprisingly consequential—software story: Nicolas Seriot argues that Jira Automation is sufficient to implement a proven Turing-complete model of computation, and backs it up with a reproducible construction. Instead of hand-waving “anything complex is Turing-complete,” he maps a classic register-machine model onto Jira concepts—using issue counts as counters and workflow status as a program counter, with automation rules acting like instruction dispatch. Why it matters is the practical implication: once your automation system crosses into general computation, you’re no longer just “streamlining workflows.” You’re effectively running programs—programs with state, branching logic, and failure modes. That’s powerful if you’re disciplined, and dangerous if you’re not. It’s a neat proof, but also a reminder that low-code and automation platforms can quietly accumulate all the complexity of software engineering, whether teams are ready for it or not.

That’s our run for May-25th-2026. If there’s a theme today, it’s that the “soft” side of tech—governance, incentives, and human expectations—keeps colliding with the “hard” side, from fabs to timing networks. Links to all stories can be found in the episode notes. Thanks for listening—see you tomorrow.

More from Hacker News