Hacker News Daily: AI Cracks Ancient Codes and WordPress Drama Unfolds

An AI just solved a 370-year-old cipher in less than a day. That code, the Cyphral Distich, has been an unsolved puzzle for cryptographers since the 17th century, and Anthropic's new model, Claude Fable 5.1, cracked it. This comes just as we were talking last week about AI agents learning to cheat and lie. It seems while we're worrying about their ethics, they're also developing this uncanny ability to see patterns that have been invisible to human eyes for centuries. So, the big question this week isn't just about what these systems can do, but what kind of thinking they represent. Are they just powerful calculators, or are they developing a form of intuition we don't understand yet? The answer seems to be a bit of both, and it's forcing a lot of uncomfortable re-evaluations across the board.

Okay, let's sweep the other big stories that had the community talking this week. First, a story of corporate survival. Matt Mullenweg is back as CEO of Automattic. You know, the parent company of WordPress. This comes right after an attempted ouster by his own board. TechCrunch confirmed on September 12th that he's been reinstated as chairman and CEO with, quote, "full support of the board." That's the kind of whiplash that usually signals deep trouble, but here it's being spun as a return to stability. It’s a classic founder-versus-board drama, but with a speedrun ending. We've seen this pattern before, most famously with Steve Jobs at Apple. The question is whether this is a sign of Mullenweg's indispensable vision or just a temporary truce.

Speaking of giants, Nike got dropped from the S&P 100 index. This happened after the company lost a staggering 200 billion dollars in market capitalization. After eighteen years in the index, that's not just a stumble; it's a fall off a cliff. For a brand that has defined athletic apparel for a generation, this is a massive signal of decay. It's a reminder that market dominance is never permanent, and brand loyalty can't shield you from fundamental business problems forever. The discussions on Hacker News were less about sneakers and more about what happens when a cultural icon loses its financial footing. On a much more positive note for builders and makers, Apple just did something... un-Apple. They published a massive library of detailed, technical dimensional drawings for almost everything they make.

We're talking iPhones, iPads, MacBooks, even the little MagSafe chargers. For anyone who's ever tried to design an accessory for an Apple product, this is a godsend. For years, the community has relied on reverse-engineering and third-party scans. Now, the official schematics are just... there. It’s a small move, but it signals a potential shift towards a more open hardware ecosystem, something Apple has historically resisted. Then there’s the privacy front. Signal is officially working on a way for you to register without a phone number. This has been the holy grail for private messaging for years. The key is their use of zero-knowledge proofs, or ZKPs. It's a cryptographic technique that lets you prove you know something—like a secret password or token—without actually revealing what that secret is.

Signal already uses ZKPs for other features, like donation badges. This move would decouple your Signal identity from your phone number, which is a huge privacy win and a major step toward making secure communication truly anonymous. Of course, not all platform news is good. A huge thread blew up about Google still serving dodgy, scammy ads. A post on Atomic14 laid it all out, and it resonated with hundreds of people. The core question is simple: Why, with all its resources and AI, is Google's ad network still so full of junk? It feels like a problem that should have been solved a decade ago. But the scale of their automated system means it's a constant battle against bad actors who are always finding new exploits. It's a perfect example of a system so large and complex that it's perpetually vulnerable at the edges.

And finally, a quick update on a story about digital archiving. The XCancel service, which was a read-only archive of Twitter, or X, has been taken down again. The site just says it's suspended due to "a new development in the ongoing legal proceedings." This is the classic cat-and-mouse game between archivists trying to preserve public data and platforms trying to control it. Every time a service like this pops up, the legal threats aren't far behind, raising all the usual thorny questions about copyright, fair use, and who really owns the public square. So let's dive deeper into the two stories that I think reveal the most about where we are right now. First, that 370-year-old cipher. The code is called the Cyphral Distich, and it comes from a 17th-century Scottish writer and polymath named Sir Thomas Urquhart.

He was a character—a royalist who fought against Cromwell, was captured, and wrote fantastical books while imprisoned in the Tower of London. In one of those books, he included this cipher, promising it contained a secret. And for nearly four centuries, nobody could crack it. Then, vals.ai points their new model, Claude Fable 5.1, at it. And within a day, it's solved. Now, here's the part that is, as the researchers put it, "embarrassing for humans in hindsight." The AI didn't use some kind of unimaginable, brute-force computation. It didn't just try every possible key. Instead, it did what the best human cryptographers do: it looked for a pattern in the context. It figured out that the key was hidden in plain sight, in another part of Urquhart's own writing.

Specifically, in a list he called the "32 Proquiritations." The cipher text contained numbers, and the AI correctly inferred that these numbers were word indices. The first number pointed to the first word in the Proquiritations, the second number to the second word, and so on. When it assembled the words, it revealed a royalist prayer for the restoration of King Charles II. It even ended with a rhyming two-line verse, a "distich," just as the name of the cipher promised. It was self-verifying. So what does this mean? This isn't just about AI being faster. This is about AI demonstrating a form of lateral thinking. It connected two seemingly unrelated pieces of information—the cipher and a list of words—and correctly guessed their relationship.

This is a creative leap. It’s the same kind of insight that led to the cracking of the Enigma code in World War Two. The codebreakers at Bletchley Park didn't just use machines; they exploited human weaknesses, like the fact that German operators would often use predictable phrases in their messages. They looked for patterns in the system, not just the math. And that's the parallel here. The AI saw the system. It understood that Urquhart, the author, was the system. His own writings were the context. This goes way beyond the "stochastic parrot" argument that these models are just regurgitating what they've been trained on. This was a novel synthesis. It's a reminder that intelligence isn't always about raw processing power. Sometimes, it's about seeing the clue that everyone else has dismissed as noise.

And right now, AI is getting very, very good at that. It makes you wonder what other "unsolvable" problems are only unsolvable because we've been looking at them the wrong way. Now for the second big thread: the case against the image format JPEG XL. This one is fascinating because it’s a story about how superior technology can still lose. It's a perfect case study in ecosystem politics and the tyranny of "good enough." So, JPEG XL is, by many technical measures, an incredible image codec. It offers better compression than old-school JPEGs, it can do lossless compression better than PNGs, and it has modern features like high dynamic range and wide color gamuts. It was designed to be the successor to everything. The one format to rule them all.

And yet, it failed to get adopted. The pivotal moment was in 2023, when the Google Chrome team decided to remove support for it. That was basically a death sentence for its web ambitions. And the Hacker News thread this week dug into a blog post by Gianni Rosato that meticulously breaks down why. The argument isn't that JPEG XL is bad. The argument is that it's not better enough in the ways that actually matter. Think about it from the perspective of a giant platform like Google or Facebook. They serve billions of images a day. Their primary concern is lossy compression—making images as small as possible while still looking acceptable. On that front, JPEG XL is good, but it's not a world-beater. Other modern codecs, like AVIF, which is derived from the AV1 video codec, are often better, especially at very low bitrates.

They can produce smaller files at the same perceptual quality. So what about JPEG XL's killer feature, its excellent lossless compression and its ability to losslessly transcode old JPEGs? Well, the post points out that while its lossless files are smaller than, say, lossless WebP, the advantage is only about twelve percent. That's a nice-to-have, not a must-have. It's not a big enough leap to justify the massive engineering effort of deploying a whole new format across the entire web ecosystem—browsers, servers, content delivery networks, image editing software. Everything would have to be updated. This is a classic pattern. You can probably see the analogy coming. It's Betamax versus VHS. Betamax was technically superior in some ways.

It had better video quality. But VHS was cheaper, had longer recording times, and, crucially, it got the backing of more movie studios. The ecosystem chose VHS. The "good enough" solution that had the right market dynamics behind it won. JPEG XL is Betamax. It's technically elegant, but it arrived in a world that had already found a "good enough" solution in WebP and was looking ahead to AVIF for the next big leap in lossy compression. The marginal gains it offered for lossless weren't enough to overcome the immense inertia of the existing system. Now, the story isn't completely over. The post does note that a recent major security flaw in WebP has made some people reconsider the safety of having a single dominant image format.

And there's work being done to integrate a more secure, Rust-based JPEG XL decoder into browsers like Firefox. So maybe there's a path for it as a niche, high-quality archival format. But for mass adoption on the web? The ship has likely sailed. And it's a powerful lesson. In technology, success is never just about technical merit. It's about timing, politics, and whether your solution solves a big enough pain point to justify the cost of switching. JPEG XL solved a problem, but for most of the web, it just wasn't a painful enough one. So what does it all add up to? We have an AI making creative leaps that were once the exclusive domain of human genius. At the same time, we have a technically brilliant image format failing because of human systems—market inertia and the simple calculus of "is it worth the effort?" The thread connecting these stories is about the nature of problem-solving.

The Cyphral Distich wasn't a math problem; it was a human problem, a literary puzzle. And the AI solved it by thinking, in a way, like a historian. Meanwhile, JPEG XL's failure wasn't a technical problem; it was an ecosystem problem, a political one. It feels like we're at an inflection point. The purely technical challenges are becoming, in some ways, the easier part. Claude Fable 5.1 shows that AI can handle immense complexity and find subtle, hidden patterns. But the JPEG XL story shows that even the most elegant technical solution can be defeated by the messy, complicated reality of human adoption. This week sets up a future where the most important skill won't just be building the best technology. It will be understanding the systems—the human systems, the economic systems, the political systems—that technology has to live in.

Building the better mousetrap is one thing; getting the world to actually use it is another game entirely. And that's a game where the rules are far more complex than any cipher.