Hacker News Daily Digest: Power Plays, AI Intrigue, and Tech’s Hottest Debates

The NSA is reportedly using an AI tool that the Pentagon has officially blacklisted. That tool is Anthropic's Mythos, and the fact that one arm of the U.S. government is using a technology another arm has forbidden tells you everything about the scramble for power in the age of AI. It’s a story about rules, and who gets to break them. And speaking of changing the rules, Apple has a new chief executive. John Ternus is taking over as CEO, with Tim Cook moving to the role of Executive Chairman. Ternus has been seen as the heir apparent for a while, running hardware engineering. But this officially marks a new era for the most valuable company in the world.

The question now is whether Ternus continues the Cook doctrine of operational excellence, or if he tries to recapture some of that Jobs-era product magic. It’s a transition from a supply chain mastermind to a product guy. We’ll see what that means for the next iPhone... or the device that comes after it. Meanwhile, the European Union is making a change that will affect every phone, not just Apple's. Starting in 2027, all phones sold in the EU will be required to have user-replaceable batteries. This is a huge win for the right-to-repair movement and a direct shot at the planned obsolescence that’s defined the smartphone industry for a decade.

The goal is to cut down on electronic waste and give consumers more control. Of course, the comments are on fire with debates about waterproofing, device thickness, and whether this will actually make a difference or just create a new market for shoddy third-party batteries. Forcing a design change this fundamental... it's the kind of move only a regulator the size of the EU could pull off. And from hardware back to software risk. Vercel disclosed a security incident from earlier this month, and the cause is a story we’re seeing more and more. A compromised third-party AI tool. Specifically, a bad OAuth app connected to that tool got access it shouldn't have had.

This is the new supply chain nightmare. It’s not about a bad component in a server rack; it's about the web of permissions and tokens we grant to all the little helper AIs that touch our code. As one person in the discussion put it, "When one OAuth token can compromise dev tools, CI pipeline, secrets and deployment simultaneously, something architectural has gone wrong." He's not wrong. The very thing that makes modern development fast—this ecosystem of integrated tools—is also what makes it so fragile. But let's go back to that NSA story, because it’s connected to something much bigger.

And it pairs perfectly with another story that was lighting up the discussion boards today: a college instructor who has gone back to using typewriters. Okay, so first, the NSA and Anthropic. The AI model is called Mythos. It's reportedly capable of finding novel, complex security vulnerabilities in codebases... autonomously. Think of it like a team of elite hackers that doesn't sleep and can read an entire operating system's source code before breakfast. The Pentagon blacklisted it as a supply chain risk—they don't want unaudited, powerful AI from a third party woven into their critical systems.

A totally reasonable policy. And yet, the NSA—the agency responsible for both signals intelligence and information assurance—is reportedly using it anyway. Why? The theory popping up in the threads is... well, it's brilliant. And a little cynical. The idea is that Anthropic has created a kind of "artificial scarcity" around Mythos. They've made it seem so powerful, so dangerous, that they can't possibly release it to the public. They give access only to a few trusted parties. This does two things. First, it acts as a safety measure, which is the official line. You don't hand out what might be a master key to the internet.

OpenAI did something similar years ago with GPT-2. But second—and this is the brilliant part—it creates insane demand. It frames Mythos not as a product, but as a strategic weapon. If you’re a government agency, and you hear there's a tool that can find flaws no one else can see, and your adversaries might get it... you can't afford not to have it. The Pentagon’s procurement rules become a suggestion. A bureaucratic hurdle to be routed around. It completely flips the sales dynamic. Anthropic isn't begging for a government contract. They've created a situation where the government is bending its own rules to get access.

We’ve seen this pattern before, but usually with hardware. Think about the race for stealth aircraft or advanced cryptography. A defense contractor builds something revolutionary, and the government classifies it to maintain an edge. Here, a private company has built the powerful thing, and they are the ones restricting access, which in turn makes the government want it more. It’s like a nightclub with a velvet rope, but for national security. And this is happening in the context of AI leaders like Sam Altman publicly saying they are scared of what they're building, comparing it to the Manhattan Project.

That kind of rhetoric only adds to the mystique and the perceived necessity of the tool. So on one hand, you have a powerful institution, the NSA, breaking its own system's rules to get its hands on a new technology it believes it can't live without. Now, let's look at the other side of the coin. The college instructor. The story is simple: to combat students using AI to write their papers, an instructor has mandated that all assignments be written on typewriters. It's such a visceral, almost comical image. A classroom filled with the clatter of keys and the ding of the carriage return.

But the discussion this sparked is what’s really important. It’s not just about "kids these days" or "lazy students." Someone in the comments made a point that reframes the whole issue. They said that for years, cheating was a privilege. If you had an older sibling who was an expert in the subject, or if your family could afford to hire a tutor or an essay writer, you had an advantage. Cheating wasn't fair. What ChatGPT and other AIs have done is democratize cheating. Now, everyone has an expert on call, for free. The outrage isn't that cheating is happening; it's that everyone can do it now.

The playing field has been leveled, but in the worst way possible. This problem is a direct result of a fifteen-year shift in education away from high-stakes final exams and toward continuous assessment—lots of smaller assignments spread throughout the semester. The goal was to reduce stress and encourage consistent learning. But it also created a system that is incredibly vulnerable to the kind of plagiarism AI excels at. So what does the instructor do? They can't un-invent the technology. They can't put a proctor in every student's dorm room. So they reach for the one thing that can't connect to the internet.

A typewriter. Where have we seen this before? Every time a new technology challenges an old way of testing knowledge. Calculators in math class. Wikipedia for history reports. The institution's response is always the same: first ban it, then try to create rules to contain it, and finally, grudgingly, accept it and change the way you teach and test. The typewriter is a retreat. It's a tactical withdrawal to a time when the rules of the game were clear. But here’s where the analogy to calculators breaks down. A calculator helps you get the right answer faster. AI can generate an entire essay, with a thesis, supporting arguments, and a conclusion, that is entirely plausible and grammatically correct.

It mimics the form of understanding, even without the substance. The typewriter doesn't solve that. A student can still generate the essay on their phone and then just manually type it out. So the typewriter isn't really a technical solution. It’s a theatrical one. It’s a statement of values. It's an attempt to force a different, slower, more deliberate process on the student. It's an attempt to control the workflow, not just the final output. So look at these two stories together. The National Security Agency and the college classroom. They seem worlds apart, but they are dealing with the exact same problem.

A technology has emerged that is so powerful it breaks the existing rules. The NSA's response is to ignore the rules to gain an advantage. They see Mythos as a key to power, and their own internal policies are an obstacle to be overcome in the race to acquire it. They are pulling the future forward, even if it means breaking things. The instructor's response is the polar opposite. They see AI as a threat to the integrity of their system, so they are rewinding the clock. They are retreating to a technology—the typewriter—from a time when the rules still made sense, to create a small bubble where the new reality doesn't apply.

One is about acquisition, the other is about preservation. One breaks the rules to get the tech, the other resurrects old tools to escape it. Both are rational responses to a world changing faster than our institutions can adapt. This week showed us the two paths for dealing with overwhelming change. You either break the rules to get an edge, or you build a fortress of old ones to keep the future out.