0:00 An interactive website letting you sift through the internal documents of the Theranos scandal just shot to number one on Hacker News.0:09 Just when you think that story is over, it finds a way back, which feels right—because as we talked about last week with Margaret Hamilton, the human side of technological progress, and its failures, is always the part that sticks with us.0:26 This new site, Theranos.world, is a stark reminder of that.0:30 It’s not just a retrospective; it’s an interactive autopsy of a dream that became a disaster, powered by modern AI that parses the original evidence for you to explore.0:42 And that’s just the top of the feed.0:44 The conversations today are all over the map, but they keep circling this theme of belief versus reality.0:52 For instance, there’s a massive thread on a 2025 study from Frontiers in Psychiatry that’s getting a fresh look.1:00 It presents compelling evidence that ADHD is deeply linked to circadian rhythm disorders.1:06 We’re talking up to eighty percent of adults and eighty-two percent of children with ADHD also having significant sleep issues like insomnia.1:16 The paper isn't just pointing out a correlation; it's proposing that for many, ADHD might be a circadian rhythm disorder.1:24 It details how things like delayed melatonin onset and blunted cortisol rhythms are common in people with ADHD.1:32 The fascinating part is the proposed solution: not just medication, but behavioral changes.1:39 Things like morning bright light therapy and low-dose melatonin to essentially reset the body’s internal clock.1:47 It reframes the conversation from a pure focus-and-attention issue to a fundamental problem of timing and sleep, which for a lot of people in tech, feels… uncomfortably familiar.1:59 Then, for something a little lighter but no less impressive, the classic 1996 shooter Quake has been completely ported to safe Rust and is now playable in a browser.2:11 This isn't some clunky emulator.2:13 It’s a full-blown port called QUAKE·SRP.2:16 It supports keyboard, mouse, gamepad, all the original mission packs, and it even saves your game state in your browser's local storage.2:25 You can pop open the console and use all the old id Software cheat codes.2:31 It’s this perfect piece of nostalgic engineering.2:34 Taking a beloved, thirty-year-old game and rebuilding it with modern tools that guarantee memory safety—something the original C code most definitely did not have—is just a beautiful encapsulation of the hacker ethos.2:49 It’s preserving history by rewriting it in a safer, more accessible language.2:55 Speaking of fundamentals, there’s a great, nerdy discussion about the subtle but maddening differences between Windows and Mac keyboards.3:04 The piece was written by Marcin Wichary, a UX designer who’s been at Google, Medium, and Figma, so he knows a thing or two about how people interact with computers.3:16 He gets into the weeds on things you feel every day but maybe never consciously name.3:22 Like how "Backspace" on Windows is "Delete" on a Mac, but "Delete" on Windows is "Forward Delete" on a Mac, a key most MacBooks don't even have.3:32 And then there are the modifier keys: Control, Alt, and Shift on Windows versus Command, Option, Control, and Shift on Mac.3:41 It’s a reminder that the most basic layer of our interaction with technology is built on these historical accidents and design decisions that we’re all still navigating decades later.3:54 And finally, before we dive deep, there are two stories about artificial intelligence and mathematics that perfectly capture the tension of this moment.4:04 On one hand, you have a guest post on Fields Medalist Terence Tao’s blog.4:10 It’s written by Álvaro Lozano-Robledo, and it directly addresses the anxiety that math students are feeling about AI.4:18 They’re worried that AI will be able to out-think human mathematicians in five or ten years, making their career paths obsolete.4:27 The advice?4:28 "Keep calm and carry on studying math." It’s a call to focus on passion and perseverance, arguing that AI is a tool, not a replacement.4:37 But then, on the other hand, there’s a new preprint from OpenAI on a foundational math topic—the Partition Principle—that’s getting absolutely roasted by a set theory expert.4:49 He basically says the paper is so unclear, uses such strange terminology, and cites sources so poorly that if it were submitted to a real academic journal, it would be rejected without even going to peer review.5:04 A "desk rejection." So in one corner, you have the fear that AI is about to become a god-tier mathematician.5:12 And in the other, you have an expert saying that a top AI lab can’t even write a coherent paper on the subject.5:20 So what does it all add up to?5:22 It seems like we're caught between the story we tell ourselves about technology, and the much messier reality.5:30 Okay, let's start with Theranos.5:32 Why, in late 2026, are we still so captivated by this story?5:36 I think this new website, Theranos.world, gives us a clue.5:40 It’s not just another article or documentary.5:43 It's an interactive archive of the collapse.5:47 The creator, who goes by kbyatnal on Hacker News, used APIs from a company called Extend.ai to parse all the publicly available Theranos PDFs—internal emails, lab reports, investor presentations—and make them searchable and explorable.6:03 So you can go on this site right now and not just read about the damning evidence, but see it for yourself.6:11 You can see the email from April 11th, 2014, where insiders were already raising red flags.6:17 You can see the manipulated reports sent to investors.6:21 It turns the user from a passive consumer of the story into an active investigator.6:27 And that’s a powerful shift.6:29 This is a pattern we've seen before, but with a modern twist.6:33 Think about other moments of intense public scrutiny.6:37 The Zapruder film of the Kennedy assassination, or the Watergate tapes.6:42 In those cases, the raw evidence became the focal point of the entire national conversation.6:48 People weren't just debating what happened; they were debating the interpretation of a specific frame of film, a specific snippet of audio.6:58 The primary source became the story.7:01 Theranos.world is the 2026 version of that.7:04 It takes a sprawling, complex corporate fraud and turns it into a dataset you can query.7:10 It’s a new form of digital archaeology.7:13 The analogy isn't perfect, of course.7:15 The Zapruder film was a single, shocking piece of evidence.7:19 The Theranos archive is a mountain of mundane corporate communications that only becomes damning when you piece it all together.7:28 But the impulse is the same: the desire to get past the narrative and touch the raw facts.7:35 What’s also telling is the social layer built around it.7:39 The site is hosting a watch party for an upcoming documentary in San Francisco on October 22nd.7:45 It’s turning this solitary act of digital investigation into a communal event.7:51 It suggests our fascination isn't just with the fraud itself, but with the shared experience of dissecting it.7:59 It’s like a true-crime podcast club, but for corporate governance.8:03 The tech behind it is also part of the story.8:07 The creator says, "hope our content marketing made your day a bit more fun!" which is such a classic Hacker News comment.8:15 They built this incredibly detailed forensic tool as a way to showcase their API.8:21 It’s a demonstration of how modern AI tools—in this case, for document parsing and understanding—can unlock insights from vast, unstructured archives.8:31 It’s a positive application of AI being used to hold power to account, which is a nice contrast to some of the other AI stories we're seeing.8:41 The site is a monument to a failure, built with the tools of progress.8:46 And that tension is exactly why we can’t look away.8:50 Now let’s pivot to that other big tension of the day: the collision of artificial intelligence and pure mathematics.8:58 This is where things get really complicated, because you have two completely opposite narratives happening at the same time.9:07 First, you have the human story, captured on Terence Tao’s blog.9:11 This isn’t just some abstract debate.9:14 A student literally wrote in, saying, "Many students are asking themselves whether going for a PhD in math is the right career move at this time," expressing this deep, existential fear that their chosen field is on the verge of being automated into irrelevance.9:32 This is the voice of real anxiety.9:35 These are people who have dedicated their lives to one of the most rigorous intellectual pursuits, and they're looking at large language models and wondering if they've made a terrible mistake.9:48 The response from Álvaro Lozano-Robledo is, essentially, therapeutic.9:53 "Keep calm and carry on studying math." He argues that the joy of discovery, the beauty of the process, and the deep thinking involved are things that can't be replaced, even if an AI can solve a problem faster.10:08 It’s an appeal to the intrinsic value of the work.10:12 Where have we seen this before?10:14 This is the John Henry story, updated for the 21st century.10:18 The steel-driving man versus the steam drill.10:21 Except here, the fear isn't just about being slower; it's about being fundamentally less capable.10:28 This has happened in every field that technology has touched.10:32 We saw it with chess, when Deep Blue beat Kasparov.10:36 For a while, people wondered if chess was "solved" or "dead." But it wasn't.10:41 Human chess is thriving.10:43 Players now use engines as training partners.10:46 The game evolved.10:47 The advice on Tao’s blog is betting on the same evolution for mathematics.10:53 AI becomes a collaborator, a tool, a "mathematical steam drill" that lets humans tackle even bigger challenges.11:00 But here's where the analogy starts to break down, and where the second story comes in.11:07 The OpenAI preprint.11:08 This isn't a story about AI being too good.11:11 It's a story about the culture of AI research being sloppy.11:15 A set theory expert named Asaf Karagila wrote a blistering critique of a new paper from OpenAI.11:22 The paper makes a claim about something called the Partition Principle, a deep, foundational concept in set theory.11:30 Karagila’s takedown is brutal.11:32 He says the paper is poorly structured, the terminology is non-standard, and it makes it hard to even understand what is being claimed.11:42 He points out that it cites unpublished lecture notes and even misrepresents his own work.11:48 His final verdict is the most damning line of the week: "If this was an academic paper submitted to a journal, it should be issued a desk rejection." That’s not an argument about whether the AI's conclusion is right or wrong.12:04 It's an argument that the work is so shoddy it doesn’t even qualify for a proper debate.12:10 It fails to meet the basic standards of communication and rigor that have been built up in mathematics over centuries.12:18 This is the other side of the "move fast and break things" coin.12:23 In software, you can ship a buggy product and patch it later.12:27 In mathematics, a proof is either right or it's not.12:31 And to even begin to determine that, you have to present your argument in a clear, verifiable way, using the shared language of the field.12:41 The critique suggests OpenAI, in this case, failed at that first step.12:46 So, while students are panicking that AI is a superhuman mathematician, the experts are pointing out that the output—or at least, the human-curated presentation of that output—can be amateurish.12:59 This isn't John Henry versus the steam drill.13:02 This is more like a slick company selling a "revolutionary" new hammer that, on closer inspection, is just a rock tied to a stick, and they've written the instruction manual in gibberish.13:16 The pattern here isn't one of technological replacement.13:20 It’s one of institutional credibility.13:22 A prestigious AI lab, trading on its name, puts out work that would be laughed out of a university department.13:30 It highlights the culture clash between the hype-driven, fast-paced world of AI development and the slow, methodical, deeply skeptical world of academic mathematics.13:42 One world values disruption and impressive demos.13:45 The other values proofs that hold up for a hundred years.13:49 So what happens when these two worlds collide?13:52 You get this exact situation: widespread anxiety about AI's capabilities, running in parallel with expert dismissal of its actual output.14:02 Both can't be right.14:03 Or maybe...14:04 they both can.14:05 Maybe the potential is real, but the process is currently broken.14:10 And navigating that gap is the real challenge.14:13 So, where does this leave us?14:15 This week, Hacker News held up a mirror to some of the biggest anxieties in tech.14:21 The ghost of Theranos reminds us how easily a compelling narrative can overwhelm inconvenient facts, and how much work it takes to piece the truth back together after the fact.14:33 The new interactive site is a testament to that—a tool for forensic accountability built long after the damage was done.14:42 Then you have the twin stories from the world of mathematics.14:46 One reflects our fear of being made obsolete by machines that seem infinitely capable.14:52 The other reveals the fallible, human-driven process behind that machine's output, showing that even the most advanced AI labs can stumble when they enter a domain built on centuries of rigor.15:06 The thread that ties it all together is the question of trust.15:10 How do we validate claims in an age of overwhelming information and powerful storytelling?15:16 The Theranos saga was a failure of venture capital, of journalism, of regulatory oversight.15:23 It was a systemic breakdown in the mechanisms we use to verify claims.15:28 The OpenAI paper controversy is a smaller, more academic version of the same problem.15:34 When a credible source produces a questionable result, how does the community respond?15:40 This week sets up a bigger question that we're going to be dealing with for years.15:46 As AI gets more powerful, the role of the human expert doesn't diminish—it becomes MORE critical.15:52 You need the domain expert, the set theorist, to tell you when the AI is hallucinating.15:59 You need the investigative user, digging through Theranos.world, to reconstruct a timeline of fraud.16:06 The technology itself is neutral.16:08 It's the human process of validation, skepticism, and accountability that gives it meaning.16:14 And right now, that process is under more strain than ever.
