Podcast  ›  Episode 17

Bitcoin's Crash, the Riemann Proof, and Media's Turbulent Week

Nov 14–21, 2026 · Nov 21, 2026 · 67 stories

Bitcoin hits a two-year low, a mathematician claims a historic proof, and the media business faces union drives, AI reporters, and billion-dollar deals.

Listen

Audio is not ready for this episode yet. You can still read the full transcript below.

Welcome to ReadFlow Weekly. It's Saturday, November 21st, 2026.

This week, we are being asked to believe in some pretty extraordinary things. To trust in systems, in stories, and in the very foundations of knowledge itself. And the results of that trust, or the lack of it, are playing out in real time, from the financial markets to the newsrooms to the far edges of human understanding.

Let's start with a system built entirely on trust, or perhaps, speculation. The cryptocurrency market is in a state of profound vertigo. Bitcoin has plummeted to a two-year low, shedding more than half its value since its peak and continuing what is now its worst month since 2022. It's a stark reminder that in the absence of traditional anchors, value is a collective story we tell ourselves. And right now, that story is one of a steep and painful slide.

Contrast that with a story of verification so meticulous, it seeks to end a 165-year-old debate. This week, we have not one, but two monumental claims regarding the Riemann Hypothesis, one of mathematics' most famous and consequential unsolved problems. First, an Anthropic mathematician stepped forward with a claimed proof. But perhaps more compelling is the parallel effort from the open-source Lean theorem-proving community, which announced it has verified a proof in a library comprising 17 million lines of code. This isn't just about a brilliant individual insight; it's about constructing a verifiable digital edifice of logic so others can walk through every step and know, with computational certainty, that it holds. In a world of shaky trust, here is an attempt to build a proof you can actually lean on.

And between these two poles, the volatile trust in crypto and the painstaking verification of a math proof, sits our entire media ecosystem, which is having something of an identity crisis. Staff are voting to unionize at institutions like the Internet Archive, while others are facing deep cuts at the BBC, CBS News, and smaller outlets like the San Francisco Standard. A potential multi-billion dollar settlement looms over YouTube. And in a particularly telling shift, several crypto news outlets appear to have replaced human reporters with AI-generated content. The question of what we read, who writes it, and who pays for it has never been more urgent. It all circles back to that core theme: what, and who, do we choose to believe?

From the trust we place in our wallets, to the trust we place in our headlines, to the trust we place in a mathematical statement about the arrangement of prime numbers, this is a week about verification and volatility. And that's just where we begin.

You're listening to ReadFlow Weekly for the week of November 14th through the 21st. Let's get into it.

Week at a glance

Welcome back. If our cold open was about the big question of trust, then this, our week at a glance, is the map of where that question is being answered, or violently ignored, across the landscape. We are covering a lot of ground, so let's chart the course.

First, in technology and AI, we are witnessing a fascinating tension between foundational verification and volatile trust. The crypto market is in a steep slide, with Bitcoin hitting a two-year low, a stark reminder of the financial systems we choose to believe in. Meanwhile, in a stunning parallel, the mathematical world is grappling with a claim of ultimate verification: a proof of the 165-year-old Riemann Hypothesis, backed by a separate, massive Lean community verification. It's a profound moment about proving what is true.

This theme of securing our systems continues with critical security advisories for tools like GPGME and urgent calls to plant "canary tokens" in our AI supply chains. And for the builders, we have deep dives into testing philosophy and a fascinating C# fork that compiles down to tiny, runtime-free native binaries.

Second, in business, the media industry is undergoing a turbulent reckoning. We're seeing unionization drives, like the one at the Internet Archive, collide with significant staff cuts at places like the BBC, CBS News, and the San Francisco Standard. This human reshuffling happens against a backdrop of massive financial moves, including a potential $3 billion settlement by YouTube and the record-shattering, nearly $4 billion opening day for Grand Theft Auto VI. And hovering over it all is the unsettling question of whether the bylines themselves are human, as several crypto news outlets appear to have replaced reporters with AI.

Finally, in science, we're marking milestones that stretch our understanding of time and origin. A Mercury-bound X-ray telescope delivers its first images after an eight-year journey, while Voyager 1 officially reaches a distance of one light-day from Earth. Closer to home, new research suggests our own human family tree needs a radical rewrite, complicating the story of our origins. And from the foundational to the futuristic, CERN detects an ultra-rare particle decay after a 40-year search, and Google partners with Blue Origin to launch experimental orbital data centers.

That's our path this week: from the verification of a mathematical holy grail to the volatility in our markets and media, all set against a backdrop of cosmic and evolutionary discovery. The thread, as you've likely gathered, is about what we choose to believe and the systems we build to prove it. Let's get into the details.

Technology & AI

Welcome back to ReadFlow Weekly. If our opening was about the grand, abstract question of trust, then our Technology and AI segment is where that question gets its hands dirty. It's about the systems we build our world on, the mathematical bedrock we assume is solid, and the volatile digital assets we're told to believe in. This week, that tension between foundational verification and spectacular volatility isn't just philosophical. It's playing out in real time, from a 165-year-old math puzzle to a cryptocurrency in freefall.

Let's start with the most dramatic story of trust, or perhaps its evaporation, in the digital economy. Bitcoin has plunged to a two-year low. As we reported last week, it had already dropped 24 percent in 46 days. This week, it fell another 9.4 percent, landing at $56,085 by Friday night. That means it has now lost more than half its value since its last peak. This is its worst month since 2022. The story here isn't just the price chart, though that's startling enough. It's about what happens when a system built on decentralized trust meets a crisis of confidence. The narrative of Bitcoin as digital gold, as an inflation hedge, as an uncorrelated asset, is being stress-tested in a very public way. Who wins in this? Short-term traders with iron stomachs, perhaps. Who loses? The everyday investors who bought into the peak of the hype cycle, and the crypto-native businesses whose models depend on perpetual growth and high transaction fees. What to watch next is whether this is a cleansing correction or the start of a more profound winter. Does trust in the underlying blockchain technology survive the brutal repricing of the asset sitting on top of it? The volatility we're seeing is a stark reminder that technological innovation and financial value are not the same thing, and their relationship is often turbulent. You can follow the ongoing slide in our coverage at Bitcoin Drops to $56,085.

Now, let's pivot from a system struggling with trust to perhaps the ultimate act of verification. This week, we may have witnessed one of the greatest mathematical achievements in history. An Anthropic mathematician has claimed a proof of the Riemann Hypothesis. This isn't just any math problem. It's a 165-year-old conjecture about the distribution of prime numbers, a holy grail with implications across number theory and cryptography. But here's the fascinating, modern twist. This isn't just a claim on a chalkboard. Separately, the Lean and Mathlib communities have released a joint statement. They report that on November 20th, a proof was released as a Lean library comprising 17 million lines of code, and they have undertaken verification. For those unfamiliar, Lean is a theorem prover. You don't just read the proof, you run it. The computer checks every logical step. This is verification at a scale and rigor previously unimaginable. Why does this matter? It transcends mathematics. It's a blueprint for how we might verify other impossibly complex systems, from AI model training runs to critical financial software. It moves truth from the realm of expert consensus, which can be slow and fallible, to the realm of computational verification. Who wins? All of human knowledge, if this holds. The entire field of mathematics gets a new foundation to build upon. What to watch next is the long, meticulous process of peer review, both human and computational. Does the Lean verification hold up to intense scrutiny? Does the Anthropic mathematician's formal paper align with the code? This is where centuries-old intellectual pursuit meets the cutting edge of how we establish truth. It's a profound story of verification, and you can find our coverage of the Lean community's landmark effort right here on ReadFlow at Lean verification of the Riemann Hypothesis.

Speaking of verifying and securing complex systems, our next arc brings the challenge down to earth, to the very code and supply chains that run our world. Two stories this week highlight the spectrum of digital defense. First, a critical security advisory. The GNU Privacy Guard team announced a new release of GPGME, the library many apps use for encryption, fixing a critical vulnerability tracked as CVE-2026-129975. This is the plumbing of privacy for countless applications. When it has a flaw, the potential for silent, widespread compromise is enormous. The fix is a reminder of the quiet, essential work of maintaining our digital infrastructure.

But what about threats we can't even see coming, especially in the frenetic world of AI? That brings us to our second story. Security researchers are issuing a call to action, arguing that your AI supply chain needs canary tokens. The idea is brilliant in its simplicity. A canary token is a piece of digital bait, a fake API key, a dummy file, a decoy model weight. You plant it in your system, your development environment, your training data pipeline. If something or someone exfiltrates that token, it triggers an alert. You've just discovered a breach you might have otherwise missed for months. The argument is that with AI, the supply chain is horrifically complex and opaque. You're pulling in models, datasets, code from thousands of sources. Traditional perimeter defense isn't enough. You need tripwires inside the house. Why does this matter? Because as we rush to build and deploy AI, security is too often an afterthought. The winners will be the organizations that bake in these paranoid, defensive measures from the start. The losers will be anyone who thinks a shiny new AI model isn't a colossal attack surface. What to watch next is whether these practices move from advanced research teams to standard enterprise DevOps playbooks. The combination of a critical crypto library patch and a strategic shift towards internal tripwires paints a picture of a security landscape that's evolving from just building higher walls to also placing sophisticated sensors in every room.

Our final arc is for the builders, the engineers who have to create reliable software in this complex environment. And this week offered two fascinating tools for the toolbox. First, a compelling argument that deterministic simulation testing is for everyone. Often relegated to the world of distributed systems, think databases, coordination services, the author argues this technique should be used for all applications. The core idea is powerful. You record all the non-deterministic inputs to your program, the time, random number generation, network messages, and then you replay them exactly. This allows you to recreate bugs perfectly, to run the same test a thousand times, to truly understand the chaotic behavior of your code. It's a methodology for imposing order on the inherent chaos of software execution. It's about verifying that your application does what you think it does, every single time.

And then, from methodology to implementation, we have a fascinating project called Zero C#. It's a fork of a compiler called bflat that takes C# code and compiles it straight to a tiny native binary. No .NET runtime. No garbage collector. Nothing to install on the target machine. You keep C#'s modern syntax and safety features, but you get a standalone executable that's incredibly small and fast to start. Memory management is handled through arenas, you allocate freely and then free everything in a chunk. This is C# escaping its traditional, managed environment and competing in spaces dominated by Rust, Go, or C. Why does this matter? It opens up new domains for C# developers: embedded systems, high-performance utilities, places where deploying a full runtime is impossible. It's about choice and control. The winner is the developer who needs both productivity and bare-metal footprint. What to watch next is whether this moves from a cool experiment to a supported toolchain. Does it gain a ecosystem? Does Microsoft, the steward of C#, see this as a fringe project or the future of the language's reach?

Together, these developer stories are about taking back control. Deterministic simulation testing gives you control over your program's behavior in time. Zero C# gives you control over your program's presence on a machine. They are responses to the complexity of modern software, offering paths to greater reliability and greater independence. They are, in their own way, acts of verification and simplification in a stack that constantly threatens to spiral into chaos.

So, let's tie this together. This week in technology and AI was a masterclass in contrasts. We have Bitcoin, a system where trust is market-driven, volatile, and currently in crisis. We have the Riemann Hypothesis, where trust is being sought through the most rigorous computational verification imaginable. We have security, where trust in our systems is maintained through both urgent patches and strategic, paranoid foresight. And we have the tools for builders, who are creating new methods to make their own software more trustworthy and controllable. The through-line is verification. How do we know something is true, or secure, or reliable? The answers this week range from the price on a screen, to 17 million lines of verification code, to a canary token silently sending an alert, to a test suite that can replay time itself. In a world asking us to believe more and more complex stories, the work of proving them has never been more critical, or more fascinating.

Business

Welcome back to ReadFlow Weekly. If our opening was about the grand, abstract question of trust, then our Technology and AI segment was about the systems we build that world upon. Now, in our Business segment, we get to see the very human and economic consequences of where that trust is fraying, and where it's being reforged under new, and often brutal, terms. This week, the story is media. It's a story of labor, of legacy, of artificial intelligence, and of staggering sums of money changing hands, all while the ground beneath an entire industry continues to shift.

Let's start with the labor movement, because it's providing a fascinating counter-narrative to the doom and gloom. Over eighty of the one hundred and twenty-four eligible US staffers at the Internet Archive have voted in favor of unionizing. They are now awaiting voluntary recognition from founder Brewster Kahle. Think about that for a second. The Internet Archive, that beloved digital library, that bastion of preservation in the chaotic web, is seeing its own workforce organize. It's a powerful signal that the mission-driven, non-profit corners of the tech-media world are not immune to the same questions of workplace equity and voice that are sweeping through corporate America. This is about building a sustainable future for the people who safeguard our digital past.

Contrast that with the stark reality playing out in more traditional newsrooms. The BBC is facing a dire warning from the UK's National Audit Office, which says that the broadcaster's planned five hundred million pound cuts could lead to a catastrophic brain drain, directly hurting its editorial integrity and hard-won reputation. Across the Atlantic, dozens of CBS News employees have accepted buyout offers from parent company Skydance, a move significantly affecting the foreign desk. And in San Francisco, the local news outlet The San Francisco Standard closed the workplace media company Charter it had acquired just last year, cutting twelve jobs. These are not abstract budget lines. These are beats going uncovered, expertise walking out the door, and local accountability weakening.

And then there's the money. So much money. In one corner, we have YouTube, which according to sources has reached an agreement in principle to pay roughly three billion dollars to US school districts over claims that its platform fosters addiction. A final deal may still be months away, but the sheer scale of that potential settlement is breathtaking. It's a direct financial consequence of the societal reckoning with social media's impact.

In another, entirely different corner, we have the pure, unadulterated power of blockbuster entertainment. Take-Two Interactive announced that Grand Theft Auto VI generated three point seven nine billion dollars in pre-order and day-one sales. To put that in perspective, that surpasses the entire global box office gross of Avengers: Endgame. It's a reminder that while one segment of media grapples with existential threats, another operates on a financial plane that dwarfs almost every other form of entertainment. The market for escapism, it seems, remains incredibly robust.

But bridging these two worlds, the struggling news industry and the ascendant world of AI, we find one of the week's most unsettling business stories. A report indicates that crypto news outlets like Decrypt and The Defiant appear to have replaced human reporters with AI-generated content. CoinDesk, notably, uses an "AI Boost" byline for stories where AI assists human editors. This comes from a Reuters report covered on Mediagazer. The context here is critical: the once-lucrative financial publishing sector, particularly around crypto, has seen advertising revenues plummet. So the business logic is coldly clear. Why pay a human when a large language model can churn out market updates and token explanations at near-zero marginal cost?

This is the trust question made manifest in business strategy. If readers cannot trust that the news they're reading about volatile assets is produced by a human with accountability and expertise, what is the foundation of that information ecosystem? It turns the entire venture into a potentially self-referential loop of automation. This isn't just about job loss; it's about the degradation of a fundamental layer of our economic discourse. When the news about what to believe in is written by a system with no belief at all, we have entered a new and very strange phase.

So, where does this leave us? We have preservationists unionizing. We have legacy broadcasters hollowing out. We have platforms paying historic settlements for societal impact. We have entertainment franchises posting GDP-level revenues. And we have the very act of business reporting being automated in the sectors that need scrutiny the most. The business of media is no longer a single story. It's a dozen simultaneous revolutions, some about preserving quality, some about cutting costs, all about deciding what we value in the information age. The market is giving us many answers this week. It's up to us to decide which ones we're willing to trust.

Science

Welcome back to ReadFlow Weekly. If our business segment was about the very human turbulence in the industries that tell our stories, then our science segment is a welcome shift in perspective. It's about the stories the universe tells about itself, the deep history written in our bones, and the patient, monumental work of figuring things out. This week, that work takes us from the very edges of our solar system to the fundamental rules of reality itself, and it suggests we might need to rethink the story we tell about our own origins.

Let's start with a pair of cosmic milestones that bookend the ambition and endurance of space science. First, closer to home, or at least closer to the sun, a Mercury-bound spacecraft has finally phoned home with its first scientific data. After a journey of eight years and roughly ten billion kilometers, a British-built X-ray telescope aboard the European-Japanese BepiColombo mission has opened its eyes and captured its first images of a distant cosmic object. This isn't just a pretty picture. Proving this instrument works in the harsh environment of space, on its way to the intense radiation and heat near Mercury, is a massive engineering triumph. It means we'll soon have a powerful new tool to study the solar wind's interaction with Mercury's surface and exosphere, teaching us more about how planets without atmospheres evolve. You can find more on that long-awaited success in our coverage of the Mercury-bound telescope's first light.

And speaking of long journeys, let's check in with an old friend who is setting records at a distance that boggles the mind. NASA's Voyager 1 probe has now reached a distance of one light-day from Earth. That means it is so profoundly far away that light itself, the fastest thing in the universe, takes a full 24 hours to travel from us to the spacecraft, or from the spacecraft back to us. After more than 49 years in space, this little machine, built with 1970s technology, is now approximately 24 billion kilometers from home. It's a staggering testament to human ingenuity and a quiet, persistent reminder that we have, literally, reached for the stars. The spacecraft is still returning data from interstellar space, a region no human-made object had ever reached before it. Every bit of data it sends back, over that immense and growing gulf, rewrites our understanding of what lies between the stars.

Now, from the vastness of space to the intricate puzzle of our own deep past. New research from Monash University is suggesting that the human family tree, that classic branching diagram we all learned in school, may be due for a radical rewrite. The traditional view classifies many of our ancient ancestors, like various australopithecines, as separate side branches that eventually died out. But this new work challenges that, proposing that several groups long considered distinct may actually belong within our own genus, Homo. This isn't just academic reshuffling. If true, it means our lineage is older, more diverse, and more complicated than we thought. It blurs the line between "them" and "us," suggesting that the traits we define as uniquely human emerged in a much wider, more varied group of hominins. It complicates the neat, linear story of human evolution into something far more interesting: a tangled, bushy network where intelligence and adaptability weren't a single winning ticket, but a experiment run by many branches of the family. For a deeper look at this potential paradigm shift, we've got the story on the proposed rewrite of the human family tree.

From the history of matter to the behavior of the most fundamental particles. In Switzerland, physicists at CERN have announced a detection decades in the making. They have finally found evidence of an extraordinarily rare particle decay, one so uncommon that it happens only about once in every ten billion beauty quarks. Scientists have been searching for this specific decay for over forty years. Why the long hunt? Because observing it, or not observing it, provides a crucial test for the Standard Model of particle physics, our best rulebook for how all fundamental particles and forces interact. Confirming the predicted rate of this decay helps shore up our understanding. Finding it happening faster or slower than expected would be a crack in the foundation, a sign of new physics waiting to be discovered. This result is a victory for patience and precision, a reminder that some of the biggest questions are answered not in a flash, but in the meticulous accumulation of data over a human lifetime.

And finally, a story that feels like science fiction becoming science fact, with a distinctly corporate twist. Google has signed a deal with Jeff Bezos's Blue Origin to send satellite-based data centers into orbit. This is part of Google's experimental Project Suncatcher, and the idea is as audacious as it sounds: launching satellites equipped with the computing capacity of a small data center into space. The proposed benefits include leveraging the cold of space for cooling and positioning computing power directly where it's needed for Earth observation and communications. It's a bold, expensive bet on the infrastructure of the future. It also opens a new chapter in the conversation about the orbital environment, energy use, and who builds the next layer of our digital world. It's a long way from a proven technology, but it signals that the next computing frontier isn't just terrestrial or in the cloud, it might literally be in orbit.

So this week in science: we're redefining our family, verifying foundational physics, checking in with our farthest robotic emissaries, and contemplating a future where data centers orbit overhead. It's a mix of profound introspection and wild extrapolation, which is, when you think about it, what science does best. That's our look at the world of discovery. Stay with us.

The thread

Welcome back to ReadFlow Weekly. If our science segment was about the patient, monumental work of figuring things out, then this, our final thread, is about the payoff. It's about the pattern that emerges when you step back from the week's individual stories. And this week, that pattern is strikingly clear. We've been watching a global stress test on the very idea of verification. What can we prove, what can we trust, and what happens when that trust evaporates.

Let's start with the purest form of verification imaginable. For 165 years, the Riemann Hypothesis has stood as perhaps the most important unproven conjecture in mathematics. This week, an Anthropic mathematician claimed a proof. But in the world of modern mathematics, a claim is just the beginning. The real news came from the Lean and Mathlib communities, who released a joint statement. They've verified that proof. Not with a handshake and a peer-reviewed paper, but with 17 million lines of code in a formal proof assistant. This is verification taken to an almost philosophical extreme. It's a mathematical truth rendered as a program that compiles. It's a level of certainty that other fields can only dream of. You can read more about that landmark Lean verification on ReadFlow.

Now, let's swing to the opposite end of the spectrum. In our business segment, we saw trust and verification breaking down in real time. The media industry is in a period of intense volatility, with unionization drives at places like the Internet Archive, staff cuts at the BBC and CBS News, and a potential multi-billion dollar settlement hanging over YouTube. But the most telling story might be the quiet replacement. Outlets like Decrypt and The Defiant appear to have swapped human reporters for AI-generated content. This isn't just about job loss. It's about the dissolution of a fundamental verification layer. Who, or what, is verifying the information in those stories? The answer, increasingly, seems to be no one we know. The human editorial process, flawed as it is, represents a chain of verification and accountability. When that's removed, what's left is a black box. You can find our coverage of that shift towards AI reporters in crypto media on the site.

And this brings us full circle to technology, where verification is both a technical necessity and a fragile commodity. Bitcoin's value plunged to a two-year low this week, continuing a brutal slide. Crypto's entire value proposition is built on a cryptographic form of verification, the blockchain. But that's a verification of transactions, not of intrinsic worth. The trust in that worth is evaporating, hence the crash. Meanwhile, security advisories were issued for critical software like GPGME, and experts called for "canary tokens" in AI supply chains. These are all tools for verification. They're tripwires and alarms designed to tell us when something we trusted has been compromised.

So here's the thread. From the cosmic patience of a Mercury-bound telescope delivering its first images after eight years, to the forty-year search for an ultra-rare particle decay at CERN, we saw the glory of slow, rigorous verification. And from the crumbling trust in crypto assets and the automated churn of AI news feeds, we saw the chaos that ensues when verification is absent, or ignored. The week asks us a direct question. In a world accelerating towards automation and abstraction, what are we willing to slow down for? What truths are worth the 17 million lines of code, the forty-year experiment, the eight-year journey? Because everything else, from the value of your digital currency to the truth of your news, hinges on the answers we choose.

One more thing

And finally, as we wrap up a week about grand verifications and systemic breakdowns, a reminder that sometimes the gremlin in the machine is not a cosmic force, but a very specific, very expensive piece of hardware on your desk. This week brought us the delightful and baffling tale of a tech enthusiast who, after bragging about his 2.5 gigabit internet connection, discovered his download speeds were being mysteriously throttled by 75 percent.

The culprit, after a classic process of elimination? His high-end Apple Studio Display. As detailed in the troubleshooting saga, the monitor's built-in Ethernet port was somehow crippling the connection. It's a wonderfully weird case of a premium feature becoming a silent bottleneck, a ghost in the elegant machine.

It serves as a perfect, humble coda to our themes. After a week of verifying century-old math proofs and watching billion-dollar systems wobble, here's a problem you can actually solve by unplugging a cable. It's a hands-on lesson in trust, but for your local network. And it proves that no matter how advanced our tools get, the most satisfying fix often remains refreshingly analog. For ReadFlow Weekly, I'm [Host Name]. Thanks for listening, and we'll see you next week.

Sign-off

And that brings us to the end of this edition of ReadFlow Weekly. From the patient, monumental verification of a 165-year-old math conjecture to the volatile trust in everything from crypto to the news we read, it's been a week that reminds us to look closely at the foundations of what we believe. Thank you so much for spending part of your Saturday with us.

If a story piqued your curiosity, like the monumental claim of a proof for the Riemann Hypothesis or the quiet, historic milestone of Voyager 1 reaching a full light-day from Earth, you can dive deeper into all of these articles and save them for later right on ReadFlow.

We're here every Saturday to help you make sense of the week that was. So until next time, keep reading, keep saving, and we'll see you right back here next week.

Key takeaways

Bitcoin's brutal slide continues

The cryptocurrency plunged to a two-year low, shedding more than half its value since its peak and testing market trust.

A 165-year math puzzle gets verified

A claimed proof of the Riemann Hypothesis was backed by a massive, independent Lean community verification, a landmark for mathematical truth.

Media faces a turbulent reckoning

From unionization drives and staff cuts to AI replacing reporters, the industry's foundations are shifting in real time.

Cosmic milestones at the solar frontier

Voyager 1 reached a distance of one light-day, while a Mercury-bound telescope delivered its first images after an eight-year journey.

Our human family tree needs a rewrite

New research suggests several ancient groups, long considered separate, may actually belong within our own evolutionary lineage.

A monitor secretly throttled internet speeds

A high-end display's built-in Ethernet port was found to be silently cutting download speeds by 75 percent, a bizarre hardware gremlin.