Nov 9, 2026 · 8 min listen · Last updated November 9, 2026
From storyflo. This is your daily audio brief. It's Theo. November 9th, tech roundup — five stories, here's number one. Let's get into it. First, from The Register. Linus Torvalds sees signs AI-prompted patch flood has peaked.
Listen · storyflo · tech
Daily Tech Brief · November 9th
0:00-7:48
Pick your daily storyteller
Subscribe to match with Theo, Jessica, Chloe, Mason, Brock — your voice, every brief.
Audio pre-rendered by Storyflo · cached + delivered from the edge
Linus Torvalds sees signs AI-prompted patch flood has peaked
Linus Torvalds shared some hopeful insights in his latest kernel update, suggesting that the intense wave of AI-driven software patches might be easing up. He introduced Linux 7.4 rc2, noting it’s significant because he had to skip some code in the previous release due to its sheer size. One of the key points he highlighted was the “various AI fixes” that have emerged since Anthropic’s Mythos model launched in April, which has helped developers uncover a staggering number of software flaws.
For instance, Microsoft tackled a record 974 flaws just last month, following a series of other major fixes throughout the year. Oracle also addressed an impressive 1,449 flaws in a single month. This flood of patches has overwhelmed IT teams, who are now racing to patch vulnerabilities that AI has identified, especially as attackers are increasingly exploiting these weaknesses.
However, Torvalds mentioned a glimmer of hope, particularly in the realm of the Extended Berkeley Packet Filter (eBPF). He noted that fixes are starting to outpace new reports in that area, which could indicate a turning point. While he cautioned against getting too optimistic too soon, there’s a sense that the worst might be behind us. If this trend continues, it could mean a significant shift in how developers manage the Linux kernel’s vast codebase, which consists of over 40 million lines of code.
The tidal wave of AI bug reports may hopefully come to an end as Linux 7.4-rc2 got more fixes than reports
You know how Linux release candidates usually come out with a focus on fixing bugs? Well, it turns out that the latest one, 7.4-rc2, is showing a promising shift. Instead of the usual trend of seeing more bug reports piling up, this time, the fixes are outpacing the reports. It's like a breath of fresh air in the Linux community, where developers have been scrambling to tackle issues over the past six months.
What’s interesting is that while we’ve been used to seeing larger release candidates filled with numerous fixes, this one seems to signal a potential turning point. It suggests that the developers might finally be getting a handle on the problems that have plagued previous releases. So, while we’ve had our fair share of headaches, it looks like there’s a light at the end of the tunnel.
If this trend continues, we could be looking at a smoother experience for users down the line. It’s a reminder of how responsive the open-source community can be when they rally together to resolve issues. Let’s keep an eye on how this unfolds, but for now, it feels like a small victory in the world of Linux.
MIT engineers have developed an AI-assisted learning platform called FLEX, which allows robots to understand and interpret human commands in a more nuanced way. This platform enables robots to not only follow instructions literally but also grasp the underlying intentions behind those commands. For example, if someone tells a robot to avoid spilling coffee on a laptop, FLEX helps the robot reason that it should steer clear of all electronics, thanks to a large language model (LLM) that has absorbed vast amounts of text data.
FLEX operates through a three-step process: first, it adds relevant features to the robot's awareness based on user commands. Next, it generalizes by identifying patterns and connections among these features, allowing the robot to make educated guesses about what to avoid, like fragile items. Finally, it masks out any irrelevant features based on the robot's perception of its environment. This means the robot can adapt quickly without needing constant, detailed instructions from users.
In tests, FLEX showed impressive results in various scenarios, such as serving coffee or moving objects. It not only followed direct commands but also demonstrated an understanding of the broader context, like avoiding flammable materials or fragile items. The goal is to create robots that can seamlessly integrate into our lives, picking up on individual preferences without requiring users to specify every detail. This could make a significant difference in how we interact with household robots in the future.
A single AI model simulates the behavior of many substances, from sliding sand to folding proteins
Researchers at MIT and UBC have developed an AI model called WorldParticle that can simulate a variety of physical phenomena using a single framework. Traditionally, scientists had to use different simulators for different processes, like fluid dynamics or protein folding, which required specific coding for each scenario. WorldParticle simplifies this by breaking materials down into tiny particles and predicting their movements over time, adapting to new materials and forces without needing custom equations for each situation.
This model learns the dynamics of materials rather than relying on predefined equations, making it versatile for applications ranging from how granular materials behave when touched by soft robots to the dynamics of sloshing liquids in storage tanks. The researchers have demonstrated its effectiveness by modeling six different physical phenomena with greater accuracy and speed than conventional methods.
The team is excited about the potential of WorldParticle to not only improve simulations but also to aid in discovering new materials and understanding complex physical processes. They envision a future where this unified approach to simulation could lead to insights and discoveries that we haven't even considered yet.
Recently I was invited to speak at the Ukrainian Institute of America with two of my boys, Amory and Conrad, about our time in Ukraine with Global Empowerment Mission. We were interviewed by Katya Soldak, a journalist from Kharkiv and editorial director of Forbes Media’s international editions. The conversation lasted an hour and a half, which is quite a lot to ask of anyone on Substack. But it’s the fullest answer I’ve given to the questions people so often ask me.
The Huang Affair: What did they know and when did they know it?
Jensen Huang, CEO of Nvidia, has confirmed he had an affair with Alice Huang, who’s linked to the chip industry and suspected of being a Chinese spy. This revelation raises serious questions about potential insider information regarding U.S. export controls that Huang might have shared, especially given the interest from the Chinese government in Nvidia’s technology. The situation is complicated further by political ties; it seems former President Trump was aware of this a year ago but chose to remain close to Huang, even calling him during public appearances.
Nvidia is now under investigation, with Congress likely to dig into who knew what and when. There’s speculation that sensitive information may have been passed between Huang and Alice, and the company’s recent claims of ignorance about these issues are under scrutiny. The implications stretch beyond personal scandal; they touch on the broader landscape of AI, international relations, and tech industry accountability.
As the news unfolds, Nvidia's stock is already taking a hit, and without a clear succession plan in place, the future of the company—and potentially the AI investment market—looks uncertain. It’s a tangled web of power, politics, and technology that’s about to unravel, and we’re all just waiting to see how it plays out.
Can CrowdStrike's New 'AI Kill Switch' Shut Down Rogue AI Agents?
CrowdStrike's "Falcon Guardian" protects Windows systems by enforcing policy, restricting unauthorized applications and blocking USB devices, notes Gartner. And now CrowdStrike says they're giving security teams "the ultimate human control over AI agents" with — an "AI Kill Switch... a defender can terminate it where it runs, block it from launching again, and cut off its enterprise network access." "When you deploy agents with a kill switch, you stay in control," said George Kurtz, CEO and founder of CrowdStrike...
Asahi Linux Drops MIT and Goes Full GPL After Spotting LLM-Assisted License Laundering
Asahi Linux, the project that enables Linux to run on Apple Silicon Macs, has decided to switch all its code from the MIT license to the GNU General Public License version 2 or later. This shift stems from concerns over license exploitation, particularly as AI tools have made it easier for bad actors to lift code without proper attribution. Asahi's team noticed attempts to hijack their work and monetize it aggressively, which prompted them to adopt stronger copyleft licenses. Under the GPL, anyone distributing modified versions must also share their source code, making it harder to profit off their contributions without giving credit.
Interestingly, this move is significant because it’s the first time a major open-source project has changed its licensing due to AI-assisted code laundering. Previously, projects typically adjusted licenses to prevent cloud providers from commercializing their work. The decision has sparked some criticism, especially from the BSD community, which initially influenced Asahi's choice of the MIT license. Asahi Linux is also taking a firm stance against AI-generated code contributions, emphasizing the importance of maintaining the integrity of their work.