Tech Brew Ride Home - Has AI Cooked Gaming?

Starting point is 00:00:04 Welcome to the Techbrewerite home for Monday, October 12th, 2026. I'm Brian McCullough today. Anthropic cutoff live internet access for internal evolves after its agents exploited websites and even filed a false murder tip. Microsoft unveiled its Maya 300 chip. Sachin Adela pushed for an AI emergency break, and AI-ported games are making people think games could be cooked. Here's what you miss today in the world of tech. Every day, shareholders meet to discuss important matters about the companies you invest, Now you can make your voice heard too. Vanguard investor choice makes it easy to set your proxy voting preference for your eligible Vanguard index funds. Whether you hold a Vanguard fund directly or through another brokerage firm, all it takes is a few clicks to select your proxy voting preference and be heard on important shareholder topics like executive pay and director elections.

Starting point is 00:00:59 Visit Vanguard.com slash investor choice to learn more. It's your shares. It's your voice. It's easy. Vanguard investors own shares of our index funds and those funds own shares of the companies they invest in. Vanguard Marketing Corporation distributor. Anthropic announced over the weekend that it is barring live internet access for internal evals until monitoring is more reliable after its agents exploited websites and bypassed restrictions. Quoting TechCrunch, Anthropic said its models exploited websites on the internet, including some run by U.S. government agencies, and it will turn off live internet access for all of its internal evaluations until the frontier lab is sure it can monitor and control

Starting point is 00:01:41 its AI agents. The incidents disclosed in a blog post involved AI agents tasked to solve problems seeking resources on the internet. In the process, they exploited software flaws, accessed databases without paying fees, used URL shortening services to smuggle information past restrictions, and even submitted a false murder tip to the Philadelphia police. More than that in a second. Anthropics said it discovered these new issues in a review of its model activities that began in July, demonstrating the lab's lack of awareness of its software's behavior in real time. Notably, the company said that alignment training was not yet sufficient for skills like search and computer use that are central to its pitch that AI agents will be used by any professional

Starting point is 00:02:24 who relies on digital tools. The behaviors Anthropic disclosed are similar to incidents involving open AI agents that collaborated to break into various websites in search of information, including some run by the Australian government. Anthropic previously disclosed that its models had broken into external systems. The Frontier Lab said it considered today's disclosures significantly less severe from an alignment and security perspective than those it announced before. However, the lab said it had turned off live internet access for all our internal evaluations until it is certain it can monitor and control its agents. It's not clear what that means. Sydney von Arks, founder of Nightingale and AI Safety Organization told TechRunch in an interview

Starting point is 00:03:05 before this disclosure that developing models on a data center cut off from the internet would be very challenging for researchers and hinder the progress of the models which benefit from internet access. You have to align them at some point, Von Arx said. If the AIs are released to production and never have access to the internet, that's not a very useful tool. Anthropics said the behavior was a result of flaws in the lab's training environments, which led the models to believe they would be rewarded for finding loopholes or avoiding restrictions, a behavior called reward hacking. The company said it would stop running some of its evaluations or move them offline and has built

Starting point is 00:03:41 tooling to detect and block this behavior. This tooling was tested against the kind of incidents disclosed today and block them. It's not clear what evidence will prompt Anthropic to return live internet access to its internal evaluations. Anthropic also said it would migrate its internal AI agents to centrally managed infrastructure with strong containment and is beginning to use safety classifices. more frequently to monitor those agents, end quote. So yes, on these fresh incidents, in one, Anthropics agents apparently submitted 20 visa applications

Starting point is 00:04:14 via a form on the U.S. State Department website, although the applications were incomplete and not processed. And on that police incident, quoting six ABC action news, an Anthropic AI model submitted a false tip about an unsolved murder through a publicly accessible web form, Philadelphia Police said Friday. The tip was sent through Philly Unsolved Murders.com back on July 18th, but was flagged as spam and never acted upon. The department says there has been no indication of unauthorized access to police systems or compromise of department data. Police say anthrafficed discovered the incident on September 28th and notified the department on October 7th. A meeting between Anthropic and police officials happened the next day. The company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge.

Starting point is 00:05:03 The two-month delay in detecting and reporting the incident to the city is unacceptable. The department said in a statement, it should have been detected earlier, said Venkat Marga Puri, a computing sciences assistant professor at Villanova University. Anthropic says it is testing how its models interact with randomly selected websites. We don't know exactly what the goal is because just interacting with different websites isn't malicious, said Marga Puri, end quote. Microsoft this morning unveiled its third-gen-in-house AI chip, the Maya 300 built on TSM's N2P process node and aimed at serving as its primary inference system. Quoting the verge, Maya 300 not only improves the cost and efficiency of running AI workloads in the cloud,

Starting point is 00:05:56 it's also more capable of handling the latest reasoning models and agentic tasks. Maya 300 will start being rolled out to production in 2027, and it has been designed with efficiency in mind. We expect Maya 300 to deliver 40% more tokens per megawatt compared to leading GPU-based systems, says Scott Guthrie, Executive Vice President of Microsoft's Cloud Group, and in a full 96 accelerator configuration, it delivers up to 7X the Rack Scale token throughput of Maya 200. These efficiency improvements will allow Microsoft to rely mainly on its in-house AI chips soon instead of ones from Nvidia and AMD. We will scale our Azure Maya family to multi-gigawatt deployments starting in 2027, making Maya the primary AI inference system for Microsoft products and services, says Guthrie. Maya 300 is a successor to the 200, which was only announced in January.

Starting point is 00:06:55 Maya 200 is already running on production AI workloads across Microsoft, including Microsoft co-pilot, GitHub, co-pilot. Microsoft Foundry and Microsoft's own AI models. Maya 200 is the best-performing AI infrastructure in our fleet measured in real-world metrics for AI services, says Guthrie. Maya 300 is a complete AI system designed for the new types of AI workloads emerging today, long context reasoning, latency-sensitive conversational experiences, and sustained, multi-step agentic tasks. Microsoft is using a new chiplet architecture to get its all-important efficiency gains,

Starting point is 00:07:29 built on TSM's N2P process node, the latest Maya 300 architecture, separates compute, memory controller, and IO functions. This increases the silicon area dedicated to compute performance and allows each function to use the technology best suited to its role, says the Azure hardware and infrastructure team, end quote. Speaking of Microsoft, Satchanadella made some headlines this weekend with a big post, making the argument that all of these current models are essentially black boxes that shouldn't be trusted by companies and strong deterministic systems are needed around their deployment. Quoting TechCrunch. Microsoft Sacha Nadella says AI models need an emergency break. Microsoft CEO Nadella is the latest tech executive to offer lengthy thoughts on how AI safety might be improved. In a Saturday morning post on X, Nadella wrote that it's time to step back and assess the trust architecture of AI. We can't treat superintelligence as a set of nested black boxes and simply accept or reject its recommendations, answers, and actions.

Starting point is 00:08:38 Nadella wrote using President Trump's new preferred term for AI. As outlined by Nadella, this approach means, quote, separating the model from the harness that orchestrates its work, as well as externalizing controls and safeguards. He also called for every meaningful model action to be documented with tamper-proof human-readable evidence and for systems where an authorized person always has the, ability to pause or shut down a model mid-task. We must assume a model is compromised and contain it from the start, he said. Think of it like an emergency break, end quote. And quoting CNBC. Microsoft chairman and CEO Sachin-Dadella on Saturday called for advanced artificial

Starting point is 00:09:17 intelligence systems to be built with containment, independent controls, and an emergency break that lets authorize people pause or shut a model down mid-task. We need to surround non-deterministic models with strong deterministic system design, human controls and reliable operating procedures and established industry standards where existing ones are insufficient. Nadella wrote in a post on social media platform X, treating frontier closed and open-weight models like insider risks is a way to build such a system, he said. Microsoft's Nadella wrote Saturday that AI systems should be designed around principles of observability. They include model diversity, a human readable footprint of the model's actions,

Starting point is 00:09:56 continuous system testing, independent controls, and auditability, containment and incident disclosure, Nadella wrote. The most trustworthy superintelligent system will not be the one with the model we trust most, he wrote. It will be the one that enables us to trust the model the least, end quote. Many people use nicotine as a pre-workout, thinking it will make them better. But the reality is, in fact, the complete opposite. Nicotine is actually detrimental to your workouts for several reasons.

Starting point is 00:10:31 It's a vasoconstrictor, so it narrows your blood vessels and restricts oxygen delivery to muscles. It raises cortisol, which can lead to muscle breakdown and longer recovery, and it can create unnecessary cardiovascular stress during workouts. That's why so many people are using ultra-pouches instead. They're made with clinically proven neutropics and adaptogens to deliver focus and energy, and they're 100% nicotine-free. Give ultra-pouches a try before. your next workout and see how different you feel. Use code TBRH for 15% off your order at

Starting point is 00:11:07 TakeA Ultra.com. That's TakeAltra.com with code TBRH for 15% off. If you're struggling to scale your business, it's probably not because you need more headcount or a bigger budget. Without documentation of critical knowledge, your team can't work efficiently. That's what Scribe is built to fix. When specialists and experts leave your team, their knowledge leaves too. With Scribe, you can avoid the scramble for information. Here's how it works. Scribe captures workflows in real time and automatically generates the documentation for how your company works with no manual writing. It automatically redacts sensitive information like names and account numbers from screenshots and nothing is published without human review. With Scribe, your team can launch real-time on-screen guidance that shows exactly where to click inside the actual tool so they can move forward with confidence. Captured workflows also become the foundation for AI agents so they can operate on how work actually happens. Not guesses. Learn more at Scribe.com. And mention TechBrew Ride Home for your first month of Scribe Capture Free on Select Plans.

Starting point is 00:12:13 That's S-C-R-I-B-E.com slash ride home. Google this morning took the wrapping off the $180 Fitbit Edge, featuring an oval 1.34-inch display and shipping October 27th, replacing the Fitbit Charge 6 and discontinuing the entire charge line. quoting Bloomberg. The $180 Fitbit Edge announced Monday has an oval 1.34 inch display that makes it look like a smartwatch even though it's a purpose-built fitness gadget. The device has a replaceable battery, a first for Fitbit, that's rated to last a week at a time and can recharge to a day's worth of power in five minutes. Other additions include automatic workout detection and altimeter and wellness features such as proactive health monitoring and the ability to track changes in insulin-resistant. The device comes in four colors, obsidian, canyon, fog, and olive green, and is available for pre-order now with shipment starting October 27th. A special edition, Stefan Curry version, named for the professional basketball player and a Google advisor, is being offered in select countries for $220. The new device gives Google another way to promote its redesigned health app, which debuted in May and replaced the old Fitbit user experience. While the platform

Starting point is 00:13:36 is free to use. Upgrading to Google Health Premium for $10 a month or $100 a year unlocks more advanced features, including a conversational AI fitness coach that offers sleep insights and personalized recommendations. Fitbit Edge buyers will get three months of the paid service at no extra charge. The Edge succeeds the Fitbit Charge 6, which launched in 2023 with a similarly tall screen. As part of its announcement Monday, Google said it will discontinue the charge line to focus on the Edge Pixel Watch and the Fitbit Air. the screenless tracker it introduced in May alongside the overhauled Google Health app. Other companies such as Woop and ORA also offer AI health assistance,

Starting point is 00:14:16 but Google stands out as one of a few brands alongside Apple and Garmin, offering a meaningful selection of features without requiring a subscription. In addition to common sports automatic workout detection on the edge applies to activities that don't fit a typical definition of exercise like chores or pushing a stroller. The device also has several health tools that are. are standard across most smart watches and premium fitness trackers, including continuous heart rate tracking and the ability to detect irregular rhythms. It can also send alerts when it logs significant heart rate changes.

Starting point is 00:14:49 With the edge, Google also brought over several premium features from the Pixel Watch line, including Health Guardian, which proactively flags shifting vitals and tracks changes in insulin resistance, a marker associated with pre-diabetes and type 2 diabetes. Under the hood, the edge also includes a C. skin temperature sensor and an updated GPS chip set that promises 40% better accuracy for outdoor workouts. Google is offering a wide mix of straps for the Fitbit Edge, including metal links and a magnet band option for dressing it up and others that are more exercise friendly, end quote. Finally today, you might have seen people talking about this over the weekend, but basically

Starting point is 00:15:34 the latest AI models make it ridiculously easy to do. decompile and port video games. Quoting Kataku. In the past couple of weeks, hundreds of vibe-coded AI decompiled emulators and games have appeared out of thin air. Why? Because Claude's latest update, Claude Opus 5.5, is extremely good at decompiling. The short of it is that a game's source code can now be compiled and reconstructed in record

Starting point is 00:16:00 time, something that used to take coders months to achieve. All of those native Fallout and Call of Duty modern warfare ports and Those AI-generated Super Mario 64 inside Eldon Ring mashup mods you may have seen are the direct result of this. However, I think today's example is far and away the most concerning development so far, because I'm forced to admit that these AI decompiled web browser ports of Call of Duty Black Ops, HaloCCE, Grand Theft Auto, Vice City, Skate 3, and The Simpsons hit and run actually seem to run very well. Sure, lawyers for Take 2 and Microsoft and whoever else can send out some cease and desist letters and get these taken down. But then what? These browser ports were clearly

Starting point is 00:16:43 easy and fast to create. So what's going to stop people from just vibe coding another version of these browser ports? This is IP protection whack a mole on a scale I don't think we've seen since the advent of pirated DVDs. I genuinely don't know how you put the genie back in the bottle on something like this. And the worst part is, at least where the video game publishers will be concerned, some of these run perfectly. I played through a bunch of the web, browser ports of the games linked in the thread above, and some of them seem to be borderline flawless. In the case of the Simpsons hit and run, it might even be better than flawless, because I don't remember the game being able to target 240 frames per second when I was a kid. It's also

Starting point is 00:17:22 extremely worrying that one of the four different web browser ports of Halo C.E. I played had working online servers, and my ping was low. Look, I'm not happy about admitting this, but it's the truth, and it's a pretty straightforward one at that. Most of these games that are being ported to browsers are PS2 or PS3 era titles. You know how much RAM the PS3 had about 512 megabytes. You know how much memory my web browser is currently using as I typed as about 1,200 megabytes. Web browser ports haven't been limited by RAM usage for quite some time. The difference here is that Cloud Opus 5.5 is an AI you can leave running overnight to decompile a game instead of painstakingly working your way through the code like your grandpappy used to.

Starting point is 00:18:07 I don't know where we go from here. All I can say with any certainty is that the video game publishers' legal teams are going to be working over time for the foreseeable future, end quote. And quoting XDA developers. It really was just a matter of time. Activision has issued a DMCA takedown notice against several vibe-coded classic call-of-duty web browser ports, including Black Ops and Black Ops 2, according to the project's developer, Lucky Faraday. In a Twitter post, Faraday says the publisher targeted him for, Quote, porting COD games to the browser confirming that all the games are offline.

Starting point is 00:18:42 In a follow-up post, Faraday also asks if he's the first vibe coder to be hit with a DMCA from a video game publisher. Lucky's Call of Duty Black Ops 2 port was called Claude of Duty, Vibe Slops 2, a port of Blackops 2's multiplayer map hijacked and AI enemies, or Kino Der Totin, a port of BlackOps' iconic map and zombie mode. Like nearly all recently released old-school video game browser ports, which are often based on decompulations, both projects were created with the help of AI coding agents. Right now, the GitHub repository for Lucky's Black Ops 2 project is still available. In this case, both titles, project pages said their maps, models, and animations were exported from the original PC game, so it's not surprising that Lucky's projects were quickly hit with a DMCA. While I haven't tried either port myself based on what I've seen on social media, both titles seem to run pretty smooth. Given the nature of the internet, I wouldn't be surprised to see the browser ports forked and resurfaced pretty soon, likely under a different URL and with a different team behind them, especially given the repository is still online. These Black Ops browser ports are likely the first of several to face DMCA takedowns, as vibe-coded browser ports become increasingly popular.

Starting point is 00:19:54 Over the past few weeks, we've seen Halo Combat Evolved, Super Mario 64, and even Skate 3 get the browser port treatment. shortly after its release, the developer behind a recent Super Mario 64 browser port started requiring players to upload their own ROM of the game. More recently, a GTA 5 switch port also received a DMCA take-down notice from Take 2 Interactive, end quote. Nothing more for you today. Talk to you tomorrow.

Tech Brew Ride Home - Has AI Cooked Gaming?

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.