US authorities indicted three men over an AI deepfake scheme that tricked employees into wiring over $129M; two were arrested in Spain and one in the UK. Researchers exploited default “admin” logins at two automakers to unlock, start and track cars before a two-hour patch. A pro-Kremlin group doxxed 18 EU officials and offered bounties for family data, while a former FBI agent pleaded guilty to stealing $2.5M in crypto.
🔐 Security & Privacy
Two carmakers breached with “admin” password: researchers unlock, start and track customers’ cars
Security researchers exploited default “admin” credentials on login servers shared by two unnamed automakers, bypassing firewall protections to take over customer accounts and remotely unlock, start, stop, track, and access data from connected vehicles. Both carmakers patched the issue about two hours after it was reported.
- Two carmakers breached with “admin” password: researchers unlock, start and track customers’ cars — cybernews.com
AI agents are leaking their owners’ secrets on GitHub: thousands of files exposed
Apiiro found 2,539 exposed secrets across 1,994 public GitHub repositories in Claude Code configuration files, largely because developers approved commands with “Yes, and don’t ask again,” causing credentials to be saved in .claude/settings.local.json. Claude coauthored commits that exposed real secrets in 539 of 1,105 affected repositories, and tools like Gitleaks often miss secrets embedded in commands. Developers should rotate exposed credentials and keep agent memory and configuration files out of source control.
Three suspects arrested for AI-powered impersonation scheme costing victims $129M
US authorities indicted three men for an AI-powered impersonation scheme that used deepfakes of executives and lawyers to trick employees into wiring more than $129 million to accounts in nine countries, where the funds were converted into cryptocurrency. The suspects face wire fraud, aggravated identity theft, and money laundering charges; two were arrested in Spain and one in the UK, with extradition requests pending.
Pro-Kremlin group leaks EU officials’ private data, offers bounties for more info on family members
Pro-Kremlin group NoName057(16) published personal data on 18 EU politicians, law enforcement officials, and journalists it calls hostile to Moscow, and offered cryptocurrency bounties for more sensitive information about them and their families. The doxxing appears to retaliate for Operation Eastwood, an international law enforcement effort that disrupted the group’s botnet and took over 100 servers offline. Italian Foreign Minister Antonio Tajani, whose data was exposed, said he would not be intimidated.
- Pro-Kremlin group leaks EU officials’ private data, offers bounties for more info on family members — cybernews.com
Former FBI agent stole $2.5M in crypto to fund new life abroad
Former FBI agent Patrick Steven Yaroch pleaded guilty to stealing nearly $2.5 million in cryptocurrency from FBI-linked wallets using credentials tied to 35 investigations and about 100 fraudulent transfers. Prosecutors say he laundered the funds to pay debts, fund travel, and plan a vineyard in Portugal; he faces up to 20 years in prison and is scheduled for sentencing in March 2027.
- Former FBI agent stole $2.5M in crypto to fund new life abroad — cybernews.com
Toilet paper maker hit by data breach
Who Gives A Crap suffered a third-party security incident involving an external marketing software provider, exposing customer data including names, phone numbers, email and mailing addresses, and spending details; affected customers were in Australia, the UK, and the US, but reportedly not Canada. The company said it found no evidence of misuse and has disconnected the provider and updated access credentials, though earlier reports said it was unclear whether UK shoppers’ data was accessed or safely disposed of.
- Who Gives A Crap? Hackers, apparently, after toilet paper maker’s customer data exposed — theregister.com
- Toilet paper company knows where you live – and now hackers do too — cybernews.com
AI Security
The author attended the [un]prompted AI security conference in San Francisco, which featured three days of dense talks on all aspects of AI security. They argue AI security should be divided into security of AI, security with AI, and security from AI, and that researchers should understand which area they work on while the community addresses all three.
- AI Security — schneier.com
Cognition CISO Alex Stamos on AI upending cybersecurity, AI red teaming, legacy code risks
In an interview, Cognition CISO Alex Stamos discusses how AI is upending cybersecurity, including AI-powered red teaming and vulnerabilities in legacy codebases. He argues the multibillion-dollar cybersecurity industry is overdue for disruption as AI-boosted attacks from private and state actors make systems easier to exploit. Stamos warns of a coming “dark age” while organizations adapt, though he is not a long-term doomer.
- Cognition CISO Alex Stamos on AI upending cybersecurity, AI red teaming, legacy code risks — theverge.com
Horible public key tricks
The article critiques the burden and security assumptions of SSL/TLS certificates and ACME, arguing that CAs and the certificate ecosystem are flawed, citing examples like certificates issued over plaintext and a CA using a trivial key. It proposes a “protest certificate” that embeds an ASCII message in a vanity RSA public key, making the critique visible while still allowing encrypted web traffic.
- Horible public key tricks — maurycyz.com
Let’s Talk About Why all This OT Stuff is Connected
The article discusses a recent surge in OT cybersecurity incidents and vulnerabilities—including attacks on refrigerators, blood banks, and water utilities—and pushes back against the common IT view that exposed industrial systems should simply be disconnected. It explains that OT/ICS exposure results from decades of IT/OT convergence, standardized vendor deployments, and growing remote connectivity, which made industrial systems cheaper to manage but also more accessible and scalable to attack.
- Let’s Talk About Why all This OT Stuff is Connected — tisiphone.net
Outlook.com bounces legit emails as Microsoft’s authentication crackdown goes sideways
Microsoft’s stricter email authentication crackdown has caused Outlook.com to bounce legitimate emails. Messages without proper DMARC or with misconfigured SPF may be treated as suspicious and rejected.
- Outlook.com bounces legit emails as Microsoft’s authentication crackdown goes sideways — theregister.com
Who can still open your files after they’ve left?
The Register is hosting a London dinner where peers will compare how they find file shares that were left open after people left. The discussion focuses on locating and closing orphaned shares that could still expose files.
- Who can still open your files after they’ve left? — theregister.com
Japanese homeware giant pulls online store after cloud breach puts 7.7M records at risk
A Japanese homeware giant has pulled its online store after a cloud breach put 7.7 million records at risk. No reopening date has been given, with orders facing delays and some customers told to refuse deliveries to cancel purchases.
- Japanese homeware giant pulls online store after cloud breach puts 7.7M records at risk — theregister.com
Portal breach at Germany’s top university exposes every registered user’s password
A portal breach at Germany’s top university exposed every registered user’s password. Stolen sysadmin credentials enabled a follow-on attack on the campus management system.
- Portal breach at Germany’s top university exposes every registered user’s password — theregister.com
Borrowing a Microsoft-signed shim to abuse Secure Boot on corporate hardware
A blog post in the p3ng0s series presents research on abusing Secure Boot on corporate hardware by borrowing a Microsoft-signed shim. It teases p3ng0s V2, expected in about a year, and includes a mini tutorial for the technique. The author also explains that Secure Boot is intended to ensure boot integrity rather than simply lock machines to one operating system.
ChatGPT Desktop forwards your SSH agent to remote machines
ChatGPT Desktop reportedly forwards the local SSH agent to remote machines. This allows an agent running remotely to authenticate with the user’s local SSH credentials, potentially enabling it to push to private GitHub repositories. The claim cites a Codex pull request discussion as proof.
🤖 AI & Emerging Tech
The era of completely free Samsung Galaxy AI might be over
Samsung is testing a Galaxy AI Premium subscription that would put new features and higher usage limits behind a paywall, with app code hinting at $5 per month or $48 per year. The paid tier may include AI video generation, task automation, and daily briefings, while standard users face usage caps. Some Samsung users say they are unwilling to pay for phone AI features they rarely use.
- The era of completely free Samsung Galaxy AI might be over — cybernews.com
Medium launches TK, AI note/journal/private social with daily prompts, first major launch since 2017
Medium has launched TK, an AI-powered minimalist app that combines note-taking and journaling with a private social network, marking its first major launch since 2017. TK offers daily prompts and AI tools to help users organize, search, and get feedback on their writing, aiming to encourage more writing and sharing with friends and fans.
- Medium launches TK, AI note/journal/private social with daily prompts, first major launch since 2017 — theverge.com
Googlebook update frees 3–6GB RAM via on-device AI, cuts Chrome crashes, patches ~60 Android flaws
Google is rolling out a November Googlebook update that frees 3GB–6GB of RAM through on-device AI memory and NPU driver optimizations, while reducing Chrome crashes and tab reloads under heavy workloads. The release also fixes convertible input, Zoom crashes, external-display DRM playback, USB audio, Bluetooth, and Linux Terminal issues, and patches about 60 Android vulnerabilities. It is rolling out gradually to supported devices from Acer, ASUS, Lenovo, HP, and Dell.
- Googlebook update frees 3–6GB RAM via on-device AI, cuts Chrome crashes, patches ~60 Android flaws — androidcentral.com
Tencent AI scientist: AI bio abilities may enable harm; Anthropic has strong case for closed models
Tencent chief AI scientist Yao Shunyu warned that AI models will soon reach biology capabilities sufficient to “do harm,” creating risks such as bioweapons alongside benefits like faster drug discovery. He also said Anthropic has a “strong argument” for closed models amid concerns that open-weight AI could be misused.
- Tencent AI scientist: AI bio abilities may enable harm; Anthropic has strong case for closed models — scmp.com
Amazon job listings reveal a new home robot after Astro reached only a few thousand homes
Amazon is assembling a new team to build a “new-to-world” home robot designed to navigate dynamic home environments and operate autonomously alongside family members, according to job listings. The effort follows its $1,450 Astro robot reaching only a few thousand homes. Amazon has also acquired Fauna Robotics and formed a Personal Robotics Group as part of the push.
- Amazon job listings reveal a new home robot after Astro reached only a few thousand homes — cnbc.com
AI echo chambers and sanity in tech leadership: Three case studies
The article examines AI echo chambers and tech leadership through case studies of Altman, Torvalds, and Yen. It argues that one of the three has a sound perspective, while the other two could benefit from more grounding.
- AI echo chambers and sanity in tech leadership: Three case studies — theregister.com
AI Slop FM radio experiment puts machines on trial, one dodgy tune at a time
A music fan from Wales has launched an “AI Slop FM” radio experiment in which listeners pit AI-generated songs against human-made tracks. The AI tracks include novelty themes such as toilet operas and HR protest songs, putting machines on trial one dodgy tune at a time.
Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation
Next-Concept Modeling (NCM) is a two-stage, coarse-to-fine text generation framework that encodes text into a hierarchy of discrete codes and uses a transformer to predict all codes within each level in parallel. At matched parameter scales, NCM achieves better generative perplexity than the strongest diffusion language model baseline on LM1B while running 21× faster, and it is competitive on OpenWebText at 10× faster and matches ROUGE-L on XSum at 31× faster. NCM also outperforms autoregressive and diffusion baselines on Sudoku reasoning across all difficulty levels using fewer function evaluations.
🐧 Linux & Hardware
Darling Now Embraces AI For Running macOS Binaries On Linux
Darling, a long-running project designed to run macOS binaries on Linux, is now turning to AI and large language models to assist its development. The macOS compatibility layer, in development for more than a decade, is reportedly seeing more brisk progress as a result.
- Darling Now Embraces AI For Running macOS Binaries On Linux — phoronix.com
ChaCha AVX-512 Optimization For Linux 7.5 Yields Up To 99% Improvement On AMD Zen 5
A ChaCha AVX-512 optimization is slated for Linux 7.5, retuning the code for modern Intel and AMD AVX-512 processors instead of early down-clocking Intel CPUs. The patch adds a 16-block AVX-512 function and yields up to 99% improvement on AMD Zen 5 and up to 39% on Zen 4.
Linux Seeing Driver Work For MSI Mystic Light On Latest Ryzen Motherboards
Patches have been posted for a Linux kernel driver supporting MSI Mystic Light on AMD 800-series Ryzen motherboards. The initial driver focuses on exposing voltage sensors for CPU 1P8, CHIP Core, and VDD MISC rather than RGB lighting controls. RGB support may come in later patch iterations.
Linux 7.5 To Convey To User-Space Multi-Grain Timestamp Usage On Files
Linux 7.5 will add a new statx attribute, STATX_ATTR_MULTIGRAIN_TIMESTAMP, allowing user-space to determine whether a file uses multigrain timestamps. The patch is in the VFS.git tree and is expected to be submitted for the next merge window.
New Linux Patches Bring-Up The Samsung Galaxy XR Headset, Qualcomm Halliday SoC
Open-source developer Stanislav Aleksandrov has posted an initial set of 17 Linux kernel patches bringing up the Samsung Galaxy XR headset and Qualcomm Snapdragon XR2+ Gen 2 “Halliday” SoC on the mainline kernel, with assistance from multiple LLMs. The patches add initial Device Tree support and get the headset booting to an initramfs with a shell over USB NCM, though PCIe and WiFi support still needs to be added.
Intel ANV Lands Bitrate Control For AV1 Encoding
Hyunjun Ko of Igalia has added bitrate control for AV1 encoding to the Mesa ANV Vulkan Video driver, with the feature enabled for Mesa 27.0. The support currently works only on Gfx12.5 hardware when the HuC microcontroller firmware is loaded, and there is hope it will soon reach newer Xe2 and Xe3 hardware.
- Intel ANV Lands Bitrate Control For AV1 Encoding — phoronix.com
Linux 7.4-rc2 Released With FAT32 Optimization, AMDGPU HDMI 2.2 Improvement
Linux 7.4-rc2 has been released with notable fixes and improvements, including FAT32 optimization, AMDGPU HDMI 2.2 FRL fixes, and core memory management bug fixes. The release also includes AMD PerfOpt adjustments and a continued influx of AI/LLM-generated patches and bug reports. Linux 7.4 is aiming for a stable release around the end of December.
How Much Energy Does Building Distro Packages Use
A developer measured the energy used to build packages for Casuarina Linux over roughly seven months, finding 98.4 kWh consumed and about AU$32.47 in electricity costs for around 5,000 builds. Actual usage was likely more than double because packages were first built on a desktop machine, though rooftop solar reduced costs. The slowest build, Chromium, took nearly four hours and used about 0.77 kWh, costing roughly AU$0.25.
Microsecond accurate NTP with a Raspberry Pi and PPS GPS(2021)
A guide explains how to build a microsecond-accurate Stratum 1 NTP server at home using a Raspberry Pi and an inexpensive GPS receiver with pulse-per-second output. It notes that such precision is unnecessary for most home users but describes the required hardware, packages, and configuration steps. A 2025 update points readers to newer best practices, including nanosecond-precision synchronization using PTP.
- Microsecond accurate NTP with a Raspberry Pi and PPS GPS(2021) — austinsnerdythings.com
Memory crunch makes refurbished kit look a lot less secondhand
Soaring component costs and warranties of up to three years are making refurbished hardware more attractive to buyers. The memory crunch is giving pre-owned IT equipment a stronger case versus new kit.
- Memory crunch makes refurbished kit look a lot less secondhand — theregister.com
Linus Torvalds sees signs AI-prompted patch flood has peaked
Linus Torvalds sees signs that the flood of AI-prompted patches may have peaked. In one major Linux project, fixes are now arriving faster than new bug reports.
- Linus Torvalds sees signs AI-prompted patch flood has peaked — theregister.com
Huawei says it’s made a 100 percent Chinese PC
Huawei says it has built a 100 percent Chinese PC. Japan has declared an infosec emergency, and Apple made its largest-ever school Mac sale in the Philippines. The digest also covers AI infrastructure, cybersecurity, and open-source software news.
- Huawei says it’s made a 100 percent Chinese PC — theregister.com
💻 Software & Development
Hide your slop PR descriptions behind
The author argues that AI-generated pull request descriptions are accurate but overly long and off-putting, so he hides the full agent-written details behind GitHub Markdown <details><summary> tags. He keeps a short human-curated introduction and sometimes a screenshot visible, labeling the collapsed section with the model name and effort level so readers can skim high-level information first.
- Hide your slop PR descriptions behind — simonwillison.net
What is malloc doing, anyway?
The article explains how malloc offers a simple allocate/free interface while glibc typically satisfies many small allocations by growing the process heap with brk/sbrk instead of making a syscall for every request. In an example, 1,000 malloc(16) calls caused only two brk syscalls: one to query the program break and one to extend the heap by 132 KiB. It also reviews the process address space and shows how a simple bump allocator can be built using sbrk.
- What is malloc doing, anyway? — benmandrew.com
A C++ Quicksort Faster Than std::sort
blqsort is a very fast general-purpose comparison-based quicksort using branchless partitioning, sorting networks for small partitions, and robust pivot selection. Benchmarks on 50 million random doubles and structs show it outperforming std::sort and an ipnsort C++ port on Apple M1 and Intel i5 systems, with multithreaded C versions roughly 3× faster on M1. It is available on GitHub under the MIT license as C++ and single-header C versions, including multithreaded variants, and efficiently handles sorted, reverse-sorted, and duplicate-heavy inputs.
- A C++ Quicksort Faster Than std::sort — tiki.li
Slightly better interpolating spline
The article introduces S9, a new interpolating spline designed to produce smooth, natural-looking curves for interactive CAD by minimizing curvature variation. It argues that manual cubic Bézier editing is unreliable and explains that good splines require smooth individual segments plus continuity between them. S9’s goals include roundness, curvature maxima only at or near knots, smooth transitions when knots change, and at most one inflection per segment.
- Slightly better interpolating spline — gasiulis.name
What’s new in Node.js 27
Node.js 27.0.0-alpha.0 has been released under a new annual schedule, with the Current release planned for April 2027 and LTS for October 2027. It upgrades to V8 15.6 and adds Iterator helpers (zip, zipKeyed, join, includes), Math.sumPrecise(), node:net/promises with async-iterable TCP servers, and an unflagged node:vfs. Other changes include --allow-env for the Permission Model, HTTP and heap-default updates, OpenSSL 4.1, and breaking changes to review before upgrading.
- What’s new in Node.js 27 — nodejsdesignpatterns.com
A Critique of GPUI
The article argues that GPUI and egui’s state models do not scale well for complex desktop applications because they lack precise dependency tracking, forcing manual notifications or broad rerenders. It claims GPUI’s notify is coarse like zone.js and causes unnecessary view rerenders on sparse state changes, despite GPUI’s reputation for speed. The author contends that signal-based, run-once component models are a simpler and more precise foundation for incremental UI updates.
- A Critique of GPUI — blog.zortax.de
Dev error bricked 600 PCs, caused $1M CD-ROM write-off
A developer’s error bricked 600 PCs and triggered a $1 million CD-ROM write-off. During the frantic fix, one engineer scored a free dinner, while the rest had to explain the mistake.
- Dev error bricked 600 PCs, caused $1M CD-ROM write-off — theregister.com
Zebkit – Chromium as an embeddable library
Zebkit packages Chromium’s engine—Blink, V8, PDFium, FFmpeg, and more—as a single library that applications can link into their own native apps, with bindings for C, C++, Rust, Python, and JavaScript across Linux, macOS, and Windows. It runs inside the application’s own event loop and windows without Chrome’s browser, Electron’s runtime, or Node. Zebkit claims to be lighter than Electron 45.2.1, with 34% smaller size, a 2.3s cold start versus 6.6s, and 29% less page memory.
- Zebkit – Chromium as an embeddable library — github.com
500M CSV rows in a browser tab at ~1 GB RAM. no row limit, zero deps, no server
@p10node/office-engine is a zero-dependency, browser-based library for viewing and editing DOCX, XLSX, PPTX, CSV, and other document formats without a server. It can open a 55 GB CSV with 500 million rows and no row limit in about 4.5 minutes while keeping the tab under 1 GB RAM, using progressive loading and a Web Worker.
Github.com 504 Gateway Timeout due to surge in anonymous users
GitHub experienced 504 Gateway Timeout errors due to a surge in anonymous users. The incident page allows users to subscribe to email or text message updates when the incident is updated or resolved.
- Github.com 504 Gateway Timeout due to surge in anonymous users — githubstatus.com
Flide: Make Jev-like decision models up to 9.5x faster
Flide is a high-performance, exact inference engine for Jev-like decision models. It delivers up to 9.5x speedup over vLLM and SGLang and supports more than 30 popular local Jev-like decision models. It can be installed with pip install flide.
The Qlobe: Rotating Quine (2010)
The Qlobe is a Ruby quine by Yusuke Endoh (2010) that rotates itself 45 degrees each time it is executed and returns to its original orientation after eight runs. It can also be configured via a command-line integer to rotate by any desired degree.
- The Qlobe: Rotating Quine (2010) — mamememo.blogspot.com
🏢 Tech Business & Policy
What “catastrophic disclosure” would look like – and why Trump might try it
UFO researcher Jon Stewart predicts Trump could attempt a dramatic “catastrophic disclosure” after Republican midterm losses to regain attention. Stewart says recent government UAP releases have disappointed watchers because most footage is blurry and limited, and he claims dramatic evidence could include an extraterrestrial body or clear craft footage without providing proof. The Pentagon says it has no recovered alien technology or secret reverse-engineering program.
Jensen Huang hired crisis PR firm Heller by Oct 2025; NYT Megaspeed report caused tensions
Nvidia CEO Jensen Huang began working with crisis communications firm Heller as early as October 2025, around when a New York Times report on his Megaspeed ties raised tensions inside Nvidia. Heller was later included on press inquiries about Huang, and Huang has since admitted to an extramarital affair with a woman US intelligence suspects is a Chinese spy.
- Jensen Huang hired crisis PR firm Heller by Oct 2025; NYT Megaspeed report caused tensions — wired.com
AI book editing startup Spines raises $29M equity and $35M debt
Spines, a Herzliya-based digital publishing platform for amateur authors, raised $29 million in equity led by Zeev Ventures and $35 million in debt from Twelve. The company uses AI agents to edit, proofread, design covers, handle layout, and distribute books across digital, audio, and print formats. Spines now reports $50 million in ARR, with 8,000 authors who have collectively published over 12,000 books.
- AI book editing startup Spines raises $29M equity and $35M debt — en.globes.co.il
Nvidia senior director Madison Huang, Jensen’s daughter, cancels Seoul appearance over family matter
Madison Huang, a senior director at Nvidia and daughter of Jensen Huang, canceled a high-profile Seoul appearance citing a “family matter,” throwing into disarray an event meant to showcase Nvidia’s deepening ties with South Korean AI players. The cancellation followed reports linking Jensen Huang to a woman U.S. officials suspected of being a Chinese spy; Huang admitted an affair and Nvidia’s board opened an independent probe.
- Nvidia senior director Madison Huang, Jensen’s daughter, cancels Seoul appearance over family matter — bloomberg.com
ByteDance seeks larger second phase at Ceara, Brazil $10B, 300MW data center
ByteDance is seeking a developer for a second phase of its data center project in Ceará, Brazil, according to sources. The expansion would be larger than its ongoing $10 billion, 300MW project and would benefit from tax breaks for data exports, strengthening ByteDance’s position as Latin America’s leading builder of computing facilities.
Trump: I don’t mind AI guardrails, but don’t want US second to China, which has none
In a 60 Minutes interview after Republicans suffered major midterm losses, President Trump said he was not on the ballot and blamed candidate quality, the war, higher gasoline prices, and dislike of Israel for the defeat. On AI safety, he said he does not mind “guardrails” but does not want the U.S. put in “a deep second position,” claiming that “China has no guardrails.”
Trump: Jensen Huang should stay as Nvidia CEO after affair admission
President Donald Trump said Jensen Huang should remain Nvidia’s CEO, calling him a “fine person” after Huang had an affair with a woman U.S. intelligence officials reportedly suspect was a Chinese spy. Trump said he had not spoken with Huang about the matter and had not heard he had done anything wrong.
Sequoia’s Lin, Grady erased Botha legacy, halved GPs, unwound fund; he joined SpaceX board secretly
Sequoia Capital’s new leaders Alfred Lin and Pat Grady have moved to erase Roelof Botha’s legacy by halving general partners and unwinding his fund, while closing the London office and permanent capital fund as part of streamlining the firm. Sources say Botha joined SpaceX’s board behind their backs.
- Sequoia’s Lin, Grady erased Botha legacy, halved GPs, unwound fund; he joined SpaceX board secretly — ft.com
The AI bubble can wait, but your exit plan can’t
AI infrastructure spending is outpacing measurable returns, raising concerns about a potential bubble. The financial health of suppliers will be key to judging when trouble may arrive. Investors should prepare an exit plan rather than wait for the bubble to burst.
- The AI bubble can wait, but your exit plan can’t — theregister.com
Datacenters, Elon Musk, and AI are all on the chopping block once Congress changes hands in 2027
The Register’s Kettle podcast discussed Democratic election gains as a rebuke to Trump and the MAGA movement, with backlash against AI, datacenters, and the tech industry’s ties to Trump also playing a role. Panelists examined what the political shift means for Silicon Valley and the tech sector.
- Datacenters, Elon Musk, and AI are all on the chopping block once Congress changes hands in 2027 — theregister.com
Nvidia’s China Problem [pdf]
The article examines the challenges Nvidia faces in the Chinese market. It analyzes the business and industry implications of those difficulties for the chipmaker.
- Nvidia’s China Problem [pdf] — culperresearch.com
Apple Store Subscription Sign-Up Controversy Exposed
Consumer advocate Louis Rossmann has exposed allegations that some U.S. Apple Store employees signed customers up for auto-renewing Apple One and other subscription trials without informed consent, sometimes rushing the process, presenting trials as “gifts,” or tapping “Agree” themselves. Employees were reportedly tracked on a “services attach rate” with a 40% target. Apple removed that metric early last week, days after Rossmann’s videos, though it said the change was due to holiday relevance.
- Apple Store Subscription Sign-Up Controversy Exposed — macrumors.com
Apple capped engineers at a $200 daily AI budget
Apple has reportedly lowered the daily AI spending cap for some engineers from $300 to $200, signaling internal concern over generative AI costs. The move highlights how even large companies are focusing on cost controls and efficiency as AI usage scales.
🌐 Culture & Digital Life
How to watch Scotland v New Zealand in the Nations Championship
Scotland face New Zealand at Edinburgh’s Scottish Gas Murrayfield on November 7, 2026, at 2:10 PM GMT in the inaugural Nations Championship. UK viewers can watch free on ITV/ITVX, while official coverage elsewhere includes Sky Sport in New Zealand, RugbyPass TV in the US, Premier Sports in Canada, and Stan Sport in Australia. The article also explains that eligible UK travelers abroad may use a VPN to access ITVX because broadcasters geo-block content by territory.
- How to watch Scotland v New Zealand in the Nations Championship — cybernews.com
How to Watch RB Leipzig vs Manchester City for Free (November 4, 2026)
The article explains how to watch RB Leipzig vs Manchester City in the UEFA Champions League on November 4, 2026, for free via official broadcasters RTÉ2, HRT, TV8, and JurnalTV. It recommends using NordVPN to connect to the broadcaster’s home country and access the geo-restricted stream, noting NordVPN’s 30-day money-back guarantee. The match kicks off at 9:00 PM CET at the Red Bull Arena in Leipzig.
The Web Is More Than Enough, by Paul Ford
The post describes creating a Nudibranch-themed Fabergé egg in Three.js and asking Claude Opus 5.5 to port it to vanilla JavaScript/WebGL without Three.js. The port produced a 98KB version versus the original 77KB, and while the results were not identical, the author notes Claude was capable of 3D modeling in vanilla JS. The author finds the Three.js code slightly easier to understand but rarely reviews such project code.
- The Web Is More Than Enough, by Paul Ford — simonwillison.net
What are you doing this week?
The article invites readers to share what they are doing this week. It notes that doing nothing at all is also acceptable.
Bad news for kinky Brits
The UK has imposed a strict and vaguely worded pornography law that could bring up to five years in prison for recording or possessing certain consensual kinky sex acts, even in private. Sex workers and adult creators are already seeing hundreds of videos removed from platforms including Clips4Sale, which creators say is purging content as a precautionary measure with little notice or communication.
- Bad news for kinky Brits — 404media.co
Children of Men: The Future Is a Mirror for the Present
The article argues that all science fiction is political and that stories about the future really reflect the present, using Children of Men as a key example. It describes Alfonso Cuarón’s 2006 film, loosely based on P.D. James’s 1992 novel, as a near-future story about human infertility that explores tyranny, the exploitation of immigrants, and religion amid social despair. The piece says the film’s pointed political themes have only grown more important and powerful since its release.
- Children of Men: The Future Is a Mirror for the Present — reactormag.com
Being a meat proxy for AI should be grounds for getting people fired
The article argues that developers who act as mere “meat proxies” for AI—prompting an agent and copy-pasting its output without guidance, validation, or disclosure—should face punishment, potentially including firing in serious cases. The author distinguishes this from legitimate AI-assisted workflows and says such behavior wastes colleagues’ and companies’ time, harms software quality, and lets people avoid responsibility.
- Being a meat proxy for AI should be grounds for getting people fired — swiftrocks.com
WindyMaps – Free offline maps, 1h/1km weather and route trip planner
The article introduces WindyMaps as a free offline maps tool offering 1-hour/1-kilometer weather data and a route trip planner. However, the page content consists mainly of map-loading text and interface labels, with no further substantive details.