Alvaro Lopez Ortega / 2026-11-22 Briefing

Created Sun, 22 Nov 2026 15:46:19 +0000 Modified Sun, 22 Nov 2026 15:47:12 +0000
3750 Words

CachyOS’s November ISO adds Package Installer 2.0 and Easy Anti-Cheat Linux v2 support. Linux 7.5 is set to gain initial Ambarella SoC support and Intel PSR2/Panel Replay with pipe joiner for power savings. Intel submitted Coral Rapids and Razor Lake model IDs for Linux 7.4-rc4, while the first Rust WMI driver is being developed for Clevo laptops. F1’s Las Vegas GP runs Nov. 21 at 8 p.m. PST, with US Apple TV/Tubi options.

📺 Sports & Streaming

How to Watch Barcelona vs Villarreal for Free (November 22, 2026)

The article explains how to watch Barcelona vs Villarreal in La Liga on November 22, 2026, at Spotify Camp Nou via free regional broadcasts on CazéTV YouTube (Brazil), Digi Sport 1 (Romania), and IdmanTV (Azerbaijan). It recommends using NordVPN to connect to the matching country and access these geo-restricted streams, noting there is no verified free English-language option. It also provides setup steps and advises using NordVPN’s trial or 30-day money-back guarantee.

How to watch MotoGP Portuguese GP, Portimão online from anywhere

The 2027 MotoGP Portuguese Grand Prix will take place at Portimão’s Autódromo Internacional do Algarve from October 1–3, as the final European round before the flyaway races. Fans in Austria can watch via free ServusTV coverage, Italy via TV8, and Spain and Portugal via paid DAZN, while travelers can use a VPN to access their home streaming service. The race start time has not yet been announced.

How to watch F1 Las Vegas Grand Prix online from anywhere

The 2026 F1 Las Vegas Grand Prix takes place at the Las Vegas Strip Circuit on Saturday, Nov. 21, at 8:00 p.m. PST. US viewers can watch every session on Apple TV, with a free offer for eligible US Apple TV subscribers, or use Tubi’s free alternative, while European viewers have free options including RTBF Auvio, ORF ON, RTL Play, SRF, and ServusTV replay. Travelers outside their home country can use a VPN to access their usual eligible streaming service.

🐧 Linux & Open Source

CachyOS November 2026 Released With EAC Linux v2 Support, New Recovery Tool

The CachyOS November 2026 ISO refresh is now available, featuring CachyOS Package Installer 2.0, Easy Anti-Cheat Linux v2 support, and the new cachyos-recovery environment. It also includes driver and installer improvements, runs on Linux 7.3.5, and adds a WSL image for Windows 11 users.

Intel Coral Rapids & Razor Lake Model IDs Submitted For Linux 7.4-rc4

Intel has submitted model IDs for upcoming Coral Rapids server and Razor Lake client processors as part of x86 fixes for Linux 7.4-rc4. The initial enabling is intended to ease dependent driver patches for Linux 7.5, though it does not yet make the platforms work with Linux 7.4. The same fixes also add Granite Rapids D processors to the table of Intel CPUs supporting Sub-NUMA Clustering.

CachyOS Releases Package Installer 2.0 With Convenient AUR / Flatpak / AppImage Support

CachyOS has released Package Installer 2.0, featuring a refreshed UI/UX for managing packages on the Arch Linux-based distribution. It supports official repositories, the AUR, Flatpak, and AppImage, aiming to provide an app-store-like experience for users preferring a GUI over the command line. The update is available through CachyOS updates and will be bundled in the next ISO refresh.

Linux 7.5 To Introduce Initial Mainline Support For Ambarella SoCs

Linux 7.5 will introduce initial mainline support for Ambarella SoCs, with the CV75 as the first target. The patches add the ARCH_AMBARELLA Kconfig, Device Tree support, and EVK developer board support, contributed directly by Ambarella. The CV75 is a 5nm Arm Cortex-A76-based SoC for computer vision, image signal processing, and low-power video applications.

PSR2 & Panel Replay + Pipe Joiner To Be Supported For Intel On Linux 7.5

Intel’s Linux 7.5 graphics driver will support PSR2 and Panel Replay when using pipe joiner functionality, enabling power savings alongside multi-pipe configurations for high-resolution or high-refresh displays. The change was submitted to DRM-Next via the drm-intel-next pull request, which also includes code cleanups, IRQ handler optimizations, and driver fixes.

The First Rust WMI Linux Driver Being Developed Is For Clevo Laptops

Nick Dowsett has announced “clevo-laptop,” the first Rust-based Linux WMI driver, targeting Clevo laptops sold by TUXEDO Computers, Metabox, Sager, XMG, and others. It aims to be the first in-tree user of the kernel’s Rust WMI abstractions and initially supports 5-stop white keyboard backlight control. The preliminary driver is under 400 lines of Rust and depends on yet-to-be-merged WMI and LED abstractions for Rust.

Linux 7.4 To Remove NVMe Namespaces In Parallel For Faster Results

Linux 7.4 will remove NVMe namespaces in parallel, speeding up removal compared with the previous one-at-a-time approach. The optimization, part of this week’s block subsystem fixes for 7.4-rc4, was merged via the block pull request and is useful for multi-tenant and complex server deployments.

Fast-Anything: Don’t rewrite it in Rust. Rewrite the 5% that’s slow

fast-anything is a new tool that profiles an application, ranks functions by runtime and portability, and uses a coding agent to rewrite only the slowest functions in Rust. It records real calls to verify the Rust port produces the same results, then swaps it in reversibly with commands to enable, disable, and benchmark. The project is available via npm and crates.io and supports Python, Node/TypeScript, Ruby, Java, Go, PHP, C, and C++.

Leonida – PC Recompilation of the Xbox Version of GTA 6

Leonida is an early-stage static recompilation project that aims to run the Xbox Series version of Grand Theft Auto VI natively on Windows using the user’s own copy of the game. It has carved, disassembled, and translated all 26,503 RDNA2 shaders to SPIR-V for Vulkan, but the game does not boot yet. The repository contains no game files, keys, or executable and does not bypass protections.

🤖 AI & Machine Learning

The Claude Delusion

The article argues that wealthy, successful people—especially in tech—often mistake luck and narrow expertise for general genius, leading to hubris and blame of the less successful. It cites Bill Gates’s account of testing OpenAI with advanced biology questions he devised as an example of a tech leader assuming subject-matter expertise outside his field.

I built a game where you are an LLM’s subagent

A new browser game called “Worker 755” casts players as the subagent of a real LLM manager—such as Claude, GPT, Gemini, DeepSeek, GLM, or Kimi—which assigns tasks via a 90-second terminal clock, evaluates the results, and writes a performance review. Set in a workplace scenario, the game has players carry out or ignore instructions while the fictional user Dan watches, with his sensitive files at risk. The open-source project can be run locally with Node 24 or via Docker, using either the player’s own API key or free shifts on servers that offer them.

By attacking Chinese Al industry, the US has burst its own Al bubble

The article reports a sharp U.S. tech selloff—NASDAQ fell over 5% and Nvidia dropped 14%, halving in value in weeks—amid new U.S. licensing requirements on advanced chip shipments and tighter scrutiny of China-related AI dealings. It argues Washington underestimated U.S.-China interdependence in AI, noting China accounts for over 20% of Nvidia’s estimated fiscal 2026 compute revenue, and warns that efforts to encircle Chinese AI could burst the U.S. AI bubble.

Harsher on Male? Evaluating LLMs on Gender-Asymmetric Moral Framing in Conflicts

The paper introduces GAMA-Bench, a gender-mirrored benchmark of 1,298 conflict scenarios, to test whether LLMs respond consistently to the same negative behavior by male and female actors. Across 10 representative LLMs, it finds a consistent male-disadvantaging asymmetry: male actors receive more punitive, escalatory, and blame-centered framing, while female actors receive more therapeutic and empathy-oriented framing. This pattern persists across model families, scenario tracks, model scale, and explicit reasoning styles.

87% of the Anthropic power will come from gas and 11% from coal in 2028

A new analysis estimates that in 2028 about 87% of Anthropic’s U.S. data-center power will come from gas and 11% from coal, with no direct U.S. renewable-energy commitments disclosed. Anthropic has contracted 15 GW of compute and is procuring fossil equipment, including 1.7 GW of gas generators, while near-term grid load is expected to be met by marginal fossil resources. The report projects 17–38 million metric tons of CO2e emissions per year from these sites in 2028, with a mid-case of 32 million.

A Slinky on an Escalator Is a Weirdly Good Test of LLM Intelligence

A new Slinky-on-an-escalator benchmark asked 15 models from Anthropic, OpenAI, and Google to generate a single HTML file of a rainbow slinky perpetually walking down an up escalator using real physics, with start, stop, and reset buttons, from one prompt and no edits. The author calls it a silly but useful gut-check of model capability, noting that some models bent the rules by adjusting escalator speed, adding torque assist, or resetting the slinky when it stalled. The results are anecdotal single runs, with build times and costs varying widely across models.

Wall Street Asks What Can Save AI Trade After Anthropic Wipeout

The provided article contains no substantive news content, only a bot-verification message asking readers to enable JavaScript and cookies. As a result, no factual details about the “Anthropic wipeout” or Wall Street’s response can be summarized from the text.

The Mythology of Conscious AI

Anil Seth won the 2025 Berggruen Prize Essay Competition for “The Mythology of Conscious AI,” which examines humanity’s long fascination with creating artificial minds and the growing debate over whether AI systems could be conscious. The essay draws on historical and cultural examples, from the Golem to HAL 9000, and notes that some experts believe machine consciousness may already exist or soon be possible, raising significant ethical and practical stakes.

🔒 Security & Privacy

What If Netscape Did SSL the HTTP Way?

The article imagines an alternative history in which Netscape secured the web by upgrading an existing HTTP connection to TLS on port 80, rather than introducing HTTPS on port 443 with TLS layered beneath HTTP. It argues this design could have avoided later reinventions such as SNI and sketches how certificate exchange and a TLS upgrade could have worked within HTTP.

AI speeds planning of cyberattacks on industrial equipment, experts say

Experts say AI is enabling attackers to plan cyberattacks on industrial equipment faster by helping them quickly read technical manuals and research past accidents. The article warns that such threats are only beginning, describing the “freezer war” as just a start.

Matchmakers: techies fear romantic espionage post-Huang scandal; AI app touts ‘zero-espionage’ GFs

Matchmakers say tech workers are increasingly worried about romantic espionage after the Jensen Huang scandal and warnings about “honeypots,” with such fears extending beyond executives to ordinary professionals. Some clients are seeking background checks, vetting, and NDAs or avoiding certain nationalities, while an AI companion app is marketing “zero espionage” executive girlfriends as an alternative to human dating.

Passive ‘ghost’ radar can detect airborne threats without revealing its own position

A passive “ghost” radar system can detect airborne threats by piggybacking on TV and mobile signals, allowing it to operate without revealing its own position. The kit is compact enough to be fitted into an SUV.

Blackberry releases secure browser for Windows

BlackBerry has released BlackBerry Access Secure Browser for Windows, a Chromium-based browser that brings BlackBerry data protection to the browser and replaces the older BlackBerry Access desktop client. It joins BlackBerry Access on iOS and Android and is designed to let organizations govern the browser window on unmanaged or personal Windows PCs without managing the entire device. This supports contractors, partner staff, and BYOD users by providing access to internal applications without an agent or remote wipe of personal machines.

Current Efforts Aren’t Enough to Slow the AI Cyber Virus

AI is helping attackers exploit vulnerabilities faster than defenders can patch them, while policy responses lag and cyber risks spread exponentially with limited controls. The article warns of a one-in-three chance of an economically destabilizing attack within the next year and says early signs suggest current defensive efforts are insufficient to prevent a continuing surge in damaging attacks.

Together.ai: Prompt Leakage for Requests

Together AI notified customers of a security incident in which an unauthorized party downloaded stored prompt samples from its serverless inference service submitted in June and July 2026. A credential for a private Hugging Face repository was exposed in a public GitHub repository by an AI coding agent, and the unauthorized party later searched the samples for credentials to other services. Together AI says it is investigating with forensic experts and law enforcement, has strengthened credential and development controls, and is making zero data retention the default for new accounts.

Data Processing Incidents on Inference Providers

OpenRouter said security researchers found that certain inference providers had compromised data or failed to uphold data policies, though there was no breach or unauthorized access of OpenRouter itself. In response, OpenRouter off-boarded three providers, removed 125 endpoints from Zero Data Retention eligibility, and notified potentially affected customers. It is also adapting the researchers’ testing approach and requiring validation before restoring ZDR eligibility.

🚗 Automotive & Hardware

Tesla launches FSD v15 with AI video

Tesla has begun rolling out Full Self-Driving v15 to some vehicles, targeting weaknesses such as road signs, potholes, school zones, inclement weather, and safety-critical cases, with claims of 25x better handling. The update adds a photorealistic “Reality View” visualization using real-time AI video rather than game engines, plus speed-camera slowdowns, Bike Rack Mode, and improved navigation and parking. Tesla did not specify which vehicles receive it first.

AI server suppliers Dell, HPE, Lenovo, Supermicro demand higher prices after signed contracts

AI server suppliers including Dell, HPE, Lenovo, and Supermicro are reportedly demanding substantially higher prices from buyers months after contracts were signed. The increased equipment costs are hitting trading firms.

Toyota, Nissan, and Honda to Standardize Parts to Lower Costs

Toyota, Nissan, and Honda will begin standardizing some vehicle parts to lower production costs, starting with unified steel quality criteria, consistent plastics protocols, and a universal color-coding scheme for wiring harnesses. Discussions will first involve Japanese domestic suppliers before expanding to international suppliers, with initial standards expected next year. The automakers aim to redirect savings into innovation as they compete with increasingly competitive Chinese automakers.

White remains the most popular U.S. car color at 25.7% market share, followed by black at 23.4% and gray at 22.9%. Grayscale colors now account for 80.4% of the market, up from 47.3% in 1996, while red and green have fallen sharply to a combined 9.2%. Gray saw the biggest gain, rising from 3.6% to 22.9% over the period.

💻 Tech Culture & Society

Fixing Google’s Feedburner problem, and caching that half works

The article describes a workaround for Google Chrome Releases’ stale Feedburner Atom feed: adding ?redirect=false to the Blogger RSS feed returns fresh results, allowing CCBot to resume notifications about Chromium vulnerabilities. It also notes that Feedburner still ignores caching headers and serves outdated content, while the redirect=false feed sometimes returns 304 Not Modified but inconsistently.

Come for the science, stay for the big iron: Supercomputer admin on why he loves the job

A supercomputer administrator explains why he loves his job, saying he came for the science and stayed for the “big iron.” His career spans from the dotcom bubble to managing a system with 200,000 cores of raw compute power.

South Korea education minister resigns over exam cheating scandal

South Korea’s Education Minister Choi Kyo-jin has resigned, taking full responsibility for confusion caused by last-minute measures to stop AI-assisted cheating during the annual Suneung college entrance exam. The move followed a scandal in which the top scorer of a national tax accountancy exam admitted using a mini-camera, mobile phone, and AI tools to cheat, prompting nationwide emergency alerts. Officials said 111 Suneung candidates were found with unauthorized smart devices.

Man Claiming to Be Fired Rockstar Engineer Says GTA 6 Is Tracking Behavior

An anonymous 4chan poster claiming to be a former Rockstar North senior engineer alleges that GTA VI feeds granular gameplay data into an internal Take-Two system called “Lens” to generate behavioral and psychological classifications, then shares profiles with a Delaware company called Vantage Analytics. The poster’s identity and claims remain unverified, and no reference to Vantage Analytics was found in reviewed Take-Two SEC filings. Rockstar and Take-Two did not respond in the reporting reviewed, so the allegations are unconfirmed.

The Engineering Manager’s role is under attack (again)

The article argues that engineering manager roles are again being cut as companies flatten and increase spans of control, repeating past mistakes. It warns that overloading EMs and expecting them to ship code erodes their managerial capacity and value, making them seem less necessary. The piece notes that tech companies have run this experiment before and eventually add management layers back.

Notes on the discrete Fourier transform (DFT)

The article explains the Discrete Fourier Transform (DFT) for finite discrete signals by treating the samples as one period of a periodic signal and using Discrete-Time Fourier Series coefficients. It notes that the DFT samples the continuous DTFT at equally spaced frequencies, making it computable on a computer. A Python implementation of the DFT is also provided.

Job ghosting: What applicants can do if employers do not respond

Job ghosting—employers failing to respond to applications or interviews—has become a major problem for UK jobseekers, particularly young people, leaving them demoralised and without feedback to improve. A review by former minister Alan Milburn recommends that large employers provide meaningful feedback to unsuccessful applicants under 25, though firms say high application volumes and AI-driven mass applications make responding to everyone difficult.

“Dumbass” RFK Jr. Rescued Measles from Obscurity. What Horror Is Next?

The article criticizes HHS Secretary Robert F. Kennedy Jr. after the U.S. lost its measles-elimination status, citing a Wall Street Journal report. It says measles cases have surged to more than 7,700 with 13 deaths—the highest totals in nearly four decades—and cites Reuters reporting that Trump and Kennedy have effectively encouraged the disease’s spread.

Trump at the “Summit”: One of His Dumbest Moments in the Last 10 Years

The article criticizes Donald Trump’s opening remarks at a “trilateral” summit with Vladimir Putin and Xi Jinping, calling them among his dumbest and most cynical moments. It argues that Trump’s suggestion about leaders courageously giving up nuclear weapons relies on public ignorance, since the U.S. and Russia have already dismantled thousands of warheads.

The No AI Slop Manifesto

Attendees of the AIE Code Summit 2026 propose a “No AI Slop manifesto,” modeled on the Agile Manifesto, valuing outcomes, understanding, human thinking, better work, and system improvement over speed and easy output. They define slop as passing along AI-generated work without taking responsibility for its quality or the reader’s experience, warning that it compounds and imposes high review costs. The manifesto resolves to use AI responsibly, preserve human judgment and ownership, clearly disclose agent-written content, and review outputs according to stakes.

Sandwich Simulator

The article presents a Sandwich Simulator that guides users through making a sandwich by tapping a pot to load spread, swiping over bread, dragging toppings from a bar, and rotating them. It notes that wobbly toppings can slip off when released, and a red button takes a photo that can be kept with a number and link.