Alvaro Lopez Ortega / 2026-10-19 Briefing

Created Mon, 19 Oct 2026 12:36:32 +0000 Modified Mon, 19 Oct 2026 12:37:18 +0000
6326 Words

Interpol-led raids across 17 countries led to 95 arrests and rescued a 5-year-old in Colombia from online child-abuse networks. Frontier will pay Connecticut $625,000 over a 2024 breach exposing 58,841 residents’ data, plus a $5.64M class settlement. Apple banned apps that create nonconsensual sexual images of real people. Microsoft scrapped its flagship UK AI supercomputer, citing grid delays.

🤖 AI & Automation

Harness Engineering 101: The LLM is the brain, the harness is the body

“Harness Engineering 101” is a blog series published as a book that teaches building AI agents from first principles. Its central thesis is that LLM APIs are stateless, so a “harness” must build, maintain, and protect the full conversation array, with “agents” understood as patches to that loop. Across 16 chapters, it develops a ~300-line toy agent in plain Python, with each chapter structured around a failure, its minimal patch, and the patch’s cost.

Mistral launches time-series foundation models

Mistral has released Timestral, a family of time-series foundation models in four sizes from 5M to 209M parameters that provides probabilistic, multivariate forecasts with past and future covariates. The 209M-parameter model ranks first on GIFT-Eval, TIME, and fev-bench among zero-shot pretrained models, outperforming larger models, while all sizes balance accuracy and size and run efficiently on CPUs. Timestral combines synthetic pre-training, covariate-aware attention, and long-context learning, with fine-tuning available via Mistral Forge.

Agent-Skills.market

Agent-Skills.market is a directory of agent skills designed for everyone, not just developers. It features automation tools like Script.it, AI Agent Email Inbox, Google Workspace workflows, and WhatsApp automation, alongside design and creative skills for video, slides, diagrams, branding, and data visualization.

Strot – Compile an LLM’s intent once, run decisions locally forever

Strot is an open-source decision compiler that uses an LLM once at compile time to turn intent and schema into a frozen Python artifact. At runtime, it performs classification, scoring, routing, and typed decisions locally on CPU at zero marginal cost, returning type-safe JSON with probabilities and confidence while abstaining on low-confidence records. It is installable via pip as strot-ai and licensed MIT.

Yila – A Codex for Scientific Research

Yila is a research-assistant platform trusted by institutions worldwide that supports CSV/XLSX data analysis, figure generation, citation-backed literature reviews, academic poster and slide creation, verified paper searches, and manuscript formatting. It delivers clear results, professional charts, reusable files, and presentation-ready academic materials tailored to specified guidelines.

doubleAI debuts code-execution model to optimize GPU/CPU; raises $220M, buys 1,000 Nvidia B300 chips

DoubleAI, a Tel Aviv-based startup, has released its debut code execution model aimed at optimizing GPU and CPU performance and lowering companies’ computing costs. The company, co-founded by Mobileye CEO Amnon Shashua, has raised $220 million and acquired 1,000 Nvidia B300 chips.

Anthropic’s opaque five-member Long-Term Benefit Trust: no equity, huge sway

Anthropic’s Long-Term Benefit Trust is an unusually opaque five-member body whose trustees own no equity but hold enormous sway, including power to shape the company’s board and block major moves. Few details about how the trust operates or its governing documents have been disclosed. The article notes this as Anthropic courts public-market investors ahead of a potential IPO and plans to put 5% of its shares into a mission-related fund.

Google DeepMind: four Gemini 4 Argon agents beat Minecraft unaided in 27h14m for $1,620

Four Gemini 4 Argon agents autonomously beat Minecraft, slaying the Ender Dragon in 27 hours and 14 minutes after being placed in a randomly generated world with no assigned roles, strategies, or game-specific tools. Google DeepMind said the agents self-organized, wrote 2,783 JavaScript programs, exchanged 229 chat messages, and died 605 times. The run cost $1,620 at Argon’s planned 2027 price.

AI extreme weather forecasts face major test amid historic El Niño as scientists question accuracy

AI is being used to forecast extreme weather and other risks, from wildfires to futures contracts, but its accuracy faces a major test during a historic El Niño. Some scientists question whether relying on the technology is a mistake.

The substitution myth & the tough spot it puts us in

The article argues that AI adoption rests on the “substitution myth”—the false assumption that AI can replace human work one-for-one without changing the broader work system. Instead, AI reshuffles roles and creates new tasks such as configuration, integration, and monitoring, making human cognitive work more critical even as adoption outpaces understanding. As a result, people are often required to intervene when AI fails at exactly the moment they are least prepared to do so.

🔒 Security & Privacy

Frontier settles data breach affecting 59K Connecticut residents for $625K

Frontier Communications agreed to pay Connecticut $625,000 to settle an April 2024 data breach affecting 58,841 residents, in which hackers used compromised employee and vendor credentials to steal unencrypted data including Social Security numbers later offered for sale on the dark web. Frontier also committed to stronger security measures and separately reached a $5.64 million class-action settlement covering about 750,000 people, including two years of free credit monitoring.

Interpol-led operation strikes online child abuse networks, resulting in 95 arrests

Interpol coordinated four investigations across 17 countries targeting online child sexual abuse material, leading to 127 search warrants, 95 arrests, and the rescue of a 5-year-old child in Colombia. Authorities uncovered distribution through messaging groups and seized devices, including more than 4,000 abuse files in Ecuador, and expect further arrests as seized devices are analyzed.

Apple tightens App Store rules to keep AI nudification apps out

Apple updated its App Review Guidelines to explicitly ban apps that create sexually explicit images of real people without consent. Apps with user-generated content must also filter harmful material and provide reporting and blocking tools, while apps cannot hide banned features during review. The change follows EU and UK efforts to curb non-consensual intimate images.

Broadcom Wi-Fi bug turns sweet 16, still lets strangers hijack your router

A Broadcom Wi-Fi bug has remained unpatched for 16 years and can still let attackers hijack vulnerable routers. A researcher claims Broadcom knew about the flaw in 2018, and millions of devices may still be at risk.

Engineer was so sick of office pranks, he ‘wiped’ his antagonist’s disk

An engineer fed up with office pranks took revenge by wiping his antagonist’s disk. The article’s summary quips that the revenge was “best served by autoexec.bat,” suggesting batch-file sabotage.

ConnectWise turns off its SaaS-y backup, security, and management services to probe security SNAFU

ConnectWise has suspended its SaaS backup, security, and management services while investigating a security incident. The move follows Huntress’s allegation that Microsoft 365 accounts were taken over.

Let’s Encrypt is adding two additional remote perspectives for domain validation

Let’s Encrypt is adding two remote perspectives for domain validation, increasing validation requests from five to seven and requiring a quorum of remote responses to improve security against BGP attacks and meet Baseline Requirements. The change is expected to have little to no user impact, with staging rollout on October 27, 2026, and production rollout on November 10, 2026.

Man arrested after discussing plans for childcare mass shooting with [Claude]

A Sydney man, David Lan Thackray, has been arrested and charged with using a carriage service to menace, harass or offend after allegedly discussing plans for a mass shooting at a childcare centre with the AI chatbot Claude. The 29-year-old was remanded in custody for a mental health assessment; his lawyer said the remarks were a stress-induced tirade with “no weight” behind them.

Footage of the Flock search that a judge ruled was uncosntitutional

Body camera footage obtained by 404 Media shows the first Flock license plate reader searches ruled unconstitutional by a federal judge. The judge found the warrantless searches in Melissa Kyle’s arrest were “indiscriminate mass surveillance” violating the Fourth Amendment, and threw out the evidence—including more than 91 pounds of meth—found after police used Flock to justify stopping and interrogating her. The footage shows a Tulsa County deputy searching Flock and another ALPR system before fully pulling Kyle over and using the traffic stop to question her travel history.

US cyber job postings fell 26.4% to 4-year low of ~236K; AI security engineer demand rose 200%+

US cybersecurity job postings fell 26.4% year over year to a four-year low of 235,786 roles from January to September, according to CyberSN. Despite the overall decline, demand surged for AI Security Engineer roles (+268.3%) and Cyber Data Scientist roles (+203.5%), with just under 15% of postings referencing AI.

Sam Altman’s World launches World ID suite for banks to verify humans, agents

World, the identity project co-founded by Sam Altman, launched a World ID verification suite for banks and financial firms, its first industry-specific offering. The suite bundles the World ID app, a now generally available World ID for Zoom deepfake detection, and World ID for Agents; World named OpenExchange, Bain Capital, VanEck Funds and Outtake as design partners but did not identify any bank using the product.

Tor VPN Beta: What we’ve learned building our own VPN for Android from scratch

Tor VPN Beta for Android extends Tor’s privacy protections beyond the browser to entire devices, giving each app its own Tor circuit to reduce cross-app tracking. Early use has focused on circumventing censorship, especially in Iran and Turkmenistan, prompting priorities like WebTunnel bridge support and stability improvements. Feedback showed exit selection can confuse censorship-circumvention users, so the design now emphasizes connecting to Tor before choosing exits.

TAINT_FORCED_BIND Sent In For Linux 7.4 To Deal With Major Fuzzing Attack

Linux 7.4’s driver core pull request introduces TAINT_FORCED_BIND to mark kernels where the sysfs bind/unbind interface is used, helping identify bug reports caused by forced driver/device combinations. The change responds to a major fuzzing “attack,” including syzbot, that randomly bound drivers to devices and generated unneeded errors and patches; the pull request also includes new Rust driver additions.

⚙️ Hardware & Infrastructure

PS6 will likely cost over $1,000 thanks to AI, hardware leaker claims

A hardware leaker claims Sony could spend at least $680 on PS6 RAM and storage alone, potentially pushing the console’s retail price above $1,000. The reported costs are linked to AI-driven shortages and higher memory and storage prices as chipmakers prioritize AI servers and chips.

Microsoft abandons UK AI supercomputer as grid struggles to keep up

Microsoft has abandoned its flagship AI supercomputer project in Essex, UK, amid power connection delays and grid constraints. The decision threatens to push the UK’s AI ambitions back until the 2030s.

Museum exhibit reveals the intricate workings of a knackered Dell

A museum exhibit is revealing the intricate inner workings of a worn-out Dell computer. It was expected to focus on the Jacquard loom instead.

AI inference is running into a memory bottleneck. Huawei wants to tackle it with OceanStor M900

Huawei is targeting the memory bottleneck in AI inference with its OceanStor M900 Context Memory Storage platform. The platform is aimed at improving the economics of large-scale AI inference.

Chicago O’Hare Digs Up 50-Ton Boulder Under Airfield, Asks Public for Name Ideas

Chicago O’Hare International Airport uncovered a 50-ton boulder about 30 feet beneath its airfield during excavation for a stormwater management facility tied to the ORDNext expansion. The city turned the discovery into a public naming contest, receiving nearly 3,000 suggestions and announcing six finalists on October 16, 2026. Online voting is open until November 2, 2026, and the winning name will be displayed alongside the boulder in a new airport exhibit.

Deutsche Telekom, Orange, Telefónica, and Vodafone have agreed to explore a joint bid for spectrum to provide direct-to-device mobile satellite coverage in Europe. The move would challenge Starlink in the European satellite service market.

Analysis: data center capacity additions to surge to 50GW+ in 2027; 24GW+ at 40 campuses advanced

SemiAnalysis reports that 60GW of data center capacity is under construction globally, with more than 24GW across 40 campuses in advanced stages and slated for delivery in 2027. The firm forecasts 50GW+ in capacity additions that year—roughly the entire cumulative global data center market through 2024 built in a single year—despite constraints on power, labor, supply chains and capital. Nearly $1 trillion in binding, non-cancellable lease contracts signed year-to-date support the outlook, though the report warns the bottleneck could shift to chips.

Q&A: Nvidia Vera team says agentic AI pushes CPU design to single-thread over core counts

NVIDIA’s Vera team says agentic AI is increasing CPU demands and pushing CPU design back toward single-threaded performance over raw core counts. The Vera CPU is positioned as a specialized conductor for sequential reasoning-loop tasks, helping keep expensive GPUs highly utilized. It combines high per-core speed, LPDDR5X memory bandwidth, and a monolithic die linked by NVIDIA’s Scalable Coherent Fabric to reduce latency and tighten CPU-GPU orchestration.

Intel unveils D-series data center GPU with 480GB LPDDR5X for inference

Intel unveiled its GPU D-series data-center accelerator, featuring up to 480GB of LPDDR5X memory—an unusually large capacity aimed at demanding AI inference workloads and a new record for GPU memory. The card is designed to support models with up to 300 billion parameters on a single GPU, while an eight-GPU system can handle models exceeding one trillion parameters. Intel rates it at 1.3 petaflops FP8, 1.5TB/s memory bandwidth, and a 350W power envelope in a standard PCIe form factor.

Crawlers, floods and fleets: a year of keeping our digital collections open

The article’s body text was inaccessible, consisting only of a notice to enable JavaScript and cookies. Its title indicates a review of a year spent keeping digital collections open amid challenges involving crawlers, floods, and fleets.

Jev-Driven SRE Diagnosis: What Worked and What Failed

The article describes a Jev-driven SRE diagnosis pipeline that runs without an LLM agent: a programmatic collector gathers and organizes Kubernetes evidence, and Jev selects among supplied options to identify likely root causes and supporting observations for a diagnosis report. Across 21 SREGym-Lite faults, the pipeline passed 80 of 105 diagnoses (76.2%) with a median diagnosis time of 14.6 seconds. Jev investigates candidates one at a time and does not generate commands or write the final report.

Clearer, Not Louder: Simona Ursan on Leading Incidents

In an interview, incident commander Simona Ursan argues that a successful incident is not defined by speed alone but by safe recovery, service stability, and lessons learned. She says credibility under pressure comes from calmly asking clarifying questions and giving regular updates to restore a shared picture, not from being the loudest voice. She also recalls first feeling like she was leading an incident more than 15 years ago as a platform engineer when she guided a colleague remotely without system access.

Key Autoscaling Metric #2 – Mean Time to Stop Sucking

The article introduces “Mean Time to Stop Sucking” (MTSS) as a proposed Kubernetes autoscaling metric for measuring how long users experience degraded performance due to autoscaling behavior. It argues that reliability should be split into scaling up and scaling down, with MTSS focused on delays in provisioning new capacity during scale-up. The author contends that despite its name, MTSS is a real metric teams should track in their autoscalers.

Stop paging on how your machines feel

Companies often struggle with customer-detected outages because their alerts focus on easy infrastructure metrics rather than user experience. Infrastructure metrics like CPU and memory can cause false alerts and fatigue, while user-facing work metrics such as request success rate and latency better indicate real incidents. The article recommends paging on work metrics and treating resource metrics as diagnostic dashboard data.

SpacemiT Announces The V100 Server-Class RISC-V Processor

SpacemiT announced the V100, which it claims is the first processor fully compliant with the RISC-V server specification. The 46-core hybrid chip combines six XiangShan Kunminghu V2R2 cores with 40 SpacemiT X100 cores, supports up to two sockets for 92 total cores, and offers 2.2–2.4 GHz clocks, a 12nm process, 120W TDP, eight-channel DDR5-4800, and 52 PCIe 5.0 lanes. It supports server features including UEFI, ACPI, and SMBIOS.

💰 Business & Finance

Microsoft’s second-hand licensing appeal gets the brush-off

The UK Court of Appeal has refused Microsoft permission to take its second-hand software licensing dispute with ValueLicensing to the Supreme Court. A “Known Adverse Document” related to the case is still to come.

AI’s epic cash call is matched only by its ability to strangle revenue streams

The article examines how AI’s massive funding needs are matched by its tendency to disrupt the revenue streams of the companies adopting it. It frames AI as a self-disrupting technology and notes the irony in the naming of AI clone products.

Verisign .com Domain Name Registry Consumer Class Action

Hagens Berman has filed a consumer class action against Verisign and ICANN alleging an illegal monopoly over .com domain names that overcharged consumers. The lawsuit says Verisign’s .com price rose more than 30% since 2021 to $10.26 per domain despite falling registry costs, and it targets people who registered or renewed a .com domain since 2022.

IMF World Economic Outlook 2026 [pdf]

The provided item is a PDF titled “IMF World Economic Outlook 2026.” No further details or findings are included in the supplied text.

The AI industry’s gigantic debts are making investors nervous

The article content is unavailable because it requires JavaScript and cookies to be enabled, so no substantive details can be summarized. The title indicates that investors are becoming nervous about the AI industry’s large debts, but the body text does not provide supporting information.

Vinted sees suspected stolen goods listings after French riots hit luxury stores

Vinted has seen an influx of listings suspected of containing goods looted during widespread French riots, which targeted stores including Sephora, Louis Vuitton, and Burberry; France is the company’s largest market. Vinted said it immediately removed linked listings, stepped up monitoring of suspicious items, and will act against accounts that breach its rules, including permanent bans.

Tenstorrent raises $1.5B Series E led by Trace/Innovation X at $6.2B valuation, up from $2.6B

AI chip startup Tenstorrent raised $1.5 billion in a Series E round led by Trace Capital and Innovation X, valuing it at $6.2 billion. The new valuation is up from $2.6 billion after it raised $700 million in 2024. Tenstorrent had attracted takeover interest earlier this year.

Foxconn to buy 50% of Mitsubishi auto unit for $727M; Mitsubishi to supply AV tech for Foxconn EVs

Foxconn’s Hon Hai Precision Industry is acquiring a 50% stake in Mitsubishi Electric’s automotive-components unit for ¥115 billion ($727 million), aiming to grow it into a business with over ¥1 trillion in sales. Mitsubishi Electric will provide technology in electric powertrains, autonomous driving, and in-car intelligence for Foxconn-developed EV platforms.

Musk: SpaceX share for every US kid in 2027, ~$10.5B; Anthropic gives 5% to APBF

Elon Musk said he will give one of his SpaceX shares to every child in the U.S. in 2027, a gift estimated at about $10.5 billion based on 72 million children and SpaceX’s $146.03 share price. The announcement followed Anthropic’s plan to commit 5% of its shares to a public benefit fund, with Musk saying he recommends giving shares directly to the public.

Stripe Link data shows agentic transactions rose 38x and payment volume 20x in September, with retail leading both and an average spend of about $100. Link now powers about 200 agents, more than double the roughly 70 signed on when the September figures were compiled, as AI assistant tools increasingly handle checkout and payments for users. Stripe’s Dan Hill says agentic commerce is growing and becoming material, though it remains a small share of Link’s 300 million users.

Abby Care, caregiver training platform, raises $82M Series C led by Transformation Capital

Abby Care, an AI-native home care startup founded by Havi Nguyen in 2022, has raised an $82 million Series C led by Transformation Capital. The company trains and certifies family members as paid caregivers and works with health plans as a Medicaid provider to address home-care workforce shortages. New investors include CVS Health Ventures, Frist Cressey Ventures, and Pivotal Ventures, with existing backers Khosla Ventures, Thrive Capital, and Sequoia participating.

Allium: stablecoin payments up 45% YoY to $482B YTD; transactions +71%, supply +0.4%

Stablecoin payments reached $482 billion in the first nine months of 2026, up 45% year over year, while transactions rose 71% to 1.04 billion, according to Allium. Circulating supply grew just 0.4% to about $305 billion, indicating existing stablecoins are being used more frequently for payments. The growth comes as Visa, Mastercard, and Stripe announced services integrating stablecoins with traditional payment rails.

China Q3 GDP to grow 4.5%, up from 3-year-low 4.3%, as AI exports offset weak spending

China’s GDP grew 4.5% year on year in Q3, up from 4.3% in Q2, as AI-driven exports helped offset weak domestic spending. The figure was below Q1’s 5% and at the low end of Beijing’s 4.5%-5% annual target, amid a prolonged property slump, falling property investment, and uncertainty over trade tensions and AI investments.

WSJ: ADI’s fee wallet funded 795 accounts for 99% of FIFA prediction market activity since late Aug

ADI Predictstreet, backed by Abu Dhabi’s royal family, paid FIFA more than $150 million for World Cup advertising rights. A Wall Street Journal analysis found that 795 accounts funded by the platform’s own fee-collecting wallet drove 99% of its activity since late August, suggesting most trading volume is inauthentic.

Europe’s top industrials cash in on AI; Schneider, Siemens push software, ABB equipment

Europe’s three biggest industrial companies are benefiting from the AI boom. Schneider Electric and Siemens are expanding further into software, while ABB is focusing on physical equipment.

SF home insurer Stand raises $70M Series C at $1B valuation

Stand, a San Francisco-based home insurer that uses digital twins of homes to simulate wildfire and hurricane risk, raised a $70 million Series C at a $1 billion valuation. The round was led by Saga Ventures, with participation from Lowercarbon, Eclipse, and others. Stand says it now insures over $10 billion of property, up from $1 billion a year ago, and is growing about 15% monthly in California and Florida.

🧑‍💻 Software & Development

Tmux and Claude Code: Sessions, Send-Keys, and the Ceiling

The article outlines a tmux-based workflow for running Claude Code and other AI coding agents in named, detachable sessions that survive terminal or SSH disconnects, with multiple agents per window and scripting via send-keys and capture-pane. It first covers Claude Code’s built-in --worktree --tmux option for single-agent branch isolation, then details custom session-per-project naming, pane layouts, and status monitoring. It also discusses practical limits encountered around four agents.

The Annotated VT100 Firmware

The VT100, launched in 1978, was DEC’s first ANSI-compatible terminal and first to use a microprocessor, the Intel 8080. Escape-sequence changes in the cost-reduced VT101 and VT102 became DEC’s definition of “VT100 compatible” in its internal Video Systems Reference Manual for the next 15 years. The documents explain the VT100’s original 8 KiB firmware through a fully annotated disassembly linked to the technical manual and print set.

Superlite: A Faster Alternatie to Superpowers

Superlite is a new Claude Code plugin built as a faster alternative to the collaborative Superpowers workflow, aiming to let users hand over a spec and return to tested, committed work without constant approval prompts. It plans builds, splits tasks to run in parallel, assigns each to the cheapest capable agent, checks seams, reviews results, and logs decisions; the article also presents benchmarks and usage details.

Starcraft: Ashfall

The article presents an in-game briefing for “Starcraft: Ashfall,” featuring a Terran Expeditionary Force mission titled “Ashfall Bridge” Chapter 01. It displays loading and combat interface elements such as reinforcement beacons, kill-speed tracking, mutation factors, supply pickups, deployment controls, route selection, and a highest kill record. It also instructs players to play the game in portrait orientation.

Sanity’s Edge – A voice BBS from 1994

Sanity’s Edge was a telephone-only voice bulletin board system run from 1994 to March 1997 on a dedicated 286 PC in Tampa, Florida, created by a father and his ten-year-old son. Callers navigated an asylum-themed board by keypad and left voice messages; over three years it received more than 26,000 calls and 46,000 voice recordings. A backup found among the father’s stored data was later fully decoded using AI and the Reverse Engineer Anything project, restoring the original audio and rebuilding missing C source.

Beginning of the Ende (a community Deno fork)

Deno’s team is moving to Cloudflare and will stop working on the runtime in a year, leaving it MIT-licensed for community maintenance. In response, the author announced Ende, a Deno fork that drops Node compatibility and other features, plans per-module permissions, sandboxes, and post-quantum crypto, and targets a December soft launch. Ende will also replace SemVer with ChronVer.

Programming Language Fundamentals in Agda

The book “Programming Language Fundamentals in Agda” introduces programming language theory using the Agda proof assistant. Its contents cover logical foundations, programming language foundations, and denotational semantics, plus an appendix on substitution. The project welcomes comments and GitHub pull requests and lists related courses and derived works.

Hot take

The author corrects their misunderstanding of “hot take,” noting it means a strong personal opinion that is usually not carefully thought out and likely to be disagreed with, not “telling it like it is.” They apologize to Alex for last week’s issue and discuss how labels, information bubbles, and algorithmic engagement shape perception, arguing that respectful debate can help people get closer to reality.

Saturation at GitHub: the saga continues

On September 13, GitHub suffered a site-wide incident when an internal data-cleanup job wrote to a shared permissions database cluster; its safeguard monitored only replica lag and missed the load building on the primary. The primary exhausted available connections, and slow database calls plus a retry loop around token creation amplified saturation and prolonged errors. GitHub declared the incident at 08:50 UTC and recovered all services by 10:44 UTC after shedding internal load and pausing the job.

Beyond synthetic testing: Capturing and replaying real database workloads at Airbnb

Airbnb describes a method for capturing real production database workloads and replaying them offline instead of relying only on synthetic testing. The approach is used to load-test systems, plan capacity, and de-risk database upgrades.

GNOME Mutter Lands Improvement For Reducing GPU Work While In Power Savings Mode

A set of 17 patches has been merged into GNOME Mutter to address high GPU usage while the monitor is sleeping. The fix prevents unthrottled repainting of CRTC back-end stage views in power savings mode and handles discarded frame presentation feedback. The changes are expected to be included in GNOME 51.2.

BeamFS Sent Out As Newest File-System Proposal For Linux

An RFC patch series has been posted for BeamFS, a new Reed-Solomon-protected Linux file-system derived from the stalled FTRFS proposal. BeamFS is designed for single-storage-device systems in radiation-prone environments, correcting corrupted bytes in place without a second copy. It incorporates feedback from FTRFS in hopes of better upstream acceptance.

Linux 7.4 Removing Its Broken & Rarely Touched Code For ARM64 Big Endian

Linux 7.4 is removing ARM64 big-endian kernel support because the code is rarely used, unsupported, and was previously marked broken. The removal clears out less than 500 lines of code and simplifies future ARM64 kernel development. The ARM64 updates also begin adding NVIDIA Rigel CPU support.

RADV Promotes VK_EXT_host_image_copy To Enabled By Default For RDNA1 GPUs

RADV has enabled the VK_EXT_host_image_copy extension by default for RDNA1 (GFX10) GPUs and added experimental support for older GFX9 GPUs. The extension was previously enabled by default only for RDNA2 and newer GPUs, and the change will be included in Mesa 27.0. It allows direct copies between host memory and Vulkan images, reducing overhead, especially for iGPUs.

🌍 Science & Society

Cybernews Readers’ Choice Awards 2026 voting is open now — help pick this year’s best tech

Cybernews has opened voting for its first Readers’ Choice Awards 2026, inviting readers to choose winners across 18 tech categories from VPNs and antivirus to website builders and AI tools. Voting closes at 00:01am PST on Sunday, November 1, 2026, and winners will be announced on November 5.

Dr Garry Hunt talks to us about the astonishing Voyager probes

Dr Garry Hunt, the UK’s sole representative on Voyager’s imaging team, discussed his role in the mission on The Register’s Kettle podcast as the probes approach a light-day’s distance from Earth. He also covered the mission’s future and the state of science in the UK, US, and beyond. Voyager launched nearly 50 years ago and is still operating.

ICANN vows to work faster, improve transparency

ICANN has pledged to work faster and improve transparency. Its leaders expect AI and geopolitics to create more governance challenges and threats.

ESA’s Hera Will View Asteroid Scars from DART’s Strike

ESA’s Hera mission is approaching the binary Didymos-Dimorphos asteroid system to study the aftermath of NASA’s 2022 DART impact. Hera will provide close-up views and deploy the Juventas and Milani CubeSats to analyze the asteroids’ dust, composition, and internal structure. The mission aims to improve understanding of asteroid deflection for future planetary defense.

Alternatives to Animal Testing Are Finally Taking Off

Organ-on-a-chip systems that mimic human organs are advancing as alternatives or complements to animal testing, with one pharmaceutical company reportedly asked by the FDA to provide such data rather than animal data. However, animal models remain the default in biomedical research, and adoption is slowed by validation needs and established habits. The shift away from animal testing is promising but still limited.

America Is Now an Ally of Russia

The provided article contains only a Cloudflare security verification notice and no substantive news content, so a factual summary cannot be produced.

Writing is no longer proof of work

The article argues that much workplace writing functions as proof-of-work evidence—demonstrating compliance, effort, or learning—rather than genuine communication, and that tech companies structurally depend on this. LLMs undermine the practice by making plausible filler cheap to generate, while firms can only partially verify such writing manually, unlike blockchains’ automatic verification.

The Rise of In-Office Remote Work

The provided page contains only a Cloudflare security verification message and no substantive news content. Therefore, no summary of the article can be produced.

Survey: US chatbot gender gap closed since 2024; men more often daily users; ChatGPT use equal (Pew)

A Pew Research Center survey finds the gender gap in overall AI chatbot use among U.S. adults has closed since 2024, with 50% of men and 47% of women reporting use in 2026. Men remain more likely to use chatbots daily (27% vs. 20%) and more likely to use brands such as Copilot, Gemini, Grok, and Claude, while ChatGPT use is identical at 44% for both genders. Women are also more skeptical about AI and its impact on their lives.

Data center industry courts kids amid backlash: 11+ firms host field trips, flag kits, cards

Amid growing backlash, data center companies and industry groups are courting young students through field trips, a children’s story about a honey badger touring a data center, and school materials tied to “International Data Center Day.” In one example, an industry group sponsored Memorial Day flag kits and informational cards for 3,100 first graders in Virginia’s Loudoun County, part of a broader corporate effort to shape the industry’s image in schools.

54% of 94 economists expect AI gains to spread across G20, not just US/China; 32 see Africa behind

A survey of 94 economists found that 54% expect AI’s gains to spread across the G20 over the next decade, rather than being limited to the US and China. Meanwhile, 32 economists said they expect African countries to fall behind.

What are you doing this week?

The article invites readers to share what they are doing this week. It also notes that it is acceptable to do nothing at all.