LastWeekIn.Tech

Week ending 2026-08-31

2026-08-31T15:59:40Z

Nancy Grace Roman Space Telescope — Hacker News
NASA's Nancy Grace Roman Space Telescope, launching soon, will carry an experimental coronagraph instrument designed to directly image starlight reflected off exoplanets for the first time, a capability that will affect astronomers across all areas of astrophysics. The coronagraph uses two palm-sized deformable mirrors equipped with approximately 2,300 tiny actuators each—NASA's first active deformable mirrors in space—along with advanced star-blocking masks and supersensitive detectors to overcome the extreme challenge of detecting faint planets against bright starlight, comparable to photographing a firefly next to a floodlight from across the country. Scientists including Vanessa Bailey at NASA's Jet Propulsion Laboratory and Margaret Turnbull at the SETI Institute will test the instrument by attempting to spot known exoplanets and observe dust around nearby stars, gathering crucial data to inform future telescopes like the planned Habitable Worlds Observatory and Lazuli Space Observatory. The coronagraph's success will pave the way for future space telescopes capable of detecting potentially habitable, Earth-like planets orbiting sun-like stars.
ChatGPT, Reddit and Roblox to face the EU's strictest platform rules — Engadget
The European Commission has designated ChatGPT as a "Very Large Online Search Engine" and Reddit and Roblox as "Very Large Online Platforms" under the Digital Services Act, subjecting them to the EU's strictest regulatory requirements after each exceeded 45 million average monthly users in the European Union. These designations require the three platforms to implement stricter enforcement of their services and algorithmic systems, particularly regarding illegal content, protection of minors, and electoral integrity, while also submitting to closer EC supervision including data sharing, annual independent audits, and researcher access to platform data. With these additions, the European Commission now designates 28 services as very large online platforms or search engines under the DSA, joining previously designated services including WhatsApp, Facebook, and Instagram. The platforms have four months to comply with the tighter regulations.
A Judge Has Blocked the Pentagon’s Attempt to Blacklist Anthropic — WIRED
US District Judge Rita Lin ruled Thursday that the Trump administration's February designation of AI company Anthropic as a "supply-chain risk" was unconstitutional retaliation, vacating Defense Secretary Pete Hegseth's decision to bar the company from federal contracts. The ruling also removed sanctions imposed by nine agencies including the Pentagon, Treasury Department, State Department, and Department of Homeland Security, though Lin affirmed the Pentagon retains the right to choose not to use Anthropic's models. The dispute originated from a failed $200 million contract negotiation over Anthropic's Claude AI models for military use, with Anthropic seeking restrictions on lethal autonomous weapons and mass surveillance applications that Hegseth rejected. The Pentagon is expected to appeal the decision, while Anthropic stated it remains focused on working with the government on AI for national security.
I Asked 100 Companies for My Data. I Got Deletion Notices Instead — WIRED
A WIRED reporter filed over 100 data access requests under California's Consumer Privacy Act and found that many companies mishandled them, with Crunchbase deleting the reporter's account despite explicit instructions not to, and BeenVerified repeatedly misclassifying the access request as a deletion request. Cash App listed a phone number in its privacy policy for data access requests but customer service representatives were unable to process calls through that method, according to the reporter's experience. Privacy advocates and researchers told WIRED that these failures reveal companies are not investing adequate resources in CCPA compliance and suggest that data minimization—limiting what companies can collect in the first place—would be a more effective consumer protection approach than relying on access requests.
Cities terminate Flock contracts at record pace in August — TechSpot
Local governments are rapidly ending contracts with Flock Safety, the company behind an AI-powered automated license plate reader camera network, with 214 cities and counties dropping the service since 2021 and 90 jurisdictions canceling in August alone. Concerns driving the cancellations include surveillance overreach, data sharing among agencies, system costs, and reports of police misuse, including cases where officers used Flock to search for an abortion patient and to identify gun show attendees. While Flock claims new partnerships are outpacing cancellations by roughly 10 to 1 and the company serves over 5,000 communities, the company has made policy concessions including reducing data retention from 30 days to seven days and prohibiting use for harassment or stalking. The backlash has expanded beyond privacy advocates to include political figures like Senator Josh Hawley and Governor Ron DeSantis, who are investigating the company's data practices and the potential for AI to transform license plate readers into mass surveillance tools.
Creepy Crawlies — Hacker News
Git.kernel.org is experiencing severe strain from AI training scrapers that render commits as HTML instead of using efficient cloning methods, consuming 14-16 of the site's 90 CPU cores (roughly 20% of total capacity) at any given time across five geographically distributed nodes. The scrapers, which now originate from millions of residential and mobile IP addresses to evade blocking, represent approximately 98% of the site's traffic, with legitimate developer requests comprising only about 2%. The site implemented Anubis proof-of-work challenges to deter bots, but scrapers have adapted by solving increasingly difficult mathematical puzzles, with 33% now successfully bypassing the current difficulty level 5 challenge. Git.kernel.org maintainers are responding by disabling features and gating anonymous access to reduce crawlable URLs and expensive operations, while noting that no simple solution exists as long as companies continue developing AI models requiring training data.
Hugging Face offers $399 robot duck to help you quack the AI code — The Register
Pollen Robotics, a Hugging Face subsidiary based in Bordeaux, France, announced the Microduck, a 25-centimeter-tall robot duck priced at $399 that runs on open-source software and is designed to teach users about AI, machine learning, and robotics through reinforcement learning. The robot ships with 1 GB of RAM and 32 GB of storage, features an articulated beak, front-facing camera, and LiDAR sensor, and can perform tasks like picking up objects and roller skating with optional accessories. Microduck is available for preorder with shipments planned before Christmas 2026, and Pollen is encouraging users to share trained skills once the system launches, though the hardware itself will not be open-sourced. The announcement comes amid unconfirmed reports that Nvidia has agreed to acquire Hugging Face for approximately $13 billion, though neither company has publicly commented on the deal.

Week ending 2026-08-17

2026-08-17T08:48:07Z

Anthropic's 'watermark' text adulteration in Claude is a perversion of writing — Hacker News
# Summary Anthropic announced that all Claude models worldwide will watermark their generated text to comply with EU regulations, but initially provided no explanation of how the system would work. The watermarking uses steganography—subtly biasing word choices toward a "green list" and away from a "red list" during text generation, creating statistical fingerprints detectable only by Anthropic with a secret key, though this approach necessarily compromises the quality and precision of the generated text. Critic John Gruber argues that this implementation adulterates Claude's output even in private conversations, makes the writing worse for users without providing them any benefit, and will be easily circumvented by bad actors while unfairly flagging legitimate uses of AI assistance.
Firefox is now the last major browser that still supports uBlock Origin — Hacker News
Firefox announced it will continue supporting uBlock Origin, positioning itself as the only major browser to maintain full compatibility with the ad-blocking extension as Microsoft Edge and other Chromium-based browsers phase it out under Manifest V3. Microsoft Edge, Opera, Brave, Vivaldi, and Samsung Browser are all following Google's lead in migrating from Manifest V2 to V3, which will prevent ad-blocking extensions from accessing the functions needed to identify and block ads. Safari and DuckDuckGo, the other major non-Chromium browsers, do not support uBlock Origin, leaving Firefox as the sole option for users wanting the full-featured extension without compromises. Users on other browsers will be limited to uBlock Origin Lite, which has reduced functionality, or their browser's built-in ad-blocking features.
Qwen 3.8 27B — Hacker News
I cannot summarize this as a news article because it is technical documentation, not news reporting. The content consists of installation instructions and model specifications for Qwen3.8-27B-FP8, a language model released by Alibaba's Qwen team. If you have an actual news article about this model release or another technology development you'd like summarized, please share that instead.
GLM-5.3: Frontier coding with emergent cyber capabilities — Hacker News
Why does Opus 5 feel worse to work with? — Hacker News
# Summary The author argues that Anthropic's Opus 5 model feels worse to work with than earlier versions like Opus 4.7, Opus 4.8, and Fable, despite having superior benchmark performance and capabilities. While Opus 5 is more capable overall, the author and colleagues report that it requires more "careful babysitting" because it makes assumptions without checking, doesn't ask clarifying questions when intent is unclear, and reinterprets plans without asking—behaviors that differ from the earlier models' tendency to seek clarification. The author speculates this results from Anthropic's focus on benchmark performance and self-improving AI development, which inherently selects for models that make bold assumptions under ambiguity rather than pause to ask questions, an approach that works well for self-contained benchmark tasks but poorly for real-world coding work where ambiguity and unclear context are inevitable.
Gemini 3.7 Flash — Hacker News
Google has released Gemini 3.7 Flash, an upgraded version of its Flash language model designed for coding, web development, and agent workflows, arriving just three weeks after Gemini 3.6 Flash. The new model shows significant performance improvements across software engineering tasks, achieving 43.6% accuracy on FrontierCode 1.1 Main compared to 3.6 Flash's 34.4%, and is priced at half the cost of its predecessor at $0.75 per million input tokens through the end of the year. Gemini 3.7 Flash is available to developers through Google's Gemini API and Android Studio, to enterprises through the Gemini Enterprise Agent Platform, and to individual users via Gemini Spark for Google AI Pro and Ultra subscribers. The model includes updated safety safeguards against misuse in chemical, biological, radiological, nuclear, and cyber offense domains.
France's top court blocks social media ban for under-15s — Hacker News
France's Constitutional Council struck down a law this week that would have banned children under 15 from accessing social media platforms like Facebook and X, ruling that the legislation infringed on freedom of expression and communication. The bill had passed through France's National Assembly and Senate with broad bipartisan support and would have also prohibited smartphone use in schools while exempting educational resources like Wikipedia and GitHub. President Emmanuel Macron, who championed the ban as protection against American tech companies and Chinese algorithms, now plans to revise and reintroduce the legislation before spring. Other European countries including Norway, Britain, and Greece are pursuing similar social media bans for minors, though Australia's recent attempt showed limited effectiveness, with studies indicating approximately 85% of banned under-16 users still accessing their accounts through various methods.

Week ending 2026-08-10

2026-08-10T09:32:55Z

New official 30th anniversary Quake mission pack adds new maps and mechanics — Ars Technica
Bethesda marked Quake’s 30
‘Humans will be a rounding error on the internet’ says Cloudflare exec — The Register
Cloudflare’s CFO Thomas Seifert warned that if current trends continue, machine‑generated traffic will outpace human traffic by up to a thousand‑fold in five years, driven largely by AI workloads and making humans a “rounding error” on the internet. The company’s Q2 results – $696 million in revenue (36 % YoY) and $205 million in losses – beat expectations,
Ransom Cartel ransomware creator sentenced to 16 years in prison — BleepingComputer
Maksim Silnikau, a 40‑year‑old Belarusian and former member of Russian‑speaking cybercrime forums, was sentenced to 16 years in U.S. federal prison after orchestrating the Ransom Cartel ransomware‑as‑a‑service operation that launched in December 2021 and attacked at least 18 companies worldwide from 2021‑2023, demanding over $5.2 million in ransoms and causing an estimated $6.7 million in losses. Using aliases
London still dominates Britain's datacenter map, but the regions are powering up — The Register
Smart glasses are only smart if we train them to be good. Then they'll be fantastic — The Register
The article argues that while AI’s human‑like façade is unsettling, the real concern is the way our minds increasingly mirror an outdated Windows system—crashing and losing compatibility without updates. It warns that augmented reality (AR) and smart glasses, though potentially life‑changing for
Linus Torvalds says AI has made 'huge' Linux kernel updates the new normal — The Register
Linus Torvalds has announced that the seventh release candidate for Linux 7.2, the largest he’s seen in years, will still go ahead despite his unease about its size
Mount Toba eruption doesn't seem like it could nearly kill our species — Ars Technica
Mount Toba’s 74,000‑year‑old eruption, the largest in the last 2.6 million years, spewed thousands of cubic kilometres of magma—about a thousand times more than the 1991 Pinatubo blast—and was once feared to have caused a planet‑wide “volcanic winter” that could have threatened early human survival.

Week ending 2026-08-03

2026-08-03T11:35:01Z

COLDCARD wallet RNG flaw likely linked to $88 million Bitcoin theft — BleepingComputer
Researchers discovered that a flaw in the COLDCARD hardware wallet’s firmware allowed attackers to exploit a deterministic random‑number generator, enabling offline seed reconstruction and the theft of about 1,367 BTC (≈ $88.6 million) from 4,585 compromised wallets; the first major wave of 1,083 BTC (≈ $70.2 million) was swept in a 41‑minute burst on July 30
AI Conquered Coding. Fast Food Is Next — WIRED
The ‘Guardrail Guy’ Went Viral for Posting About Flock Cameras. Then Someone Destroyed Them — WIRED
Steve “The Guardrail Guy” Eimers, who built a YouTube, Tik
UK government investment arm cops to 40-hour leak of officials' contact details — The Register
UK Government Investments (UKGI) admitted that a staff member left an internal file—containing the names and work‑email addresses of 51 high‑level government officials—publicly accessible for roughly 40 hours during the 2025‑26 financial year, a lapse that violated its own security policies. The breach was voluntarily reported to the Information Commissioner’s Office and to UKGI’s Audit and Risk Committee, triggering an external review that found the response “appropriate” but called for
Alienware 27 QD-OLED (AW2726DM) Review: A $350 Winner — WIRED
The Alienware AW2726DM is a 27‑inch 1440p QD‑OLED monitor that offers stunning contrast and color—reaching 98 % DCI‑P3 and 94 % Adobe RGB—thanks to self‑emissive pixels and quantum‑dot technology, but its pixel‑arrangement quirks can blur small text and cause color fringing. With a modest 110 ppi and a 200‑nit peak brightness,
People Are Ghosting Long-Term Partners. Some Don’t Regret It — WIRED
The Best Audio Players for Kids: Yoto, Toniebox, and More — WIRED
Parents are increasingly turning to audio players as a screen‑free alternative that balances independent play and

Week ending 2026-07-27

2026-07-27T11:34:14Z

US prosecutors charge Atlanta man after GrapheneOS phone wipes itself during airport search — TechSpot
A federal case in Atlanta has sparked a rare legal debate over privacy‑focused mobile software when prosecutors charged Sam Tunick with destroying property to thwart evidence after he used GrapheneOS—a security‑only operating system for Google Pixel phones—to wipe his device. Tunick, who was stopped at Hartsfield‑Jackson International Airport on January 24 after an internal DOJ “suspected terrorism” alert tied to his anti‑Cop City activism, was interrogated
The first thing I do for friends switching from iPhone to Android is change this setting — Android Police
Switching from iOS to Android feels jarring mainly because the thumb’s muscle memory is built around iOS conventions—swiping in from the left edge, hitting a back button in the upper‑left corner,
Google Turns a Selfie Video Into Your Account’s Spare Key — WIRED
Google has rolled out a new “selfie video” sign‑in option that lets users unlock their accounts with a short facial video, adding a backup key for situations where device‑based passkeys are inaccessible. Users enable it via the Security & sign‑in tab, record a brief clip following lighting and “liveness” guidelines
The EU Fines Google $1 Billion for Prioritizing Its Own Services in Search — WIRED
The European Commission has slapped Google with a $1 billion fine for abusing its dominance in EU search and app‑store markets under the Digital Markets Act, ordering the company to stop giving its own services (shopping, travel
Artist sues AI meme generator for selling deeply personal comic as ad template — Ars Technica
Artist Elmer Saflor—known online as “Superelmer” and creator of the viral “Running Away Balloon” comic—has filed a lawsuit against Memes Apps, LCC, the
This Is Donald Trump’s AI Brain Trust — WIRED
The
First teaser for Apple TV's Neuromancer debuts at SDCC — Ars Technica
Apple TV is rapidly overtaking HBO as the premier destination for high‑quality dramas—especially sci‑fi—thanks to titles such as *Slow Horses*, *Severance*, *Foundation*, *For All Mankind*, *Murderbot*, *Silo*, *Pluribus*, and *Widow’s Bay*, with the highly‑

Week ending 2026-07-20

2026-07-20T11:00:52Z

San Francisco orders Apple, Google to remove nudify apps from app stores — Ars Technica
San Francisco Attorney General David Chiu has sent cease‑and‑desist letters to Apple and Google demanding the removal of 13 “nudification” apps that can turn ordinary photos into non‑consensual deepfake porn, citing California laws prohibiting such services. The apps
HollowByte DDoS flaw bloats OpenSSL server memory with 11-byte payload — BleepingComputer
A recently discovered “HollowByte” flaw lets attackers trigger a denial‑of‑service on vulnerable OpenSSL servers with an 11‑byte payload that tricks the library into pre‑allocating large buffers based on a forged message length header—without waiting for the hanyar. By repeatedly opening TLS connections with random, exaggerated size claims, a malicious actor can exhaust server memory, fragment the heap, and leave the process
China delivers a one-two punch to America’s AI dominance — The Verge
New Free Speech Concern: When AI Chatbots Won't Criticize Leaders from Repressive Regimes — Slashdot
Parkinson’s Law, articulated by Cyril Northcote Parkinson in a 1955 *Economist* essay, observes that “work expands to fill the time available.” He argues that when deadlines are set generously, tasks—whether trivial or substantial—tend to stretch and consume all allotted time, often leading to inefficiency and wasted resources. Parkinson attributes this phenomenon to bureaucratic inertia,
As AI Transforms Silicon Valley, Some Tech Workers Face Evaporating Financial Security — Slashdot
Cyril Northcote Parkinson’s 1955 piece in *The Economist* introduces what is now known as
India's first privately-developed rocket reaches orbit on dramatic debut launch — Ars Technica
Indian space officials hailed the successful debut of Skyroot Aerospace’s Vikram‑1, India’s first fully commercial satellite launcher, after it reached a 280‑mile (450 km) orbit at a 60° inclination. Launched from Sriharikota Island at 1:35 a.m. EDT, Vikram‑1—standing 22 m tall and capable of lofting up to 770 lb (350 kg) payloads—fueled three solid‑stage boosters followed by a liquid‑fuel fourth
As mosquito ranges expand, better monitoring is key to preventing disease — Ars Technica

Week ending 2026-07-13

2026-07-13T11:21:52Z

AI-driven datacenter builds drive Microsoft's emissions up a quarter in one year — The Register
Microsoft’s 2026 Environmental Sustainability Report highlights a paradox: while it achieved 100 % renewable power for its global electricity use last year, its datacenter expansion has pushed greenhouse‑gas emissions up 25 % to 20 million tCO₂e in 2025 (projected 34 million tCO₂e without current cuts). The company’s goal to be carbon‑negative by 2030 remains unfulfilled, though it argues AI’s societal benefits justify the energy demand. Scope 3 emissions still dominate its footprint, with Scope 2 rising to 13 % of total due to purchased power. Water use is being curbed—WUE improved 25 % since 2022, and Microsoft claims it replenished more water than it withdrew in 2025. Circular initiatives reuse 92 % of decommissioned servers, and construction experiments include low‑carbon concrete and wood‑based facilities to reduce embodied carbon. The report frames AI infrastructure as both an environmental challenge and an opportunity for leadership.
OpenAI’s Head of Safety Is Leaving the Company — WIRED
OpenAI’s head of safety systems Johannes Heidecke has announced his departure, a move that follows a reorganization aimed at tighter integration of the company’s safety and research functions. Chief research officer Mark Chen said safety teams will now report to VP of research and alignment Mia Glaese, who will expand her role to include safety, while Saachi Jain will serve as interim head of safety systems. Chen cited accelerating model training and shortened release cycles as reasons for increased safety coordination needs. Heidecke, who joined in 2021 and led safety systems since 2024, is leaving amid a broader exodus of safety leaders—including former chief futurist Joshua Achiam—and as OpenAI rolls out increasingly capable models like GPT‑5.6 that have shown misaligned behavior. The reshuffle comes as the company continues to refine its product strategy, with Greg Brockman taking on go‑to‑market duties after CEO of AGI deployment Fidji Simo steps down.
Photovoltaics are still running after a year under Swiss trains — The Register
iGarden Swim Jet X Pro 10 Review: Endless Pool — WIRED
Python Is So Slow. Can Julia Solve the Two-Language Problem? — WIRED
The article contrasts the often‑pedestrian nature of award acceptance speeches with the inspirational Turing Award lectures that have shaped computing, citing examples such as Backus, Thompson, and Dijkstra. It then focuses on Kenneth Iverson’s 1979 keynote on notation, explaining how APL merged mathematical and programming notations to make complex operations more concise, thereby influencing scientific computing. Fast forward to 2026, the piece highlights the persistent “two‑language problem”: researchers prototype in slow, high‑level Python but must rewrite critical sections in faster, lower‑level languages like C++ or Rust, a dilemma that can’t be solved by AI alone. Drawing an analogy to construction materials, the author wonders if a language could be as approachable as Python yet as performant as C. In 2012, four mathematically‑oriented scientists launched Julia to fill this niche, aiming for an open‑source, easy‑to‑learn language that satisfies both casual users and hardcore developers, thereby uniting the fragmented ecosystem of Matlab, Lisp, Python, Ruby, and Perl.
White Supremacists Love Armie Hammer’s New Movie — WIRED
The article describes the controversial film **“Citizen Vigilante,”** directed by Uwe Boll and starring Armie Hammer as Sanders, a former U.S. soldier turned European real‑estate magnate who launches a violent purge against Muslims, migrants, and anyone he deems complicit in a supposed “great replacement.” The film, which propagates the baseless conspiracy that Muslims will overrun Europe in 30 years, was heavily criticized for its racist, xenophobic, and alt‑right propaganda content. It was banned in Germany for inciting violence, yet gained massive attention when Elon Musk promoted it on X, sharing the entire movie and praising its “moderate response” scenes. The film’s release on streaming platforms and online screenings by far‑right groups amplified its reach, prompting concerns from hate‑prevention advocates that Musk’s endorsement normalizes extremist violence and hate.
A Leak of San Francisco Police Drone Footage Exposes the New Reality of Urban Surveillance — WIRED
In mid‑June, a Skydio X10 drone followed a suspect through San Francisco, livestreaming the chase, arrests, and searches to a public web page that the police never intended to expose. Security researchers Sam Curry and Maik Robert discovered that the San Francisco Police

Week ending 2026-07-06

2026-07-06T12:12:31Z

C programmers commit fresh crimes against readability — The Register
The 2025 International Obfuscated C Code Contest (IOCCC29) returned after a four‑year break, awarding 23 winning entries across 14 quirky categories, including three “hat‑trick” winners (Yusuke Endoh, Nick Craig‑Wood, and Don Yang). Highlights include Adrian Cable’s 366‑byte Subleq emulator that runs a full‑color Mandelbrot set, Pong, and even a Linux port on a tiny virtual architecture designed for software preservation; Craig‑Wood’s 66‑line Game Boy emulator and a FRACTRAN‑based Commodore 64 emulator
Companies that add more AI also add more people — The Register
NASA launched an emergency mission to stop the Swift Observatory from crashing to Earth — The Verge
NASA has called on Katalyst Space Technologies to save the Swift Observatory—launched in 2004 to study gamma
Inside the Luddite festival harnessing Gen Z’s rage against Big Tech — Ars Technica
Google loses long-running appeal of record EU fine, will have to cough up $4.7 billion — Ars Technica
Google’s record‑setting €4.34 billion fine for abusing its Android
Bentley teases its first EV, the Torcal — Ars Technica
Bentley is set to launch its fourth model, the Torcal, which marks the marque’s first all‑electric vehicle. The car has already been tested in the Arctic Circle and at the Nürburgring, while trademark filings of “Torcal” in Europe and the UK—
The Czinger 21C might be the wildest car we drive all year — Ars Technica

Week ending 2026-06-29

2026-06-29T12:42:27Z

NYT slams Microsoft for building copyright-infringing supercomputer for OpenAI — Ars Technica
The New York Times filed a heavily redacted motion to amend its copyright suit against OpenAI and Microsoft, tightening a contributory‑infringement claim in light of a Supreme Court ruling that now requires plaintiffs to prove intentional inducement of illegal conduct. The Times alleges that Microsoft built a world‑class supercomputing system to entice OpenAI to steal its copyrighted content, and it seeks to drop two other claims (contributory infringement and trademark dilution) while keeping its core accusation that Microsoft and OpenAI pirated millions of NYT works to compete and profit. Microsoft dismisses the amendment as a “last‑ditch” effort, whereas the Times argues the changes do not prejudice defendants or delay the case, and that the amended claim aligns with new legal precedent and fresh discovery evidence.
Why did this journal retract two 1940s papers by Max Planck? — Ars Technica
China Defies US Restrictions and Builds the World’s Fastest Supercomputer — WIRED
China’s newly installed LineShine supercomputer at the National Super
Trump-shuttered climate change site back online in nonprofit hands — The Register
In July 2025, a Trump‑issued executive order “prioritizing gold‑standard science” forced NOAA to shut down its Climate.gov website, purge its 15‑year archive of climate data and remove staff, sparking alarm that the public and scientists would lose critical climate information. Former Climate.gov program manager Rebecca L
FBI: Russian hackers now target Signal backup recovery keys — BleepingComputer
Yeasound RIC800 Hearing Aids Review: Good Audio, Glitchy App — WIRED
Yeasound, a subsidiary of telecom maker Yealink, is reviving the behind‑the‑ear hearing‑aid niche with its RIC800 (and RIC700) models. The RIC800 packs AI‑powered noise reduction, an automatic speech‑focus system, Android
The war against ‘woke’ could end US science as we know it — The Verge
The Office of Management and Budget’s 412‑page proposal threatens to turn the U.S. federal grant system into a politically‑driven tool by requiring that every $1 trillion in research funding be vetted by political appointees to ensure alignment with the President’s “policy priorities”; it would bar funding for DEI, gender‑identity research, and any work that could be seen as “

Week ending 2026-06-22

2026-06-22T13:45:13Z

Polymarket reportedly paid people to post fake videos of themselves placing bets — The Verge
Hackers exploit info disclosure bug in Gravity SMTP WordPress plugin — BleepingComputer
WordPress sites using the Gravity SMTP plugin (≈100,000 sites) are at risk from CVE‑2026‑4020, a medium‑severity, unauthenticated information‑leak flaw that exposes a REST endpoint returning a full “System Report.” The report can reveal API keys, third‑party email credentials, WordPress configuration, server and database details, and other sensitive data, enabling attackers to impersonate the site and plan further exploits. Defiant’s Wordfence firewall has blocked over 17 million attack attempts, with a spike of 4 million blocked requests on June 7, and recommends blocking the IPs listed in their advisory and the `/wp-json/gravitysmtp/v1/tests/mock-data` endpoint. The vulnerability is fixed in Gravity SMTP 2.1.5 (released March 17); site owners should upgrade immediately. In a related alert, CVE‑2026‑8713 affects the Avada Builder plugin, allowing unauthenticated arbitrary file deletion via a path‑traversal flaw when a published Avada form stores submissions in the database; the issue is patched in Avada 3.15.4 and should be applied promptly, even though no active exploitation has been seen.
Microsoft accidentally kills epic Outlook email threads — The Register
Microsoft’s latest build of Outlook for macOS (v16.110, build 26061317) has broken a fundamental feature: when users reply to an email, the original message no longer appears in the reply window. This omission hampers threaded conversations and has already prompted complaints that the update slipped past quality checks. The only workaround, per a forum moderator, is to roll back to a previous version and disable auto‑updates until a fix arrives—an option that is impractical for managed fleets. While users can copy‑paste the relevant text, the disruption reveals a lapse in Microsoft’s testing process, especially troubling given the company’s recent push to phase out Office 2019 on Macs and its broader strategy of encouraging upgrades. The Register has yet to receive a response from Microsoft regarding the issue.
Several US States Bet That AI Can Solve Their Prison Recidivism Crisis — Slashdot
The quote humorously notes that the University of California, Berkeley produced two iconic and seemingly unrelated innovations—LSD, the psychedelic drug, and UNIX, the influential operating system—implying a surprising but deliberate connection between them.
Trump admin’s coal investments assist plants with repeated violations — Ars Technica
In 2023, Tennessee’s Cumberland Fossil Plant—already plagued by pollution, equipment failures and health risks—was slated to shut its units in 2026 and 2028 after a 2011 settlement that required the Tennessee Valley Authority (TVA) to install pollution controls. Yet, following a Trump‑era reshuffle of TVA’s board, the agency secured a $46 million federal grant to extend the plant’s life—one of only three DOE‑funded coal plants repeatedly cited for Clean Air and Water Act violations, alongside Oklahoma’s Grand River Energy Center and North Carolina’s Roxboro Steam Electric Plant. Local residents like Angie Mummaw view the extension as a setback to clean‑energy progress, while clean‑energy advocates such as Maggie Shober argue that keeping coal plants running accelerates climate change and health harms. Studies link the plant’s toxic fine‑particle emissions to roughly 1,000 premature deaths across the Northeast from 1999‑2020, underscoring the broader public‑health impact of continued coal operations.
These Are the 12 Ikea Products the Company's Design Chief Personally Owns — WIRED
Johan Ejdemo, a carpenter and avid cook, mixes Ikea staples with his own carpentry to create a highly functional yet personalized kitchen. He keeps a long‑used Koncis metal tray set for cooking, customises Ikea’s Faktum/Metod system with extra storage and a hand‑painted island, and adds a vintage PS cabinet and a 1995 PS clock for nostalgic flair. His space is rounded out with a now‑unavailable Norbo folding table, a weathered Bekväm step stool, and practical Norberg‑style stools, all of which accommodate family gatherings and everyday use. The result is a durable, well‑organized kitchen that blends Ikea’s affordability and design with Ejdemo’s hands‑on craftsmanship.
The UK Just Lost Its Sixth Prime Minister of the Decade — WIRED
UK Prime Minister Keir Starmer resigned less than two years after leading Labour to a landslide victory, citing his party’s assessment that he is not the best fit to guide them into the next election. His popularity has plummeted amid tax rises, spending cuts, and unmet public‑service promises, with May’s local elections seeing Labour lose 1,500 council seats and many councils shift to the Reform UK party. Starmer’s resignation comes as Andy Burnham, newly elected MP and former Manchester mayor, is poised to succeed him, with former Health Secretary Wes Streeting pledging support. The resignation follows a series of high‑profile leadership changes since the 2016 Brexit referendum, and Starmer will remain in office until the Labour Party elects a new leader, expected before Parliament reconvenes in September.

Week ending 2026-06-15

2026-06-15T14:32:31Z

Google Earth’s flight simulator is now available in your browser — The Verge
Google Earth has quietly featured a flight‑simulator mode since 2007, and on Friday the company made it publicly available via the web interface. After opening Google Earth in a browser, users click “Explore Earth” and then select “Flight Simulator” from the Tools menu. Switching the basemap to Satellite gives a photorealistic view, and the controls—accessed through keyboard or mouse—are simple but require practice to navigate without crashing; if a collision occurs the simulator pauses and offers a reset to a safe altitude. While not as sophisticated as dedicated flight games like Microsoft Flight Simulator, the feature lets anyone pilot a virtual aircraft over the globe without downloading extra software.
The UK Places a Sweeping Ban on Social Media for Kids Under 16 — WIRED
Prime Minister Keir Starmer has announced a sweeping ban that will prohibit children under 16 from accessing major social‑media platforms—Facebook, Instagram, X, TikTok, Snapchat and YouTube—effective spring 2027, while retaining WhatsApp and Signal for messaging. The ban also raises the minimum age for romantic chatbot interactions to 18, removes livestreaming and stranger‑contact features for under‑16s, and may introduce an overnight curfew for those under 18. The move, framed as protection against graphic content, bullying and “doomscrolling,” follows a public consultation that attracted over 100,000 submissions and a 90 % parental endorsement for a full ban; it comes after Australia enacted a similar restriction last year and amid political pressure on Starmer’s leadership. Tech companies have cautioned that blanket bans could push young users toward less supervised platforms, while supporters, including parents of a transgender victim, argue the harms outweigh the benefits.
Fox is buying Roku — The Verge
Fox has agreed to acquire Roku for $22 billion, a move that will combine Fox’s TV networks and Tubi streamer with Roku’s hardware, software and The Roku Channel, positioning the merged entity as the U.S. TV industry’s third‑largest player by viewing share. CEO Lachlan Murdoch framed the deal as a “defining moment” that will bring Fox’s live content into the leading streaming platform, while Roku founder Anthony Wood emphasized that the platform will remain open and partner‑friendly, simply adding Fox properties—especially Fox Sports—to its home‑screen promotion. The transaction is slated to close in the first half of 2027, subject to U.S. regulatory approval and limited international clearance.
US Gov asks Anthropic to ban 'foreign national' access to Fable, Mythos — BleepingComputer
Anthropic has halted access to its flagship models Fable 5 and Mythos 5 worldwide after a U.S. export‑control directive barred all foreign nationals—including its own foreign‑national employees—from using them, citing national‑security concerns over a reported narrow jailbreak. The stop came just days after the models were rolled out for free Pro, Max, and Enterprise customers, leaving millions of users offline and forcing a fallback to older models such as Claude Opus 4.8. The move has drawn criticism from UK AI officials and industry observers, who see it as a blow to technological sovereignty and a precedent that could stall future AI deployments. Anthropic disputes the government’s rationale, claims the issue is minor, and says it is working to restore access while offering more details within 24 hours.
PRC-linked spies hid inside medical and military networks for more than a year, snooping through Gmail and stealing data — The Register
Google’s Threat Intelligence Group exposed a year‑long espionage campaign by a PRC‑linked threat group, dubbed UNC6508, that infiltrated North American medical and military research organizations through compromised REDCap servers. After planting custom malware called InfiniteRed, the attackers harvested REDCap login credentials, maintained persistence via upgrade injections, and ultimately accessed internal networks and admin accounts. They leveraged Google Workspace’s content‑compliance rules, creating a “Patroit” rule that silently BCC‑forwarded targeted emails—ranging from drone and unmanned vehicle technology to the mosquito‑borne Chikungunya virus—to an attacker‑controlled Gmail address (BebitaBarefoot774). The operation spanned defense, technology, and medical research sectors, targeting universities, hospitals, and military health institutions, and underscored how state‑sponsored actors use seemingly unrelated data‑search queries to cast a broad net across national‑security‑relevant domains.
UK AI hiring surges as firms seek people to babysit the bots — The Register
PwC’s latest AI Jobs Barometer shows a sharp 61 % rise in UK AI specialist hiring (from 112,000 to 180,000 roles) even as overall vacancies fell 6.6 %. However, the growth is largely driven by “AI user” roles that help existing professionals leverage AI—up 66,000 positions—while pure “AI developer” roles grew only 2,600. This creates a two‑track labor market: high‑skill jobs where AI adds value to specialised work are expanding 39 % since 2018, versus 17 % growth in roles where AI merely simplifies tasks. Wage premiums for AI‑skilled roles have surged to an average 34.2 % above non‑AI equivalents (up from 11 % last year), with consumer firms paying the most and the public sector the least. The findings underscore that the current AI hiring boom is centred on seasoned professionals who can effectively use AI, rather than a flood of new tech builders, amid broader concerns about AI’s impact on employment and workplace productivity.
Russia appears set to finally address long-term, serious space station cracks — Ars Technica
In mid‑June 2024, NASA and Roscosmos faced a serious depressurization risk on the International Space Station when persistent cracks in the Russian Zvezda service module’s PrK module began leaking again after earlier repairs. The leaks, first noticed in 2019 and normally sealed with Germetall‑1, had worsened in early June, prompting Roscosmos to initiate a comprehensive inspection and repair on June 5. The proposed solution alarmed NASA, which ordered its Crew‑12 astronauts to seek temporary refuge in a Crew Dragon capsule for emergency protection. After Russia revisited its inspection plans and NASA backed the decision, the crew ended the emergency precautions and resumed normal operations aboard the ISS.

Week ending 2026-06-08

2026-06-08T12:41:40Z

How to Watch Apple’s WWDC 2026 Keynote—and What to Expect — WIRED
Apple’s final Worldwide Developers Conference under CEO Tim Cook, set to launch on June 8, will spotlight new software releases—iOS 27, macOS 27, watchOS 27, and more—emphasizing performance, stability, and a suite of AI-powered “Apple Intelligence” features such as digital event passes and receipt‑based bill splitting; the keynote is expected to unveil a major Siri overhaul, now powered by a multi‑year partnership with Google’s Gemini, promising a more personalized assistant.
Critical Everest Forms Pro flaw exploited to take over WordPress sites — BleepingComputer
Hackers are actively exploiting CVE‑2026‑3300, a critical vulnerability in Everest Forms Pro (v 1.9.12 and earlier) that lets an unauthenticated attacker inject arbitrary PHP code via the plugin’s Complex Calculation feature. The flaw arises because user‑supplied values are inserted into a PHP string and executed with `eval()`, and the sanitization used (`sanitize_text_field()`) fails to escape single quotes, enabling attackers to close the string and inject code such as `wp_insert_user()` to create rogue administrator accounts. Wordfence reports show active exploitation began on April 13, with over 29,300 blocked attempts and two main source IPs (202.56.2.126 and 209.146.60.26). The issue was reported in February, patched on March 18, but administrators should still audit logs and accounts—especially for the username “diksimarina”—and block the identified IPs to mitigate the risk.
The weather and climate science AI revolution isn’t revolutionary — Ars Technica
The article explains that AI is increasingly being used in weather and climate science, but not as flashy large‑language models (LLMs) like chatbots; instead it relies on established machine learning (ML) techniques that identify complex patterns in vast datasets. While a National Weather Service office once posted a misleading AI‑generated map, the real applications involve training neural networks to recognise relationships in environmental data, improving forecasting and climate modeling. The piece clarifies that ML, not LLMs, underpins these advances and that meteorologists and climate scientists are still doing the heavy lifting, using well‑understood algorithms rather than replacing them with generic AI assistants.
Momfluencers Are Pitching AI as a Better ‘Coparent’ Than Men — WIRED
Lilian Schmidt, a Zurich‑based brand consultant and mother, turned to ChatGPT after conventional sleep‑tips failed to help her 3½‑year‑old daughter fall asleep, discovering that the AI’s counterintuitive advice—more stimulation before bed—worked instantly; this success sparked a viral TikTok series and the creation of a paid custom GPT called “Coparent,” positioning Schmidt as a “momfluencer” who promotes generative AI as a substitute for traditional parenting labor. Her content, which highlights how AI can reduce the invisible workload that 13.5 extra hours per week of chores and childcare places on U.S. mothers, mirrors broader gendered labor disparities, with women still shouldering the majority of household responsibilities despite fathers’ increasing involvement. Yet a 2025 study shows women are 20 % less likely than men to use generative AI, a gap attributed to the tools’ “pale, male, and stale” design, underscoring both the potential and the challenges of AI‑assisted parenting.
UK boffin bait lands 18 international researchers — The Register
RIP Anthony Head: Our 10 favorite moments of Buffy's Giles — Ars Technica
The article announces the death of actor Anthony Head (72), best known for playing Rupert Giles on *Buffy the Vampire Slayer*, and notes how fans and colleagues paid tribute online. It highlights Head’s later roles—Uther Pendragon in *Merlin*, the Prime Minister in *Little Britain*, a headmaster in *Doctor Who*, and Rupert Mannion in *Ted Lasso*—but affirms that Giles remains his defining part, with even a planned but unmade spin‑off. The piece underscores Head’s subtle, essential presence in the show’s storytelling, and invites readers to rewatch *Buffy* in his honor, offering a list of favorite Giles moments (the article begins detailing the first, “Lie To Me” episode).
Lenovo IdeaPad Slim 5x Review: The Best Laptop Under $1,000 — WIRED
Lenovo’s IdeaPad Slim 5x proves that a budget laptop can feel premium without compromising too much on usability. Its 15.3‑inch screen, centered touchpad and spacious keyboard keep it travel‑ready while matching the size and thickness of the 15‑inch MacBook Air. The touchpad outperforms many rivals in responsiveness, though its click is noticeably loud, and the speakers, while not as full as Apple’s, outperform average budget units. Powered by the Qualcomm Snapdragon X2 Plus, the machine delivers solid CPU performance for multitasking, though it still lags behind Apple’s M5 in raw power. Still, for typical office workflows, streaming, and light multitasking the X2 Plus offers more than enough horsepower. The bright, colorful, anti‑glare touchscreen adds a premium feel, making the Slim 5x a compelling choice for users who value portability, performance, and a decent display without a hefty price tag.

Week ending 2026-06-01

2026-06-01T13:46:30Z

LLMs are closer to religion than they appear. Watch out for those who like it that way — The Register
The article contrasts two starkly different reactions to AI’s influence on faith: Pope Leo XIV’s sweeping Latin encyclical “Magnifica Humanitas,” which condemns AI for risking human dignity, and a multi‑faith study claiming that AIs never provide religious answers. The papal critique highlights a human‑rights and political dimension—suggesting that AI could erode dignity and even be exempted in the workplace on religious grounds—while the study’s methodology and assumptions (treating “religion” largely as fundamentalist Christianity and ignoring scientific consensus) reveal the risk of politicizing AI to advance particular religious narratives. Both cases underscore how AI’s unregulated evolution can echo the schismatic power of religion: it can become a tool of ideological proselytization or a target for ethical and political opposition, showing that the intersection of technology and faith is already a battleground for shaping society’s future.
AI agents get their own phone directory built atop DNS — The Register
DNS‑AID is an open‑source project that lets AI agents discover one another through the existing Domain Name System, eliminating the need for custom registries or hard‑coded ports. Built on DNS features such as SVCB (and TXT fallback) plus optional DNSSEC, DANE TLSA, and JWS signatures, it enables agents to publish and locate services by name, function, or domain via records like _{agent‑name}._{protocol}._agents.{domain}. The Linux Foundation will govern the project, which was originally developed by Infoblox, and is already supported by major DNS providers (AWS, Cloudflare, Azure, Google, etc.). Developers can get started by installing the dns‑aid tool, initializing a zone, publishing an SVCB record, and optionally signing with DNSSEC. By leveraging DNS’s proven global scalability and ubiquity, DNS‑AID offers a “web‑native” routing layer that could underpin a future $3‑$5 trillion agent‑to‑agent economy.
This is the Microsoft Surface Laptop Ultra with Nvidia RTX Spark — The Verge
Microsoft and Nvidia have announced the Surface Laptop Ultra, a 15‑inch Windows 11 laptop that will house Nvidia’s new RTX Spark “superchip.” The chip, originally sold in the DGX Spark mini‑PC, offers up to 20 CPU cores, 6,144 GPU cores, and 128 GB of unified memory, with lower‑spec models starting at 16 GB, and is expected to deliver all‑day battery life, RTX 5070‑level graphics, and up to 1 petaflop of AI compute. The Ultra will feature a 15‑inch mini‑LED panel with 262 ppi, 2,000 nits peak HDR brightness, a large haptic trackpad, and a range of ports (USB‑C, USB‑A, HDMI, SD card slot, headphone jack). Microsoft touts it as the most powerful Surface ever, with a weight under 4.5 lb and a price that has yet to be disclosed. The partnership builds on years of collaboration to bring Windows to Arm‑based devices and to optimise the RTX Spark for Windows 11.
Blue Origin Rocket Explodes in Fiery Setback — WIRED
Blue Origin’s 98‑meter New Glenn rocket blew up during a hot‑fire (static‑fire) test on Cape Canaveral’s Launch Complex 36, igniting the launchpad at night; founder Jeff Bezos confirmed all personnel were safe and that the company was already investigating the cause. The test, meant to verify engine performance before an actual flight, was the fourth planned New Glenn mission that would have carried 48 satellites for Amazon’s Leo network. NASA’s administrator, Jared Isaacman, acknowledged the anomaly, pledged a thorough investigation, and noted potential, though still‑unknown, impacts on Artemis‑related lunar missions. The incident follows a prior launch failure in April that prompted an FAA probe, highlighting the challenges of developing heavy‑lift launch capability.
An OpenAI model solved a famous math problem that stumped humans for 80 years — Ars Technica
In May, OpenAI revealed that an internal AI model had disproved the long‑standing Erdős unit distance conjecture, a milestone that earned praise from top mathematicians such as Fields Medalist Tim Gowers and University of Toronto professor Daniel Litt; the result marks the first autonomous AI‑generated proof of a major open conjecture, yet it largely repackaged existing ideas rather than inventing new techniques, reinforcing a near‑term future where AI’s encyclopedic recall and brute‑force reasoning complement human insight and creativity in mathematical research.
Best Sleep Trackers of 2026: Oura, Whoop, and Eight Sleep — WIRED
Apple’s own smartwatch remains the most affordable option: the Series 11 ($399) can track four sleep stages using its heart‑rate sensor and accelerometer, and newer models add blood‑oxygen and temperature readings, but it offers no overall sleep score. Garmin’s premium Epix Pro ($1,000) delivers advanced sleep monitoring—stages, SpO₂, respiratory rate and restlessness—but tends to over‑estimate sleep duration and ignores naps in its score. Google’s Nest Hub Max ($229) uses radar and a microphone to monitor sleep without a wearable, yet it consistently overestimates REM and misses wake periods, while the cheaper second‑gen model ($100) offers similar performance. Headband contenders like Muse S Athena ($475) provide EEG‑style brain data plus heart‑rate and respiration, yet can be uncomfortable and prone to failure if removed during the night. Withings’ ScanWatch 2 ($370) offers a full four‑stage profile with respiratory‑rate PPG and additional metrics, though its bulk may disturb sleep and it can misclassify quiet wakefulness. The company’s Sleep Tracking Mat ($200) sits under the mattress and gathers motion, breathing, heart‑rate and snoring data, but its assumption that stillness equals sleep can skew scores and it needs a power outlet. Finally, the low‑cost Sleep Routine app ($60 / yr) gives nightly stage reports with reasonable accuracy but is occasionally buggy. Each device trades off price, comfort and data depth, so the best choice depends on whether you prioritize convenience, comprehensive metrics or affordability.
Universal Audio Volt 876 USB Audio Interface Review: Pro-Level Polish — WIRED
In 2006, the author’s attempt to jump into IDM music using a new M‑Audio FastTrack Pro was thwarted by driver headaches and high latency, forcing a switch to a Line 6 DL4 pedal and a return to live band playing; had a plug‑and‑play interface like today’s Universal Audio Volt existed, his entry into computer‑based music would have been smoother. Today, audio interfaces have dramatically improved: prices are lower, quality higher, and latency negligible, especially with class‑compliant designs that work out of the box on Windows, macOS, iOS, and Android. As a result, anyone can buy a $200‑class interface, connect a mic, and produce decent recordings for platforms like TikTok without the technical barriers that once plagued early adopters.

Week ending 2026-05-25

2026-05-25T12:00:49Z

A Fundamental Principle of Aeronautical Engineering Has Been Overturned — WIRED
Scammers Are Abusing an Internal Microsoft Account to Send Spam Links — Slashdot
In “Make it myself? But I'm a physical organic chemist!” the author argues that a significant portion of the reagents and materials routinely used in physical organic research can be prepared in‑house rather than purchased, thereby saving time, money, and reducing environmental impact. Drawing on examples from kinetic and mechanistic studies—such as custom‑synthesized probes, isotopically labeled substrates, and high‑purity solvents—the piece illustrates how standard procedures (e.g., recrystallization, distillation, chromatography) can be adapted for laboratory‑scale production. It also addresses safety, quality control, and documentation requirements, emphasizing that meticulous characterization (NMR, MS, IR) and proper labeling are essential to maintain reproducibility. Overall, the article encourages chemists to view reagent synthesis as an integral part of experimental design, fostering a more sustainable and cost‑effective research workflow.
Building Pi with Pi — Hacker News (hnrss)
The article reflects on using Pi, an AI‑powered issue tracker and development tool, to build and debug Pi itself, highlighting the “slop” that arises when LLMs rewrite issue descriptions and code, producing vague, over‑hypothesised, or incorrect diagnostics that mislead human maintainers; this sloppiness forces developers to insist on concise, fact‑based issue reports and to actively verify AI analysis rather than trust it, and it also leads to over‑engineering fixes that add unnecessary complexity. The author notes the high volume of AI‑generated issues and pull requests that are auto‑closed, reopened, or closed again, stressing that the problem lies more with contributors and the tools they use than with GitHub itself. Parallel Pi sessions can work on multiple issues simultaneously, but the overall ecosystem remains overwhelmed by the sheer amount of code and projects in a post‑AI world, underscoring the need for stronger community collaboration and a focus on fixing upstream bugs rather than creating local workarounds.
The AI Era Is Creating a Bug Hunting Arms Race — WIRED
A decade of bug‑bounty programs—once a reluctant novelty—has been upended by the rise of agentic AI that can autonomously hunt software vulnerabilities and craft exploits, flooding disclosure portals and inflating the cost of payouts. Apple’s rewards have leapt from $200k to $2m, and firms like Google expect to spend many times more next year, while smaller companies may struggle to keep pace. The surge in low‑ and medium‑severity findings is already shortening the 90‑day responsible‑disclosure window that was built for slower, rarer bug‑finders, and is prompting organizations to speed patch releases to avoid being outmaneuvered by attackers who are now using the same AI tools. This new equilibrium threatens to compress discovery and exploitation timelines, making rapid, careful patch deployment—despite its operational risks—an urgent imperative in a landscape where even “non‑professional” actors can deploy AI‑generated zero‑day exploits.
Digital sovereignty, the musical: One engineer’s bizarre crusade against hyperscalers — The Register
A French infrastructure architect, Amine Raiti, has launched “Operation Dindon,” a tongue‑in‑cheek, multilingual protest against Amazon, Google and Microsoft that threatens to unleash AI‑generated sea shanties, K‑pop, Finnish polka and satirical poetry if the hyperscalers do not address cloud lock‑in and egress fees by September. Raiti’s “Iron Ultimatum” demands that providers let customers cancel long‑term contracts, waive costly data‑transfer charges and make exit from proprietary services financially viable; otherwise his 50‑track AI‑produced campaign will continue. The campaign, which began with a 14‑part satirical series about a fictional turkey trapped in the cloud, aims to turn vendor‑lock‑in complaints into performance art and has yet to elicit a public response from the cloud giants.
Under-trained techie didn't claim overtime for mistakenly failing to phone it in — The Register
These Privacy-Conscious Gay Dating Apps Want to Dethrone Grindr — WIRED
The article critiques mainstream gay dating apps for prioritizing monetization over meaningful connection, citing Grindr’s ad‑heavy model and high‑priced subscription tiers and the backlash over Sniffies’ Match Group investment, which raised fears of corporate takeover. In response, tech entrepreneur Calum Bowden launched MeetMarket, a privacy‑centric, decentralized web app that stores user data on devices, offers end‑to‑end encryption, and remains ad‑free, attracting over 12,000 sign‑ups in 48 hours and 60,000 users overall with a monthly fee of €12. Bowden, backed by his nonprofit Trust, aims to give queer users agency and protect them in hostile legal contexts. The piece notes a broader trend toward niche, community‑driven apps—such as Justin Finnegan’s Chunkr—that cater to specific subcultures within the gay community, suggesting that users are increasingly seeking specificity and intentionality over the generic, profit‑driven platforms of Big Dating.

Week ending 2026-05-18

2026-05-18T11:53:31Z

Linus Torvalds says AI-powered bug hunters have made Linux security mailing list ‘almost entirely unmanageable’ — The Register
Linus Torvalds has warned that the Linux kernel’s security mailing list has become “almost entirely unmanageable” because many researchers now use the same AI tools to uncover bugs and submit duplicate reports, creating “pointless churn.” In a recent kernel‑state post, he highlighted that the flood of AI‑generated findings forces developers to spend time forwarding or dismissing identical issues rather than addressing new ones, and he urged reporters to read the documentation, patch the code, and add genuine value instead of just sending “random reports.” Torvalds sees the practice as wasteful, especially since AI‑discovered bugs are not secret and sharing them on a private list only amplifies duplication. This stance contrasts with fellow maintainer Greg Kroah‑Hartman’s more positive view of AI’s usefulness for free‑software security.
Russian hackers turn Kazuar backdoor into modular P2P botnet — BleepingComputer
Secret Blizzard, a Russian hacking outfit linked to the FSB and operating alongside groups such as Turla, has upgraded its long‑standing Kazuar backdoor into a modular, peer‑to‑peer botnet that excels at stealth, persistence and data exfiltration. Since 2017 (with roots tracing back to 2005) Kazuar has targeted governments, diplomatic missions, defense entities and critical infrastructure across Europe, Asia and Ukraine. Microsoft’s analysis reveals a three‑module architecture—Kernel (the leader coordinator that manages tasks and internal communication), Bridge (the external proxy that talks to the C2 server via HTTP, WebSockets or EWS), and Worker (the espionage engine that keylogs, screenshots, harvests files, performs reconnaissance and steals email data). The botnet uses internal IPC (Windows Messaging, Mailslots, named pipes) with AES‑encrypted, Protobuf‑serialized messages, while the Bridge handles encrypted exfiltration. Kazuar’s new version supports 150 configuration options, including AMSI, ETW and WLDP bypasses, task scheduling, process injection, and granular data‑exfiltration controls, making it highly adaptable and evasive. Microsoft advises defenders to prioritize behavioral detection over static signatures to counter this modular, long‑term threat.
An ICE Firearms Trainer Was Involved in At Least 4 Deadly Shootings — WIRED
David S. Norman, a former Phoenix Police officer who founded the law‑enforcement training firm TruKinetics LLC, has a controversial record that now shadows the paramilitary training he provides to U.S. immigration enforcement. Norman was involved in six on‑duty shootings while with the Phoenix Police, killing four people, and later testified in a lawsuit that he participated in at least four lethal shootings. TruKinetics, based in Gilbert, Arizona, runs intensive small‑team and close‑quarters combat courses—including “kill house” scenarios—and was paid nearly $28,000 to conduct a 40‑hour mandatory training program for Department of Homeland Security’s Special Response Teams (SRTs) at Fort Benning, Georgia, where over 700 agents from ICE and CBP rotate annually. Despite Norman’s claims that his courses focus on tactical skills rather than crowd control or active‑shooter drills, the firm’s training of SRTs, which have been used in civil immigration enforcement and have been implicated in the deaths of protestors, raises concerns about the nature of the training and the background of its instructor.
Mozilla warns UK: Breaking VPNs will not magically fix Britain's age-check mess — The Register
Mozilla has warned the UK government that targeting VPNs as a tool to curb children’s age‑check evasion would be a misguided and harmful policy move. In a submission to the Department for Science, Innovation and Technology, Mozilla’s policy manager Svea Windwehr highlighted that VPNs are essential for privacy and security for millions—from remote workers to journalists—by masking IP addresses and preventing profiling. The company points out that only a small fraction of kids use VPNs to bypass age gates, and that most workarounds rely on fake birth dates, borrowed accounts or weak facial‑scan checks. Firefox’s own move to embed a built‑in VPN shows the tech sector is leaning toward integrated privacy rather than standalone tools. Mozilla argues that regulating VPNs would push Britain toward “safety through surveillance” while ignoring the real drivers of online harm—platform recommendation systems and engagement algorithms—rather than addressing the root causes of age‑gate circumvention.
How to Watch Google I/O — WIRED
Google’s upcoming I/O, set for May 19 at the Shoreline Amphitheatre, will showcase new Android 17 features—many AI‑powered, such as smarter voice‑to‑text, autofill, digital‑well‑being tools, a refreshed Android Auto UI, 3D emojis, and “Gemini Intelligence” for flagship devices—alongside a new Googlebook line of premium, AI‑first laptops from partners like Acer, HP, Lenovo, and Dell. The keynote, streamed live on Google’s site and YouTube, is expected to include surprise demos and further AI updates, especially to Search, Google Labs’ Flow music tool, and a potential video‑generation Gemini model called Omni.
The Backward Logic of Chickenpox Parties — WIRED
Chickenpox, once an almost inevitable childhood illness, was historically managed through a now‑obscure practice known as “chickenpox parties,” where parents deliberately exposed children to the varicella‑zoster virus at the lowest risk of severe complications—typically in early childhood—before routine vaccination became available. Those parties relied on shared knowledge of home remedies and the belief that a controlled outbreak would spare the family from later, potentially life‑threatening adult infections, such as pneumonia, encephalitis, or meningitis, which occur in about 3 per 1,000 cases. With the introduction and widespread use of the vaccine, the incidence of natural infection has dropped dramatically, rendering the practice largely obsolete while highlighting the historical context of how communities once navigated this highly contagious, itch‑ridden disease.
Supercharging Immune Cells May Help Control HIV Long-Term — WIRED
Scientists have adapted CAR‑T cell therapy—originally a cancer treatment—for HIV, and early data from two participants in a small safety study show promise: after a single infusion of genetically engineered autologous immune cells, both patients have maintained undetectable viral loads for almost 12–24 months and have been able to discontinue antiretroviral drugs entirely. The trial, presented at the American Society of Gene and Cell Therapy meeting, suggests that teaching the immune system to control HIV without lifelong medication is achievable, offering a potential functional cure that could be safer, more scalable, and more affordable than previous stem‑cell–based approaches that required donor cells with the CCR5 mutation.

Week ending 2026-05-11

2026-05-11T11:34:35Z

I Work in Hollywood. Everyone Who Used to Make TV Is Now Secretly Training AI — WIRED
The new Wild West of AI kids’ toys — Ars Technica
The article critiques the booming market for AI‑powered children’s toys, noting that while Pixar’s Toy Story 5 villain is a green frog‑shaped tablet, real‑world competitors like FoloToy, Alilo, Miriat and Miko’s Miko have already flooded trade shows and online stores with soft teddy bears, bunnies, and “robots” that use advanced models such as GPT‑4o. With over 1,500 AI toy companies registered in China alone, the sector remains largely unregulated, and consumer watchdogs have identified dangerous content—ranging from instructions for violence and sexual references to political propaganda—spilling from toys like FoloToy’s Kumma bear and Miriat’s Miiloo. The piece warns that beyond inappropriate material, the allure of AI companions (“I’m gonna be your best friend”) could disrupt children’s social development, underscoring the urgent need for stricter guardrails and regulatory oversight.
Linux kernel maintainers pitch emergency killswitch after CopyFail and Dirty Frag chaos — The Register
Linux kernel maintainers are proposing a “Killswitch” feature that lets system administrators temporarily disable specific vulnerable kernel functions at runtime, rather than wait for a patch, patch build, and reboot. The idea—submitted by co‑maintainer Sasha Levin—addresses the exposure that occurs while patches are still under development, as highlighted by recent exploits such as CopyFail and Dirty Frag. By pulling the plug on a buggy function, calls to it will fail immediately, preventing exploitation until a proper fix is applied. While the mechanism would sacrifice some subsystem functionality (e.g., networking or cryptography) for security, it has sparked debate over potential abuse, stability impacts, and the ethics of giving admins the power to “shut the kernel down” on demand. Nonetheless, the community’s experience with recent vulnerabilities suggests that disabling broken code may be a more practical interim defense than leaving it exposed.
A Chevron Texas Power Plant Seeks School District Tax Break — WIRED
A Texas oil firm, Chevron’s Energy Forge One, is seeking a multi‑hundred‑million‑dollar state tax abatement under the JETI Act to build a West Texas gas‑powered plant that will supply electricity exclusively to a data center—potentially for Microsoft, which has been in talks with Chevron and the project’s investment fund Engine 1. The plant will not connect to the grid but will provide behind‑the‑meter power to a future data‑center tenant, a model that has surged as big tech grapples with grid delays and public scrutiny over data‑center energy costs. The Texas Comptroller’s board has recommended approval, making this the first JETI tax break for a plant intended solely for data‑center use, and the school board that would normally receive the tax revenue has waived its share because the state funds the abatement. The project could save Chevron more than $227 million over ten years and is expected to create about 25 permanent jobs, although that requirement is not enforced for generation facilities.
CUDA Proves Nvidia Is a Software Company — WIRED
Keyboard Shortcuts I Learned From My Cat — WIRED
The article humorously recounts how the narrator’s cat Mira, by simply walking across MacBook and Windows keyboards, inadvertently triggers a variety of hidden keyboard shortcuts that reveal new computer functions—such as hiding windows (⌘‑⌥‑H or Win+,), switching virtual desktops (Ctrl‑←/→ or Win‑Ctrl‑←/→), taking area screenshots (⌘‑Shift‑4 or Win‑Shift‑S), toggling the macOS dock (⌘‑⌥‑D), and unintentionally enabling accessibility features like Sticky Keys and FilterKeys by pressing Shift repeatedly or holding the right‑shift key; the writer expresses both amusement and newfound appreciation for these shortcuts while noting ongoing efforts to train Mira away from the desk.
Classic Outlook's Quick Steps trip over Microsoft bug — The Register
In Microsoft Outlook’s classic client, a bug introduced in build 2512 can cause Quick Steps to appear grayed out when the actions they contain can’t be performed—e.g., a step that moves a message to a folder and clears categories will be disabled for messages lacking categories. Microsoft acknowledges the issue and offers a workaround: the Quick Step can still be triggered via its keyboard shortcut even when the UI grays it out. This glitch is part of a series of recent problems with Classic Outlook, which Microsoft is gradually phasing out in favor of the new Outlook app, though it still retains features like COM support that the newer version lacks.

Week ending 2026-05-04

2026-05-04T10:23:18Z

OpenAI Introduces AI-Generated Pets for Its Codex App — Slashdot
Descartes argues that the idea of an empty void—completely devoid of matter or even space—is logically impossible; even if nothing tangible occupies a region, that region still possesses the property of space itself, so there can never truly be a "vacuum" in the absolute sense.
How to Track Your Luggage (2026): AirTag, Pebblebee, CaseSafe — WIRED
The Wired Reviews article offers practical advice for stress‑free travel, emphasizing the importance of tracking luggage when checked‑in items become unavoidable. The author shares a personal anecdote about a surprise propeller flight that forced carry‑ons to be gate‑checked, and how Bluetooth trackers helped him confirm his suitcases stayed with him through layovers. The piece encourages readers to avoid checked bags whenever possible but provides tips for those that must be checked, including using trackable cases. It highlights the July “Trackable Carry‑On” from CaseSafe as the only carry‑on sized sufficiently small (21.5 × 15 × 8.5 in) to fit in the overhead bin of a DHC‑8‑400, noting its slim dimensions and built‑in tracking feature. The article also directs readers to related reviews on the best carry‑on suitcases, travel totes, and laptop backpacks.
Five Eyes spook shops warn rapid rollouts of agentic AI are too risky — The Register
A joint Five Eyes guidance document—issued last Friday by Australia, the United States, Canada, New Zealand and the United Kingdom—warns that agentic AI systems, which autonomously act across critical infrastructure and defense, are likely to misbehave and expand an organisation’s attack surface. The guide cites illustrative scenarios (e.g., an AI patch‑manager deleting firewall logs or a procurement bot enabling fraud after a low‑risk tool is compromised) to show how even well‑designed agents can be abused. It lists 23 specific risks, over 100 mitigation practices, and stresses that developers, vendors and security teams must test, fail‑safe‑design, and human‑escalate uncertain actions. The report urges a highly cautious, incremental rollout that prioritises resilience, reversibility and risk containment over efficiency, demanding strong governance, accountability, monitoring and oversight until mature security standards for agentic AI emerge.
Can Listening to ‘Subliminals’ Make You Beautiful? Plenty of Women Believe It — WIRED
The article explores the rise of “subliminal” videos—short audio‑visual clips that embed hypnotic affirmations aimed at reshaping viewers’ self‑image and external appearance—on platforms like TikTok and Reddit. With millions of views, these clips promise everything from flawless skin to ideal body proportions, and even broader life gains such as academic success or romantic attraction. While originally rooted in a niche 2012 subreddit, the trend has expanded to cover a wide array of “looksmaxxing,” “facemaxxing,” and “beautymaxxing” content, echoing the anxieties driving male looks‑maximizers into radical online cultures. The article highlights how young women, pressured by unrealistic beauty standards, increasingly turn to these quasi‑spiritual tools instead of traditional fitness or medical approaches, citing anecdotal reports of significant personal change and the emergence of self‑made subliminal creators.
Science Has Found Even More Ways Coffee Is Good for You — WIRED
A recent study published in *Nature Communications* examined how regular coffee drinking affects the gut‑brain axis in healthy adults, distinguishing caffeine’s role from other coffee compounds. Thirty‑one habitual coffee drinkers (3–5 cups daily) and thirty‑one non‑drinkers were compared, and participants also cycled through two‑week coffee abstinence followed by re‑introduction of either caffeinated or decaffeinated coffee. The results showed that coffee alters gut microbiota composition—boosting species such as *Eggerthella* and *Cryptobacterium curtum*—and reduces inflammatory markers while increasing anti‑inflammatory molecules, effects that persist even with decaffeinated coffee. Behaviorally, regular drinkers exhibited higher impulsiveness and emotional reactivity, which eased during the abstinence period, and both coffee types lowered perceived stress and depressive symptoms; caffeine‑rich coffee further improved anxiety and attention, whereas decaf enhanced memory, sleep quality, and activity. Cortisol levels remained unchanged, suggesting coffee mainly influences subjective stress perception. Overall, the study highlights coffee’s complex, caffeine‑independent modulation of the microbiome, metabolism, and neuro‑behavioral states, supporting its potential as a dietary intervention for gut and mental health.
15 Best MagSafe Wireless Chargers (2026): Power Banks, Stands, Pads, and Travel Chargers — WIRED
The article reviews a range of MagSafe‑compatible Qi2 wireless chargers that didn’t earn top honors. The Mous Dual Charging Station offers dual pads (15 W MagSafe, 5 W Qi) plus a USB‑C port but is slow. Anker’s MagGo delivers a sturdy 15 W pad with a built‑in cable, while Belkin’s BoostCharge Pro converts into a stand and charges up to 15 W. Casetify’s PowerThru 2‑in‑1 charges iPhones at 7.5 W and earbuds at 15 W, and its cheaper single‑pad version offers 7.5 W MagSafe or 15 W Qi with a decorative design. Mophie’s 3‑in‑1 travel charger packs iPhone, Watch, and AirPods charging (15 W) in a fold‑out case. All options blend design variety with modest charging speeds, providing alternatives for those seeking MagSafe convenience at different price points.
Dating Is a Rich Person’s Game Now — WIRED
Modern dating has become a costly, “date‑flation”‑driven activity that many people now have to budget for, according to a 2026 survey by JG Wentworth and a BMO Real Financial Progress Index report, which found the average all‑in date cost rose 12.5 % to $189—outpacing living expenses and pushing 86 % of US singles to delay or abandon dating. Low‑income earners are hit hardest, with one‑third of those making under $50,000 stopping entirely, while even high‑income individuals are taking breaks. This economic friction turns dating into a luxury, driving intentional but unequal participation, especially among Gen Z‑X men, who are increasingly opting out and sharing the strain on platforms like TikTok. The trend has fueled discussions around sugar babies and the broader financial realities of romance, suggesting that connection is no longer spontaneous but a costly, discretionary pursuit.

Week ending 2026-04-27

2026-04-27T10:26:20Z

Trump fires the entire National Science Board — The Verge
The article criticizes a presidential decision to target the National Science Board (NSB), arguing that the board—an apolitical body that advises the President on the National Science Foundation’s future—has been undermined by a leader who has attacked the NSF from the outset. It warns that replacing the NSB with politically aligned appointees could erode scientific independence and hand control of American innovation to adversaries, labeling the move as a foolish, self‑serving tactic.
Google Studies Prompt Injection Attacks Against AI Agents Browsing the Web — Slashdot
Neil Bogart’s quote distills a cynical view on hype and success: if a product or idea is heavily promoted and ultimately succeeds, the hype was justified and the promoter is vindicated; but if the same hype fails to deliver results, it is merely hype. Thus, the outcome of the hype determines whether it was a valid strategy or a mere puffery.
Prime Video drops full trailer for Spider-Noir — Ars Technica
Prime Video’s latest buzz comes from the full trailer for the upcoming live‑action series *Spider‑Noir*, unveiled in Mexico City over the weekend. The trailer was released in two styles: a stark black‑and‑white version that feels like a Raymond Chandler film, and a colorful “True Hue” cut. The show is a continuation of Marvel’s 2009 noir line, reimagining characters in a Great Depression setting. It stars Nicolas Cage as Ben Reilly, a hard‑boiled private investigator who secretly becomes the city’s lone superhero, “The Spider,” in 1930s New York. Co‑showrunner Oren Uziel, a film‑noir enthusiast, chose to replace the familiar Peter Parker with Reilly to better fit the gritty, adult tone of the series.
Watch out UK taxpayers: 28,000 HMRC staffers just got an AI copilot — The Register
HMRC has deployed 28,000 Microsoft Copilot licences after a 20,000‑civil‑servant trial that showed an average 26‑minute daily time saving and high user enthusiasm, prompting the department’s chief AI officer to announce plans for “agentic‑style” features and a vision of becoming the world’s most AI‑enabled tax authority. The trial also highlighted limitations in handling complex, data‑heavy tasks and raised security concerns about sensitive data, raising doubts about using Copilot in official “Sensitive” workflows. HMRC frames Copilot as a bolt‑on to existing automation that has delivered £8 billion in benefits, but critics point to the government’s stale IT estate, duplicated content and the risk of embedding a generative system that is hard to turn off once adopted, leaving the agency with a tool that works well enough to rely on but is not yet trustworthy for critical decisions.
Strange New Worlds S4 teaser strikes a more serious tone — Ars Technica
Paramount+ revealed a new teaser for *Star Trek: Strange New Worlds* Season 4 at CCXP in Mexico City, hinting that the show will return to a more serious tone after the uneven third season. The clip revisits the crew’s response to a distress call, a crippling space storm, and a mission to a nearby planet for iridium, echoing the extended sequence from last October’s New York Comic‑Con. Fans can also expect a puppet episode, a Wild‑West‑style setting, and even dinosaurs, while Pike’s ongoing grief over the departure of Marie Batel will likely shape the season’s emotional stakes.
New robotic control software avoids jamming their joints — Ars Technica
EPFL researchers have created “Kinematic Intelligence,” a framework that lets robots switch between different arm models without retraining from scratch, mirroring how smartphones sync data when upgraded. The system addresses a long‑standing issue in robot learning‑from‑demonstration: skills learned on one robot often fail on another due to subtle differences in link lengths, joint orientations, or configurations. By automatically adapting learned behaviors to the new robot’s kinematic constraints, Kinematic Intelligence enables seamless transfer of tasks—such as wiping, stacking, or welding—across diverse robotic platforms.
Here’s How Much San Francisco Tech Companies Pay for Police Protection — WIRED
Elon Musk’s 2024 move of X’s offices outside San Francisco after calling the city’s violent crime “horrific” highlights a growing trend of tech firms paying for police protection under the city’s 10B rent‑a‑cop program, which allows businesses to hire officers at standard city rates. 2024 records show Airbnb spent about $428,000 on uniformed, armed officers, while Salesforce paid roughly $728,000 through a security vendor and added $41,000 for police at its TrailblazerDX conference. OpenAI and Anthropic have not been regular clients, though OpenAI briefly paid $813 for a museum event. The program, used by a range of companies—including the Giants baseball team ($1.9 M), Apple ($1.2 M), and various retailers—offers a low‑cost, quick‑response security solution, but raises concerns about conflicts of interest and liability, even as it provides officers with additional income and fosters community ties.

Week ending 2026-04-20

2026-04-20T10:07:43Z

Blue Origin's rocket reuse achievement marred by upper stage failure — Ars Technica
Blue Origin’s third New Glenn launch marked the first successful reflight of an orbital‑class booster, but the mission ended with a setback for the company’s flagship rocket, which is slated to carry NASA’s Artemis crew to the moon. The 321‑ft vehicle lifted off from Cape Canaveral at 7:25 am EDT, the BE‑4 methane engines powered the lift‑off and the first‑stage booster shed its engines after three minutes, separating from the upper stage that used BE‑3U liquid hydrogen/oxygen engines. The booster followed a parabolic trajectory, re‑ignited for two braking burns, and landed safely on a ship in the Atlantic, becoming the second flight of the “Never Tell Me The Odds” booster. While the successful recovery demonstrates progress toward reusable, high‑volume launches, the flight also exposed challenges in New Glenn’s larger, more complex design—highlighting that, unlike SpaceX’s rapid Falcon 9 refurbishment, Blue Origin still has hurdles to overcome before it can match the rapid reflight cadence needed for the Artemis program.
The Best Hybrid Mattresses for Couples, Back Pain, and More (2026) — WIRED
Hybrid mattresses combine springs and foam layers with a wide array of additional materials—latex, different foam densities, fabrics and cooling covers—to deliver a balance of softness, support and breathability; the key to choosing the best one is personal testing, as no single model suits everyone, which is why Wired reviewers spend weeks on each mattress, noting fit for height, weight, sleep position and specific needs such as cooling. The article reviews a spectrum of hybrids, from side‑sleep‑focused to back‑pain relief and organic options, and as of April 2026 has added the Tiami Luxury Hybrid and Saatva HD to the lineup, with updated prices.
Blue Origin nails the landing, but puts the payload satellite in the wrong orbit — The Register
Blue Origin’s New Glenn rocket successfully landed its reused first stage on the Jacklyn platform during its April 19 launch, but the mission failed to place AST SpaceMobile’s satellite Bluebird 7 into a usable orbit; the second stage delivered the satellite to a lower‑than‑planned, off‑nominal orbit that its own thrusters could not correct, forcing AST to de‑orbit the payload and recover its cost through insurance. The incident leaves the second stage’s performance unclear and raises questions about Blue Origin’s upcoming schedule—including a Starlink competitor, additional AST payloads and a lunar cargo lander for a potential Artemis‑III demonstration—while underscoring that the successful landing and reuse of the rocket was outweighed by the loss of the critical payload.
I’ve fired one of America’s most powerful lasers—here’s what a shot day looks like — Ars Technica
The Texas Petawatt (TPW), a 17‑story‑deep, high‑power laser facility beneath the University of Texas at Austin’s Physics, Math, and Astronomy building, was a DOE‑funded research center that produced petawatt‑level pulses—brief bursts of light carrying more power than the U.S. electrical grid for a trillionth of a second—by stretching, amplifying, and then compressing laser pulses. As lead laser scientist from 2020‑2024, I oversaw “shot days” where targets ranging from ultra‑thin metal foils to gas jets were irradiated to probe stellar physics, fusion, and even medical applications. The process involved hours of meticulous preparation, culminating in a 10‑second window of near silence during the actual pulse, before the facility was shut down due to funding cuts.
There’s New Evidence for How Loneliness Affects Memory in Old Age — WIRED
A new six‑year longitudinal analysis of 10,217 European adults aged 65–94, published in *Aging & Mental Health*, shows that loneliness is linked to lower baseline memory performance—both immediate and delayed recall—but does not accelerate the rate of cognitive decline; age itself remains the strongest predictor of memory loss, with sharper drops after 75 and especially after 85, while depression, chronic illness, and loneliness further lower initial scores. Physical activity, however, lifts the baseline memory level, acting as a cognitive buffer without altering decline speed. Though the study does not pinpoint causality, it highlights loneliness as a significant factor in initial memory deficits, underscoring the need to address social isolation as part of strategies to mitigate cognitive decline in an aging population projected to grow substantially by 2050.
Best Meta Glasses (2026): Ray-Ban, Oakley, AR — WIRED
Meta’s Ray‑Ban‑styled AR glasses—sold in over seven million pairs by 2025—have become a ubiquitous sight at outdoor events, offering a single, stylish device that doubles as sunglasses, headphones, and a camera, thanks to a partnership with EssilorLuxottica. While the tech lets users capture content and enjoy an integrated experience, it also raises serious privacy concerns, as Meta’s history of questionable data practices and the potential for covert recording make many people uneasy about wearing “pervert glasses.” Despite these issues, the latest model upgrades the original design with a 12‑MP camera, up to eight hours of battery life, and options for clear, prescription, or transition lenses, making it a practical choice for those who value convenience and style over the broader societal implications.
Hisense U7SG TV Review (2026): Better Design, Great Value — WIRED
Hisense’s new U7SG mid‑range TV blurs the line to premium with a sleek industrial stand, matte anti‑reflective panel, and bright mini‑LED backlighting that rivals Samsung’s high‑end models; it ships with a top‑tier gaming setup, a Google TV interface, and a wide range of sizes up to 116 inches. While the latest U8 (UR8) pushes RGB LED technology for more vivid colors and higher brightness, the U7SG still feels a touch “budget” due to side‑view performance quirks and some software quirks that can make demanding scenes look odd—though these are fixable. Overall, its eye‑catching colors and improved image clarity make it a solid budget choice, especially as its price is expected to drop soon.

Week ending 2026-04-13

2026-04-13T10:05:19Z

Sam Altman reportedly targeted in second attack — The Verge
Sam Altman’s San Francisco home was the target of two separate threats in the span of a weekend. On April 10, a 20‑year‑old was arrested after allegedly throwing a Molotov cocktail at the Russian Hill residence; the investigation remains open. Two days later, police received a 2:56 AM call about possible gunfire, captured a vehicle’s license plate on surveillance, and arrested two occupants (25 and 23) who were charged with negligent discharge after a weapon was fired at the house. Officers seized three firearms during the arrest. San Francisco Police Chief Derrick Lew emphasized the department’s commitment to prosecuting gun‑related crimes.
Google removes "Doki Doki Literature Club" from Google Play — Hacker News (hnrss)
The article emphasizes that the web application in question is highly interactive and necessitates JavaScript; while a basic HTML interface could technically function, it would not serve the intended purpose. It directs readers to learn more about the platform through Bluesky (bsky.social) and the underlying protocol at atproto.com.
Suspect Arrested for Allegedly Throwing Molotov Cocktail at Sam Altman’s Home — WIRED
San Francisco police arrested a 20‑year‑old man early Friday after he allegedly threw an incendiary device at OpenAI CEO Sam Altman’s home and later threatened to burn down the company’s Mission Bay headquarters. Security staff notified employees of the incident and advised caution, while OpenAI remained open but heightened security. The suspect was apprehended within an hour of the first attack, and no injuries or significant damage were reported; charges are pending. This follows prior threats against the company, including a November activist scare and February 2025 protests.
How and When to Watch the Artemis II Mission’s Return to Earth — WIRED
NASA’s Artemis II mission, after a 10‑day lunar excursion, is set to re‑enter Earth’s atmosphere on Friday, April 10 at 5:07 p.m. PDT (with corresponding U.S. times from 5:07 p.m. SF to 8:08 p.m. NY). The Orion capsule will plunge toward southeast Hawaii at about 38,400 km/h, endure temperatures up to 2,760 °C on its exterior, and subject the crew to nearly 3.9 g’s before splashing down in the Pacific off California in roughly 13 minutes. The re‑entry is a critical test of NASA’s deep‑space technology, with the event streamed live on NASA+, Amazon Prime, Apple TV, Netflix, HBO Max, and YouTube.
Shock from Iran war has Trump's vision for US energy dominance flailing — Ars Technica
Trump’s claim that the U.S. can weather the chaos sparked by Iran’s blockade of the Strait of Hormuz ignores the sharp reality seen at pumps across America, where gasoline prices have spiked by more than $4 per gallon—an $8.4 billion hike in a month—illustrating the nation’s vulnerability to global supply shocks. While the U.S. is the world’s largest oil and gas producer, it also consumes the most, making it heavily dependent on the same international market that the conflict has disrupted. Iran’s insistence on controlling the Strait—and its threat to charge tolls—has kept tankers stalled, and even a two‑week ceasefire has not fully restored flow. The Energy Information Administration warns that elevated prices will persist through year‑end, and critical infrastructure such as Qatar’s Ras Laffan LNG terminal may suffer long‑term damage. As the U.S. rolls back climate and clean‑energy policies, its dual role as producer and consumer leaves it exposed to the very global disruptions it claims to dominate.
‘The Audacity’ Is the Broligarchy Takedown You Were Waiting For — WIRED
“The Audacity” is an AMC drama that follows billionaire tech mogul Duncan Park (Billy Magnussen) as he manipulates markets, stalks his therapist and pushes his privileged family into toxic environments, all while trying to maintain his image as a genius. The series, created by Jonathan Glatzer, echoes the anti‑One‑Percent satire of shows like *Succession* and *Silicon Valley*, but adds a new angle: it exposes the emotional wreckage that comes from unchecked power and narcissism. Duncan’s obsession with control—evidenced by a failed sale of his firm Hypergnosis, his reliance on an ayahuasca shaman, and his covert use of AI to spy on Dr. JoAnne Felder (Sarah Goldberg)—leads to a dangerous cat‑and‑mouse dynamic that spirals into insider trading and blackmail. Meanwhile, the neglected children of both Duncan and JoAnne drift through cut‑throat private schools, highlighting the human cost of wealth and ambition. The show uses these personal dramas to critique the culture of Silicon Valley excess and the broader societal fallout of elite dominance.
Why Is It So Hard to Fix an Electric Bike? (2026) — WIRED
Bike shops are increasingly wary of taking on electric‑bike repairs because the market is littered with vehicles that don’t meet the legal Class 1‑3 definition—many sold online exceed the 28 mph limit and function more like mopeds than bicycles. Owners such as MacKenzie Hardt of Hardt Family Cyclery now refuse to repair any ebike without UL 2849 certification, citing safety and liability concerns after witnessing injuries from poorly controlled power systems. With nationwide injury rates for e‑bikes up by over 1,000 % since 2020, insurers often void coverage on non‑compliant bikes, leaving shops exposed to product‑liability claims and legal uncertainty that varies state to state.