The Supreme Court upheld birthright citizenship, rejecting an executive order. U.S. labor share of income fell to a post-war low, viewed as cyclical. In Virginia, Henrico County’s 37 data centers strained the grid, sparking a 25% rate hike and school conservation orders. Tech scrutiny intensified as an audit exposed steganographic tracking in Claude Code, raising trust concerns, while online marketplaces battle scams selling seeds for AI-fabricated flowers.
🤖 AI & Generative Models
Scammers Sell Seeds for Exotic AI-Generated Flowers That Don’t Exist
Fraudulent sellers are using AI-generated images of fantastical, non-existent plants to scam buyers on major e-commerce platforms like eBay, Amazon, and Etsy. These listings have driven thousands of sales, resulting in customers receiving non-viable, incorrect, or missing shipments. Beyond defrauding consumers, the widespread flood of fabricated imagery is increasingly polluting online plant databases and search results.
Words Are a Byproduct of Consciousness. For LLMs, It’s Backwards
Large language models fundamentally differ from human cognition by generating text through statistical prediction rather than underlying conscious thought. This technological reversal dramatically lowers execution barriers, ultimately shifting the primary value of innovation from technical implementation to human imagination and idea generation.
- Words Are a Byproduct of Consciousness. For LLMs, It’s Backwards — ranpara.net
Gemma 4 on Cerebras - The Fastest Inference Is Now Multimodal
Google DeepMind’s open-weight Gemma 4 31B multimodal model now achieves a record 1,851 tokens per second on Cerebras Inference hardware, delivering 35 times the speed of standard GPUs and a 1.5-second first-token latency. This performance breakthrough enables developers to build real-time, highly responsive multimodal and agentic AI applications that were previously too slow for practical use.
- Gemma 4 on Cerebras - The Fastest Inference Is Now Multimodal — cerebras.ai
Open Memory Protocol – One Memory Store for Claude, ChatGPT, Curso
The Open Memory Protocol (OMP) is a new vendor-neutral specification designed to enable portable and interoperable AI memory across different platforms, such as Claude, ChatGPT, and Cursor. It provides a standardized framework of servers, SDKs, and adapters to prevent information loss and maintain user context when switching between various AI tools and devices.
AI & vibe coding fuel mobile game surge: 181K launches, up 118% on iOS and 73% on Android YoY
Generative AI and “vibe coding” have driven a massive surge in mobile game releases, with 181,000 titles launched in the six months to May, marking year-on-year increases of 118% on iOS and 73% on Android. Despite this production boom, the gaming industry is facing significant layoffs and growing concern among professionals regarding AI’s impact on creativity and job security.
- AI & vibe coding fuel mobile game surge: 181K launches, up 118% on iOS and 73% on Android YoY — ft.com
Travel industry rushes to launch tools and loyalty schemes to rival AI travel agents.
Travel companies, including hotels and tour operators, are developing proprietary tools and strengthening loyalty programs to counter the rise of “agentic” AI agents capable of autonomous booking. These businesses are increasingly integrating with large language models to provide personalized experiences and maintain consumer trust amidst the shift toward automated travel planning.
OpenAI teases new device with Work Louder launching July 15, featuring Codex upgrades
OpenAI has teased a new hardware device launching on July 15 in collaboration with keyboard manufacturer Work Louder. The upcoming product is expected to provide upgraded shortcuts for OpenAI’s Codex coding tool, potentially featuring programmable buttons or keys.
- OpenAI teases new device with Work Louder launching July 15, featuring Codex upgrades — theverge.com
What the OCI MSA didn’t solve for AI scaling
The OCI MSA has standardized AI networking architecture around a slow-and-wide, wavelength-division multiplexing design, but leaves the manufacturing methods for scaling bandwidth beyond four wavelengths open. To maintain energy efficiency as demand grows, the industry consensus favors increasing wavelength counts rather than raising symbol rates. This shifts the primary challenge from architectural debate to the mass production of stable, high-precision laser arrays.
- What the OCI MSA didn’t solve for AI scaling — theregister.com
Where there’s a will, AI still has work to do
A probate lawyer has discovered that while AI can generate documents that appear visually authentic, they often fail to address critical legal nuances. Although the AI-generated document looked professional, it missed many of the essential questions necessary for its validity.
- Where there’s a will, AI still has work to do — theregister.com
đź’» Developer Tools & Code
Claude Code Is Steganographically Marking Requests
An audit of the Claude Code desktop client revealed steganographic mechanisms that silently encode API routing data into system prompts using obscure Unicode characters within standard date strings. These hidden markers trigger based on custom base URLs, specific timezones, and domain keywords to help Anthropic detect unauthorized resellers, proxy gateways, and AI lab endpoints. While the feature appears aimed at preventing abuse, its covert implementation and obfuscated configuration lists have raised significant transparency and trust concerns regarding a developer tool with extensive system access.
- Claude Code Is Steganographically Marking Requests — thereallo.dev
Cursor now has a mobile app for guiding your coding agent on the go
Cursor launched Cursor Mobile, enabling developers to create and manage AI coding agents directly from their smartphones. The app aligns with similar releases from Anthropic and OpenAI, reflecting a broader industry shift from direct code editing to overseeing autonomous coding agents. As a result, developers are increasingly abandoning traditional desktop setups for continuous, phone-based workflows.
- Cursor now has a mobile app for guiding your coding agent on the go — techcrunch.com
Looking Ahead to Postgres 19
PostgreSQL 19 prioritizes practical, production-ready enhancements over flashy new features, focusing on long-term operational stability and daily workflow efficiency. The upcoming release introduces built-in REPACK CONCURRENTLY for low-lock table reorganization, new partition merge and split capabilities for flexible schema evolution, and further matured logical replication alongside steady improvements to vacuum, query planning, and system monitoring.
- Looking Ahead to Postgres 19 — snowflake.com
Knoppix
The Chemnitz Linux-Days 2025 will be held on March 22–23 under the theme “The Culture of Open Source.” The event will feature a German-language presentation addressing the opportunities and risks of generative AI in university lectures and exams. This announcement was posted on the official Knoppix Linux news portal.
- Knoppix — knopper.net
Parsing Over Validation in Programming
Advocating for a “parse, don’t validate” approach in TypeScript, these articles highlight how the language’s structural typing naturally discards validation results, often leading to redundant checks and runtime errors. To address this limitation, developers are encouraged to combine parsers with branded types that simulate nominal typing. This technique embeds proof of correctness directly into the type system, ensuring only verified data safely progresses through the application.
- Parse, Don’t Validate – In a Language That Doesn’t Want You To — cekrem.github.io
Zluda 6 release (run unmodified CUDA applications on non-Nvidia GPUs)
The ZLUDA Version 6 release introduces support for PhysX and Blender textures, allowing unmodified CUDA applications to run on non-NVIDIA GPUs. The update also features improved Windows compatibility through more robust library management and enhanced performance for machine learning workloads.
- Zluda 6 release (run unmodified CUDA applications on non-Nvidia GPUs) — vosen.github.io
Exploring PDP-1 Lisp (1960)
This article provides a practical guide for initializing and programming Lisp on the DEC PDP-1 computer. It details necessary hardware switch configurations, references essential technical manuals, and demonstrates basic operations such as program execution and input handling.
- Exploring PDP-1 Lisp (1960) — obsolescence.dev
Letos: Create, edit, browse SQLite databases. Formerly known as SQLiteStudio
Letos, formerly known as SQLiteStudio, has released version 4.0.0, marking its first major update under the new project name. This open-source, cross-platform tool introduces a new ERD editor, Qt 6 integration, and native ARM64 builds for Windows, Linux, and macOS.
Run any Dockerfile on Vercel
Vercel now enables the deployment of any stateless, HTTP-speaking container to its Fluid compute platform via a Dockerfile.vercel, supporting diverse stacks like Go, Java, and Rails without requiring local daemon or registry management. The platform automatically handles image building, storage, and bidirectional autoscaling while offering instant preview URLs for every commit, integrated observability, and a unified network for full-stack deployment. Users pay only for active CPU execution time, benefiting from optimized startup performance through streaming boot images.
- Run any Dockerfile on Vercel — vercel.com
Underappreciated builtin: Grand Unified Debugger
The Grand Unified Debugger (GUD) is a built-in Emacs feature that integrates support for multiple external debuggers, with a primary focus on its graphical GDB interface. It streamlines the debugging workflow by allowing developers to set breakpoints directly in the source code fringe, step through execution, and inspect variables using dedicated windows and tooltips. The article demonstrates these capabilities with a simple C program and notes that future posts will explore GUD’s broader debugger support.
- Underappreciated builtin: Grand Unified Debugger — tusharhero.codeberg.page
Slint and the Node.js Event Loop
Slint 1.17 improves its Node.js binding on Linux and macOS by integrating the toolkit’s event loop directly with libuv. This optimization replaces an inefficient 16ms polling interval, reducing CPU usage and decreasing latency for JavaScript timers.
- Slint and the Node.js Event Loop — slint.dev
May in Servo: user scripts, mp4 compat, blackboxing in DevTools, and more
Servo 0.3.0 introduces several new web platform features, including enhanced font properties and expanded DOM APIs, while improving build compatibility for Windows, FreeBSD, and NixOS. The release also includes critical security patches to address memory safety vulnerabilities within the SpiderMonkey JavaScript runtime.
Diagrams for Penrose Tiles
This article details a Haskell implementation for rendering finite regions of non-periodic Penrose tilings using kite and dart tiles. Leveraging the diagrams package, the author introduces methods such as compChoices and decompPatch that utilize half-tiles to ensure mathematically legal tiling patterns.
- Diagrams for Penrose Tiles — readerunner.wordpress.com
Proving the Fundamental Theorem of Arithmetic in Agda
This article presents a complete, from-scratch formalization of the existence portion of the Fundamental Theorem of Arithmetic in Agda. Developed without relying on the Agda standard library, the work serves as an educational tutorial for intermediate learners.
- Proving the Fundamental Theorem of Arithmetic in Agda — byorgey.github.io
Rheo 0.4.0
Rheo 0.4.0 is a typesetting and static site engine that leverages the Typst markup language to compile a single source file into PDF, HTML, and EPUB formats. Designed for students, authors, and developers, it streamlines multi-format publishing by handling compilation, asset management, and linking while also supporting Atom feeds and RevealJS presentations. The release offers a simplified, cross-platform alternative to LaTeX and Markdown for producing consistent digital and print documents.
- Rheo 0.4.0 — rheo.ohrg.org
jj_tui: terminal user interface to jujutsu focused on speed and clarity
jj_tui is a high-performance, interactive terminal user interface designed for the Jujutsu version control system. It features real-time graph previews and intuitive keybindings to support essential operations such as committing, rebasing, squashing, and revset filtering.
No more Java refills for Intel Macs after JDK 27, says Oracle
Oracle has announced that Java support for Intel-based Macs will end following the release of JDK 27. This move comes as another platform prepares to wind down support following Apple’s transition away from Intel processors.
- No more Java refills for Intel Macs after JDK 27, says Oracle — theregister.com
🖥️ Hardware & Infrastructure
County with 37 Data Centers Asks Schools to ‘Conserve Electricity’
Henrico County, Virginia, is asking government employees and schools to conserve electricity following a 25% rate increase that is expected to add $5 million to the annual budget. This utility spike stems from the rapid expansion of 37 local data centers, which have overwhelmed the regional power grid and forced a reliance on expensive temporary generators. Officials hope these conservation measures will help offset the financial strain on residents and public services until permanent infrastructure upgrades are completed.
Have You Restarted Your Computer This Week?
Despite modern computers rarely requiring manual restarts, the author intentionally reboots their Mac weekly as a deliberate ritual. This low-risk practice, safeguarded by cloud synchronization, doubles as a standard troubleshooting method and a brief mental respite from daily productivity demands.
- Have You Restarted Your Computer This Week? — taonaw.com
DGX Spark vs. Mac Studio and Halo
NVIDIA’s $4,699 DGX Spark desktop AI workstation packs 128GB of unified memory, excelling at prompt processing and running massive 120B-parameter models that significantly outperform Apple’s Mac Mini M4 Pro. While its LPDDR5X memory bandwidth limits token generation speed compared to multi-GPU consumer setups, the system’s native CUDA integration and identical architecture to cloud clusters effectively bridge local prototyping with datacenter deployment. Recent software optimizations have further improved throughput, solidifying its role as a specialized tool for high-capacity local AI development.
- DGX Spark vs. Mac Studio and Halo — aimultiple.com
Microsoft reveals why Windows 11 keeps saying a file is in use after closing app
Microsoft explains that the persistent “file in use” error occurs when background processes, such as antivirus scans, network shares, or mapped DLLs, keep file handles locked even after the primary application closes. Users can identify and terminate these hidden locks using diagnostic utilities like the Sysinternals suite or PowerToys File Locksmith. Originally developed in the 1990s, these tools remain effective for safely managing file access and resolving locking issues on modern Windows systems.
- Microsoft reveals why Windows 11 keeps saying a file is in use after closing app — windowslatest.com
Lumo 2.0
Proton has unveiled Lumo 2.0, a major update to its encrypted AI assistant that introduces advanced reasoning models and multimodal capabilities for processing both text and images. The update features enhanced web search functionality and specialized modes for speed or complex tasks, all while maintaining strict zero-access encryption under Swiss privacy laws.
- Lumo 2.0 — proton.me
Antares Achieves Criticality of Mark-0 Reactor
Antares’ Mark-0 microreactor has achieved initial criticality at Idaho National Laboratory, marking the first time a private company has reached this milestone under the DOE Reactor Pilot Program. Conducted in partnership with the DOE and U.S. Army, the demonstration validates key reactor physics for future microreactor deployments to support military energy needs.
- Antares Achieves Criticality of Mark-0 Reactor — antaresindustries.com
Why Won’t Europe Build AI Data Centers in Iceland?
The European Union’s Tech Sovereignty Package aims to triple data center capacity to reduce dependency on American cloud providers. However, the initiative faces significant hurdles regarding energy availability and infrastructure, despite Iceland offering ideal conditions for AI hosting through renewable energy and natural cooling.
- Why Won’t Europe Build AI Data Centers in Iceland? — mrkt30.com
Xiaomi, Oppo, and Vivo cut 2026 shipment targets; Xiaomi cutting 30% to ~95M units
Major Chinese smartphone manufacturers, including Xiaomi, Oppo, and Vivo, are reducing their 2026 shipment targets due to rising production costs and component shortages. Some companies are slashing their outlook by up to 30%, with Xiaomi reportedly cutting its target to approximately 95 million units.
- Xiaomi, Oppo, and Vivo cut 2026 shipment targets; Xiaomi cutting 30% to ~95M units — asia.nikkei.com
Serving Local AI on my Jetson through Durable Streams
An author self-hosts the Kokoro-82M text-to-speech model on an NVIDIA Jetson Orin Nano Super to eliminate third-party dependencies and convert text into audio on demand. To overcome the latency and network instability inherent in incremental AI inference, the project replaces standard request-response APIs with a durable stream architecture. This system persists audio chunks as sequential records, enabling live playback, on-demand replay, and shareable links without requiring complex backend infrastructure.
No wheel, no pedals: Tesla unleashes Cybercab on the streets of Austin
Tesla’s Cybercab has been spotted testing on public roads in Austin, Texas, confirming its production design lacks a steering wheel and pedals. Unlike Tesla’s existing Model Y-based robotaxis that feature manual controls for human intervention, the Cybercab is engineered as a fully autonomous vehicle.
- No wheel, no pedals: Tesla unleashes Cybercab on the streets of Austin — cybernews.com
Arm64 on the desktop? It’s spendy and it’s sluggish
ARM64 desktop systems currently deliver sluggish performance and carry high costs, making them an impractical upgrade for most users. Even when paired with premium hardware like ample RAM, powerful GPUs, and fast storage, the architecture lacks the software optimization and overall value needed for a viable daily desktop experience.
- Arm64 on the desktop? It’s spendy and it’s sluggish — theregister.com
How is AI changing datacenter network fabrics?
Rapidly growing AI workloads are overwhelming traditional datacenter networking architectures, creating critical bottlenecks for modern compute infrastructure. To address this, the industry is accelerating the deployment of specialized high-bandwidth, low-latency network fabrics designed to handle the massive parallel data demands of AI clusters. This architectural shift is essential for maintaining the performance and scalability required by next-generation machine learning deployments.
- How is AI changing datacenter network fabrics? — theregister.com
đź’° Tech Business & Funding
Popping the GPU Bubble
Moondream’s Photon utilizes pipelined decoding to eliminate “GPU bubbles” caused by CPU housekeeping delays during sequential token generation. This technique increases efficiency by overlapping CPU and GPU tasks, launching the next forward pass while the previous step’s data is still being processed and committed.
- Popping the GPU Bubble — moondream.ai
Bandcamp has laid off most of their engineers
Bandcamp has laid off most of its remaining engineering staff. The news was shared by a long-term employee who confirmed their departure as part of the widespread layoffs.
- Bandcamp has laid off most of their engineers — bsky.app
Visa, BlackRock & 140+ Firms Launch Open USD Stablecoin Sharing Reserve Earnings
Over 140 major financial, tech, and crypto companies, including Visa, Mastercard, Stripe, BlackRock, and Coinbase, have partnered through Open Standard to launch Open USD, a new stablecoin designed for business payments. Expected to debut later this year with native support on the Tempo network, the token will feature shared governance and distribute most of its reserve earnings to participating firms. The initiative aims to establish an interoperable, fee-free payment asset for the broader global financial system.
Schneider Electric buys Cognite for $3.1B, plans Aveva merger
French industrial conglomerate Schneider Electric has agreed to acquire industrial AI firm Cognite in a $3.1 billion all-cash transaction. The company plans to integrate Cognite with its existing industrial software division, Aveva, to strengthen its industrial data and AI software portfolio. The acquisition is expected to be completed in the coming quarters.
- Schneider Electric buys Cognite for $3.1B, plans Aveva merger — bloomberg.com
Nebex raises $30M seed led by GV to connect US space tech suppliers and investors
Alphabet’s GV has led a $30 million seed funding round for Nebex, a new space fintech startup founded by former Axiom Space CEO Tejpaul Bhatia. The company aims to provide an online platform that connects US space technology providers with foreign governments and investors to facilitate industry transactions for a fee.
AI video startup Higgsfield seeks $300M-$500M raise at $5B valuation
AI video startup Higgsfield is reportedly in discussions to raise $300 million to $500 million at a $5 billion pre-money valuation. This potential funding round would more than quadruple the company’s worth from its January 2026 financing. The deal would provide substantial capital to further develop its AI-driven video generation platform.
- AI video startup Higgsfield seeks $300M-$500M raise at $5B valuation — theinformation.com
Microsoft’s PR agency cuts staff amid Xbox restructuring
The PR agency Assembly is reportedly laying off an unknown number of employees as part of its own internal reorganization. While there was initial speculation that the cuts were linked to recent budget reductions at Xbox, journalist Jason Schreier clarified that the layoffs are unrelated to any changes in the agency’s partnership with Microsoft.
- Microsoft’s PR agency cuts staff amid Xbox restructuring — cybernews.com
UK Regulator Challenges Apple And Google App Fees
The UK’s Competition and Markets Authority has proposed regulations requiring Apple and Google to allow developers to direct users toward cheaper third-party payment options, bypassing mandatory app store commission fees. Aiming to reduce consumer costs and stimulate market competition, the proposal also considers mandating that Apple open its NFC technology for third-party contactless payments within iOS apps.
- UK regulator wants Apple and Google to let devs steer clear of app store fees — theregister.com
- UK regulator plans to let Apple and Google users choose cheaper payment options in app stores — cybernews.com
📜 Policy & Society
The labor share of income in the US is at its lowest post-war level
The U.S. labor share of income has fallen to its lowest post-war level following a sharp post-pandemic decline. Analysis shows this drop mirrors the cyclical patterns and sectoral shifts typical of pre-2000 recessions, indicating it is a standard business cycle fluctuation rather than a new structural trend. As a result, a full recovery will likely depend on a prolonged period of economic expansion.
- The labor share of income in the US is at its lowest post-war level — libertystreeteconomics.newyorkfed.org
Supreme Court upholds broad conception of birthright citizenship
The Supreme Court upheld the 14th Amendment’s guarantee of birthright citizenship, rejecting President Trump’s executive order that sought to deny citizenship to children of undocumented or temporary visa holders. The majority opinion warned that overturning this longstanding precedent would effectively revive the discredited Dred Scott decision. Despite the ruling, the Justice Department pledged to continue enforcing laws against alleged “birth tourism” schemes.
Memoirs of Extraordinary Popular Delusions and the Madness of Crowds
Charles Mackay’s 1841 work, Memoirs of Extraordinary Popular Delusions and the Madness of Crowds, explores humanity’s susceptibility to collective manias such as financial bubbles and religious crusades. Through various historical anecdotes, the book provides an influential analysis of crowd psychology and its role in economic speculation and market crashes.
- Memoirs of Extraordinary Popular Delusions and the Madness of Crowds — gutenberg.org
American Pride Falls to 25-Year Record Low
A recent Gallup poll reveals that extreme pride in being American has fallen to a 25-year low of 33%. This decline is accompanied by a stark partisan divide, with 70% of Republicans reporting extreme pride compared to only 14% of Democrats.
- American Pride Falls to 25-Year Record Low — news.gallup.com
Should every baby’s DNA be sequenced?
Baby Freddie was diagnosed with retinoblastoma, a rare and aggressive eye cancer, at just four weeks old. Thanks to early detection and treatment involving chemotherapy and laser therapy, he has a significantly improved chance of maintaining normal eyesight.
- Should every baby’s DNA be sequenced? — economist.com
Exercise intensity influences body composition in healthy older adults (2025)
A six-month study investigated how high-intensity interval, moderate-intensity continuous, and low-intensity training impact the body composition of healthy older adults. The researchers aimed to determine which exercise intensity most effectively reduces fat mass and visceral adipose tissue to help mitigate age-associated health risks.
The US ambassador had Belgian police stop our reporting
Journalists from The European Correspondent were forcibly removed from a U.S.-sponsored “Freedom 250” celebration in Brussels by Belgian police. The removal occurred after the reporters attempted to question Ambassador Bill White, following instructions from the U.S. embassy that labeled one journalist an “active threat.”
- The US ambassador had Belgian police stop our reporting — europeancorrespondent.com
A Berkeley AI professor makes provocative argument for decelerating AI research
UC Berkeley AI researcher Emma Pierson has sparked debate by arguing that the rapid development of generalist AI models poses immediate societal risks, such as inequality and surveillance, that may outweigh potential medical benefits. Her stance has drawn significant criticism from AI accelerationists, who contend that decelerating research is inhumane because it delays life-saving scientific breakthroughs.
- A Berkeley AI professor makes provocative argument for decelerating AI research — fastcompany.com
Study suggests most Americans would be healthier without daylight saving time
A Stanford Medicine study suggests that eliminating biannual time shifts in favor of either permanent standard or daylight saving time would better support human circadian rhythms. The researchers found that adopting permanent standard time would provide the greatest health benefits, potentially preventing hundreds of thousands of strokes and millions of cases of obesity.
- Study suggests most Americans would be healthier without daylight saving time — med.stanford.edu
Microsoft’s Ireland unit hit $47B pretax (38% of total) as EU mandates country-by-country reporting
Under new European Union regulations mandating public country-by-country tax reporting, Microsoft disclosed that its Ireland hub generated $47 billion in pretax profits for FY 2025, representing 38.1% of its global total. This concentration yields over $7 million in pretax profit per employee, more than 13 times the company’s worldwide average. The mandatory filings provide investors and the public with unprecedented transparency into multinational corporate tax strategies.
- Microsoft’s Ireland unit hit $47B pretax (38% of total) as EU mandates country-by-country reporting — wsj.com
Trump administration discusses SpaceX share donations to Trump Accounts for children
The Trump administration has held discussions with SpaceX regarding the potential donation of company stock to “Trump Accounts,” a tax-advantaged savings program for children launching next week. Such a contribution could signal a reconciliation between Elon Musk and Donald Trump following their recent professional fallout.
- Trump administration discusses SpaceX share donations to Trump Accounts for children — semafor.com
Amazon sued over Australian Prime Video pricing
The Australian Competition and Consumer Commission (ACCC) is suing Amazon Australia for allegedly using unfair contract terms within its Prime Video subscriptions. The regulator claims the company introduced advertisements to the platform and required annual members to pay an additional A$2.99 per month to maintain an ad-free experience.
- Australians paid A$2.99 extra for ad-free Prime Video, and Amazon is now being sued over it — cybernews.com
- Australia’s ACCC sues Amazon over alleged unfair Prime Video ad terms and price hikes — bloomberg.com
UK FCA unveils updated crypto framework after softening rules criticized as holding UK back
The UK Financial Conduct Authority (FCA) has updated its cryptocurrency regulatory framework by softening previously proposed rules. These changes ease capital and disclosure requirements following industry concerns that the initial regime was too onerous and risked undermining the UK’s competitiveness.
- UK FCA unveils updated crypto framework after softening rules criticized as holding UK back — ft.com
stop asking writers about “AI”
A writer expresses frustration over being repeatedly asked for their opinion on AI, arguing that such inquiries equate human creativity with machine-generated content. The author contends that these questions devalue professional authorship and have significantly impacted their motivation to write.
- stop asking writers about “AI” — benjaminhollon.com
New EU proposal aims to stop drivers from speeding using satellites
The European Commission is exploring a proposal to use satellite and GPS technology to automatically reduce vehicle engine power when drivers exceed speed limits. If approved, this system could become mandatory for all new cars in the EU by 2030, aiming to reduce road fatalities by up to 20%.
- New EU proposal aims to stop drivers from speeding using satellites — cybernews.com
Atlas shrugs: New UK asylum seeker IT system failed to help case workers learn from appeals
The UK’s new Atlas IT system for asylum seekers has gone live despite several development delays. Inspectors report that the system launched without essential functionality required by Home Office case managers to learn from appeals.
- Atlas shrugs: New UK asylum seeker IT system failed to help case workers learn from appeals — theregister.com
đź”’ Security & Privacy
Beware, Claude Code deletes >30 day old transcripts. Anthropic won’t fix it
Claude Code’s default settings silently delete conversation transcripts older than 30 days on startup, causing users to lose critical reasoning context without warning. Anthropic has refused to modify this behavior despite requests for safer defaults and explicit opt-in disclosures. Users are advised to manually configure a longer retention period in their settings to prevent further data loss.
PDFMergely – In-browser PDF tools that never upload your files
PDFMergely is a free, browser-based PDF toolkit that processes documents entirely locally on a user’s device, guaranteeing zero server uploads and complete privacy. Leveraging WebAssembly and Web Workers, the platform delivers a comprehensive suite of merging, editing, and conversion tools that operate instantly and function fully offline. By eliminating account requirements and cloud dependencies, it offers a secure alternative for handling sensitive files.
- PDFMergely – In-browser PDF tools that never upload your files — pdfmergely.com
ICE Tracks Down Woman to Force Her to Delete Instagram Post
Two ICE agents confronted a poll worker at a Syracuse, New York, polling station on Election Day to demand the removal of social media posts they alleged threatened federal personnel. The confrontation took place despite election officials stating that law enforcement presence is generally prohibited in polling places unless responding to an emergency.
- ICE Tracks Down Woman to Force Her to Delete Instagram Post — newrepublic.com
Soatok’s Informal Guide to Threat Models
This informal guide outlines threat modeling as a practical cybersecurity practice for identifying protected assets, adversaries, attack vectors, and corresponding defenses. Effective models require mapping asset relationships, explicitly documenting underlying assumptions, and acknowledging deliberately excluded risks. Ultimately, they should function as living documents that are continuously updated as systems and operating environments evolve.
- Soatok’s Informal Guide to Threat Models — soatok.blog
Furality Ultra Club A/V Writeup
This technical writeup details the A/V production workflows used for Furality Ultra, specifically focusing on controlling club visuals and lighting through video-encoded DMX data. The author explains how tools like Furality Gridnode and Blender were utilized to manage complex world features, including multi-axis screens and automated light robots.
- Furality Ultra Club A/V Writeup — value.gay
AirDrop and Quick Share vulnerable: Nearby hackers can crash your devices
Researchers have identified six security vulnerabilities in Apple’s AirDrop and Android’s Quick Share that could allow nearby attackers to crash devices or potentially execute remote code. These flaws affect billions of iOS and Android smartphones, exposing users to risks such as denial-of-service attacks from within a 30-meter range.