Alvaro Lopez Ortega / 2026-02-24 Briefing

Created Wed, 25 Feb 2026 02:09:51 +0000 Modified Mon, 13 Apr 2026 13:43:18 +0000
9667 Words

A new protocol, RAGS, is being implemented to filter out low-quality, AI-generated content across projects, instructing AI systems to halt and return errors upon rejection. Security researchers have uncovered a vulnerability, “Parsing Differentials,” that allows attackers to bypass HTML sanitizers, injecting malicious code despite implemented filters. Cloudflare engineers rapidly developed vinext, a Next.js replacement utilizing AI, for faster, more efficient serverless deployments. Concerns are rising about a potential overemphasis on AI within software engineering research, neglecting crucial areas like system reliability and architectural resilience after recent widespread cloud outages.

πŸ€– AI & Machine Learning

Next.js Rebuild: AI-Powered Development

Cloudflare engineers have created vinext, a Next.js replacement built on Vite, to improve performance and simplify deployments to serverless platforms. The project, developed in just one week with AI assistance for around $1,100, boasts significantly faster build times and smaller file sizes while maintaining Next.js compatibility. Vinext is already in production use and aims to address limitations encountered when adapting Next.js for broader serverless deployments.

AI Functionality Updates

AWS has launched AI Functions, dubbed “Software 3.1,” an experimental library enabling Large Language Models to generate and execute code directly within applications. This innovative system returns native Python objects and employs automated verification to continuously ensure code correctness, removing the need for manual review. Developers can define function behavior through natural language, receiving automatically generated and validated code via the Strands Agents SDK.

How I got Claude to teach me dbt

The author experimented with using the AI language model Claude as a personalized “Instructor” to learn dbt, a data transformation technology. This approach combined interactive explanations, troubleshooting assistance, and a tailored learning environment, mimicking the benefits of tutorials, courses, and support systems. The author believes this AI-powered tutoring has the potential to revolutionize self-paced learning.

Amazon would rather blame its own engineers than its AI

Following an AWS outage affecting Cost Explorer in China, Amazon has blamed human error and overly broad permissions, avoiding responsibility for potential issues with its AI coding tool, Kiro. The company’s response appears intended to protect the image of its AI technology, prioritizing it over acknowledging engineering mistakes. This incident highlights a reluctance to attribute problems to AI amid growing competition in the field.

Intel backs SambaNova’s $350M bid to challenge GPUs in AI inference

SambaNova has secured $350 million in funding, including a significant investment from Intel, to develop its dataflow architecture as an alternative to GPUs for AI inference. Intel and SambaNova are collaborating on a multi-year project to integrate SambaNova’s RDUs with Intel Xeon CPUs, aiming to provide a more cost-effective AI solution. The new SN50 accelerator is designed to offer improved performance compared to Nvidia’s offerings, leveraging a dataflow architecture to reduce data movement overhead.

Amazon AGI Lab Head David Luan Departs After Short Tenure

David Luan, head of Amazon’s AGI lab, is departing the company after less than two years, following Amazon’s acquisition of his startup Adept. He cited a desire to focus entirely on developing new AI capabilities. His departure follows a recent reorganization of Amazon’s AGI division and is facing scrutiny from regulators.

Anthropic Updates Scaling Policy, Separates Safety Commitments

Anthropic has revised its Responsible Scaling Policy, removing a key pledge to only train AI models when safety can be guaranteed. The company now prioritizes transparency and matching industry safety practices to remain competitive in the rapidly evolving AI landscape. This shift reflects Anthropic’s recent commercial success and acknowledges the lack of comprehensive AI regulation.

Waymo Robotaxis Now in Dallas, Houston, & More Cities

Waymo has expanded its robotaxi service to Dallas, Houston, San Antonio, and Orlando, bringing its total number of U.S. operating cities to ten. The company is initially offering rides to select users through its app, with broader availability planned by late 2026. This expansion strengthens Waymo’s position in the autonomous vehicle market as it faces competition from rivals.

AI Agent Startup Nimble Raises $47M Series B

Nimble, a startup using AI agents to structure real-time web data into queryable tables, has raised $47 million in Series B funding. Their platform validates web data and integrates with existing data warehouses, addressing limitations of current AI systems by providing reliable and accessible information. This technology enables businesses to leverage web data within their current data infrastructure for tasks like competitor analysis and pricing research.

New AI Models Released, Improving Performance

Inception, an AI startup, has launched Mercury 2, a new language model leveraging diffusion technology to achieve significantly faster and more cost-effective text interaction compared to competitors. Simultaneously, a separate startup released open-weight speech-to-text models that surpass OpenAI’s WhisperLargev3 in accuracy, demonstrating competitive results with limited resources. These advancements highlight progress in both language modeling speed and speech recognition capabilities within the AI space.

AI Model Training Hurdles: Parsing Trillions of Tokens From PDFs

Despite advances in AI, the PDF format poses a significant challenge for extracting usable data due to its design prioritizing visual preservation over machine readability. Companies like Reducto are developing specialized AI to parse these difficult files, enabling the creation of searchable applications and unlocking insights from previously inaccessible documents. This technology is crucial for tasks such as analyzing large document sets related to figures like Jeffrey Epstein.

Proposal: Add “AI generated” as a flag reason

A proposal suggests adding a specific “AI generated” flag to better identify and address the growing presence of AI-written content on the platform. Currently, users rely on imprecise methods like spam flagging and the “vibecoding” tag, leading to confusion and hindering quality control. This new flag aims to streamline content identification and educate users on recognizing AI-generated text.

OpenAI resets spending expectations, from $1.4T to $600B

OpenAI has lowered its projected compute spending to $600 billion by 2030, a significant decrease from previous estimates, aligning it with anticipated $280 billion in revenue. The company is finalizing a funding round exceeding $100 billion with investments from Nvidia, SoftBank, and Amazon, while its chatbot ChatGPT now has 900 million weekly active users. This revised spending plan comes as OpenAI faces increasing competition and seeks to refine its growth strategy.

Pilo – open-source agentic web automation engine by Mozilla

Mozilla has open-sourced Pilo, a new engine powering its Tabstack automation platform. Pilo uses natural language goals to autonomously navigate web pages, relying on an accessibility tree and context compression to improve efficiency and stability. The project is available on GitHub and includes a browser extension for real-time debugging.

AI-generated replies really are a scourge these days

The increasing prevalence of AI-generated responses online is causing frustration and confusion. Writer Simon Willson highlights the difficulty in distinguishing between genuine human replies and those created by artificial intelligence. This proliferation of AI content is being described as a “scourge” due to its impact on online communication.

Firefox 148 Launches with AI Kill Switch Feature and More Enhancements

Firefox 148 now includes an “AI kill switch” allowing users to permanently disable AI features and prevent future re-enablement. The update also provides users with increased control over remote updates and data collection, alongside various enhancements to web platform capabilities and accessibility. Detailed release notes are available for those seeking more information.

Blood test boosts Alzheimer’s diagnosis accuracy to 94.5%, clinical study shows

A new clinical study demonstrates that a blood test measuring the p-tau217 protein significantly improves the accuracy of Alzheimer’s diagnosis, raising it from 75.5% to 94.5%. The test also increased clinicians’ confidence and, in some cases, led to changes in initial assessments. This offers a potentially less invasive and more accessible diagnostic option compared to current methods.

I turned off ChatGPT’s memory

The author argues against using ChatGPT’s memory feature, introducing the concept of “context rot” where accumulated errors degrade results. He illustrates this with examples of ChatGPT misinterpreting past instructions, leading to unpredictable and sometimes humorous outputs. Ultimately, disabling memory ensures more unbiased and consistent responses from the AI.

NIST Seeking Public Comment on AI Agent Security (Deadline: March 9, 2026)

The National Institute of Standards and Technology (NIST) is seeking public comment on AI agent security. The Federal Register is displaying a prototype “Web 2.0” version, which is currently an unofficial informational resource and lacks legal authority. Comments must be submitted by March 9, 2026.

Steerling-8B, a language model that can explain any token it generates

Steerling-8B is a new language model notable for its inherent interpretability, allowing users to trace each generated token back to its origins in input context, understandable concepts, and training data. This smaller model achieves competitive performance while enabling capabilities like concept-level steering and training data provenance, offering unprecedented control and transparency in its output. The model’s architecture achieves this by decomposing embeddings into explicit pathways for both known and discovered concepts.

ChatGPT finds an error in Terence Tao’s math research

Mathematician Terence Tao acknowledged that ChatGPT identified a significant error – a sign error – in his research regarding small primes. He subsequently corrected the mistake by referencing a paper by Hildebrand and utilizing a specific inequality related to the Dickman function. Tao confirmed that this approach resolved the issue and repaired his argument.

βš™οΈ Development & Tools

ansigpt: c89 implementation of microgpt

A developer has created an ANSI C implementation of Karpathy’s microGPT, a simplified language model. The project, built primarily in Vim, requires manual data download and comes with disclaimers about the code’s complexity and development process. It was partially debugged with the assistance of LLMs.

Introducing jjq, a local merge queue for jj

jjq is a new command-line tool designed to provide a local merge queue system for the jj version control tool. It automates testing and updates a “trunk” commit, enabling teams to manage concurrent development and maintain a stable codebase. This local system offers an alternative to centralized forge merge queues, aiming to increase development speed while minimizing integration issues.

Charm v2: Major releases for Bubble Tea, Lip Gloss, and Bubbles for terminal UIs in Go

Bubble Tea, Lip Gloss, and Bubbles have released version 2.0.0, marking the end of their beta phases and introducing significant performance improvements for terminal user interfaces. These updates, which have been used in the company’s AI coding agent, optimize rendering and compositing for increased efficiency. The new ecosystem aims to support the growing demands of AI and coding tools within the terminal environment.

Extending C with Prolog

The integration of Prolog and C allows developers to combine the strengths of both languages. Prolog excels in pattern matching and search due to its dynamic memory management and unification capabilities, which are advantageous for symbolic comparisons. This hybrid approach enables more efficient solutions for complex projects that require both procedural and declarative programming elements.

TLA+ By Example

This resource provides a practical introduction to the TLA+ specification language. It features a tutorial on TLA+ syntax, data structures, and model checking, alongside a detailed walkthrough of a blocking queue example. The tutorial builds upon existing work by Markus Kuppe.

Reducing the size of Go binaries by up to 77%

Datadog engineers significantly reduced the size of Go binaries in their Agent software, achieving reductions of up to 77% between versions 7.60.0 and 7.68.0. This was accomplished through dependency auditing, code refactoring, and linker optimizations, reversing previous growth without removing features. The improvements also led to enhancements in the Go compiler and linker, benefiting other projects.

Rust debugging survey 2026

The Rust community is conducting a survey to better understand and address the challenges developers face when debugging Rust code. The anonymous survey, open until March 13, 2026, aims to gather data to improve debugging support across debuggers, operating systems, and code types. Results will inform future improvements and create a more robust and consistent debugging experience.

Racket v9.1

Racket v9.1 has been released, featuring improvements to documentation navigation, the DrRacket IDE with new visual elements, and several performance enhancements. The update also includes library updates like OpenSSL and Typed Racket, along with numerous bug fixes and documentation improvements. Users are encouraged to migrate from version 9.0 and contribute to the open-source project.

Where Do Specifications Fit in the Dependency Tree?

Software projects often struggle to manage runtime dependencies, with critical language and compiler specifications frequently overlooked. This lack of standardized tracking leads to build failures and debugging difficulties, as dependencies are declared inconsistently across different tools and files. Package managers like Spack demonstrate the importance of explicitly declaring these underlying dependencies to improve reliability and resolution.

Desk Setup: From Multi-Monitor to Single Screen with Virtual Desktops

The author has refined their computer desk setup over time, ultimately opting for a single 27-inch monitor paired with virtual desktops accessed via keyboard shortcuts. This streamlined approach prioritizes focus and efficient task switching, often utilizing a laptop screen for video conferencing. The current setup was inspired by others, including ThePrimeagen, and incorporates various productivity tools.

Remote Terminal Access via WebRTC

New tools are emerging to simplify accessing and distributing terminal applications. Trolley offers a streamlined way to package and run text-based interfaces (TUIs) across Linux, macOS, and Windows, while a new iPhone app leverages WebRTC for secure, remote access to Mac terminals with AI coding assistance. Both solutions aim to enhance the accessibility and development experience for terminal-based workflows.

Go library maintainer brands GitHub’s Dependabot a ’noise machine'

Go library maintainer Filippo Valsorda has criticized GitHub’s Dependabot, calling it a “noise machine” due to excessive and inaccurate security alerts. He argues the tool generates false positives, hinders proper security practices, and advocates for developers to disable it in favor of more targeted analysis. Valsorda also expressed concerns about Dependabot’s automatic dependency updates, recommending controlled upgrades instead.

Microsoft teases β€˜reimagined SharePoint experience’ landing in April

Microsoft will release a redesigned SharePoint experience in April, emphasizing simplified knowledge discovery and AI integration. Users will also gain the ability to customize their OneDrive folder names for better organization and shorter file paths. The company has also resolved an issue affecting the recognition of Advanced Security add-ons for the Business Premium + Copilot bundle.

Pi – a minimal terminal coding harness

Pi is a new, minimal terminal coding harness designed for flexible workflows and customization. Users can extend Pi’s functionality with TypeScript extensions, skills, and prompt templates, allowing for tailored AI interactions and feature development. It supports numerous AI models, stores sessions as navigable trees, and prioritizes user-defined features over pre-built solutions.

LibreOffice Online: A Fresh Start

The Document Foundation has restarted the LibreOffice Online project, aiming to create a community-driven, web-based version of the office suite. The project’s repository will soon reopen for contributions, though users should anticipate an early and potentially unstable state. The foundation will provide the technology but will not offer hosting or enterprise support.

Tag Promptless on any GitHub PR/Issue to get updated user-facing docs

Promptless, an AI tool that automatically drafts documentation updates, has launched version 1.0 with improved features like automatic screenshot updates and more natural-sounding writing. Users can now easily request doc updates by tagging @promptless on open-source GitHub repositories, and the tool is being offered for free to CNCF/Linux Foundation projects. The platform aims to streamline documentation processes by monitoring workflows and suggesting relevant updates.

Why the KeePass format should be based on SQLite

KeePass, a popular password manager, is facing limitations due to its aging XML-based KDBX file format. The rigid structure hinders the seamless integration of modern security features, often forcing developers to rely on workarounds and custom attributes, leading to fragmentation and compatibility issues. Switching to a more flexible format like SQLite could resolve these problems and improve the user experience.

SNKV – SQLite’s B-tree as a key-value store (C/C++ and Python bindings)

SNKV is a new library that allows developers to utilize SQLite’s B-tree engine directly for key-value storage, bypassing the SQL parsing and planning layers. This approach offers significant performance improvements for read-heavy workloads compared to standard SQLite, while still providing ACID properties and concurrency. The library is available in both Python and C/C++ implementations and is designed as a drop-in replacement for certain use cases.

Decimal-Java is a library to convert java.math.BigDecimal to and from IEEE-754r

The decimal-java library facilitates conversions between Java’s BigDecimal and IEEE-754r decimal byte representations (decimal32, decimal64, and decimal128). It provides encoding and decoding methods with overflow handling options, and is designed for Java 17 and later. The library is stable, modularized, and licensed under the MIT license.

πŸ›‘οΈ Security & Privacy

Parse Me, Baby, One More Time: Bypassing HTML Sanitizer via Parsing Differentials (2024)

Researchers have discovered a new vulnerability allowing attackers to bypass HTML sanitizers by exploiting differences in how various parsers interpret the same HTML code. This technique, dubbed “Parsing Differentials,” can enable the injection of malicious code even when sanitization is implemented. The finding highlights the complexity of creating truly secure HTML sanitization filters and poses a risk to web applications relying on them.

Goodbye innerHTML, Hello setHTML: Stronger XSS Protection in Firefox 148

Firefox 148 introduces the new standardized Sanitizer API, a tool designed to simplify and improve cross-site scripting (XSS) protection for developers. This API offers a safer default for HTML insertion by sanitizing untrusted content and is compatible with Trusted Types for enhanced security. Replacing innerHTML assignments with the new setHTML() method allows for easier XSS prevention with minimal code changes.

Patch these 4 critical, make-me-root SolarWinds bugs ASAP

SolarWinds’ Serv-U file transfer software has four critical vulnerabilities that could allow attackers to execute code as root. Patches are available in version 15.5.4 to address issues including broken access control and insecure direct object references. While not currently exploited, SolarWinds products are frequent targets, emphasizing the importance of immediate patching.

North Korea’s Lazarus Group targets healthcare orgs with Medusa ransomware

North Korean state-sponsored hackers, the Lazarus Group, are now employing the Medusa ransomware to target healthcare organizations in the US and Middle East. This marks a shift in their tactics, potentially aimed at generating revenue for the North Korean regime, and follows their previous use of other ransomware strains. Security researchers have linked the attacks to Lazarus Group through technical indicators, though definitive attribution remains complex.

Korean cops charge teens over bike hire breach that exposed data on 4.62M riders

Two South Korean teenagers have been charged with hacking Seoul’s public bike service, Ttareungyi, and stealing data on 4.62 million users. The breach, discovered during an investigation into a separate DDoS attack, exposed sensitive personal information like addresses and phone numbers. While authorities believe the teens intended to profit, there’s no evidence the data was leaked or sold.

West Midlands Police earn red card over Copilot’s imaginary football match

A parliamentary committee found that West Midlands Police improperly used Microsoft Copilot, an AI chatbot, to generate false information about a football match, which influenced a decision to ban fans. The report criticized the force’s lack of oversight and transparency in utilizing AI, leading to the former chief constable’s retirement. This incident highlights concerns about the expanding use of AI in policing and the need for caution and verification.

Ex-L3Harris Exec Sentenced to 7+ Years for Trade Secret Theft

A former L3Harris executive, Peter Williams, was sentenced to over seven years in prison for stealing and selling zero-day exploits to a Russian broker. Williams pleaded guilty to theft of trade secrets and admitted to selling the stolen information for millions of dollars in cryptocurrency. The U.S. Treasury also sanctioned Operation Zero, the Russian brokerage involved in the transaction.

ADT to Buy Origin Wireless for $170M, Boosts AI Security

ADT has acquired Origin Wireless for $170 million to integrate advanced AI motion classification into its home security systems. Origin’s technology uses existing Wi-Fi signals to identify the type of motion detected, differentiating between people, pets, and other objects. The integration, slated to begin in 2027, will enhance ADT’s security offerings without requiring additional hardware.

Meta Execs Warned Encryption Could Hinder CSAM Reporting

Meta executives pushed forward with end-to-end encryption for Facebook and Instagram messaging despite internal warnings that the feature would impede the company’s ability to identify and report child exploitation material to law enforcement. Internal documents revealed concerns about the impact on safety operations and potential reduction in harmful content reporting. New Mexico’s Attorney General alleges Meta misled the public regarding the safety of this encryption implementation, and a jury trial is currently underway.

Every privacy concession in history has been permanent

The article cautions against accepting privacy compromises, arguing that historical concessions are permanent and erode privacy baselines. It cites examples like the Bank Secrecy Act and the Patriot Act to illustrate that once privacy is relinquished, it’s rarely regained. The author urges a long-term perspective on privacy loss, rather than focusing solely on short-term benefits.

Cell Service for the Fairly Paranoid

Cape is a new mobile carrier prioritizing user privacy and security through features like data minimization, identifier rotation, and encrypted texting. The company partners with Proton to provide additional privacy tools and offers nationwide 4G and 5G coverage. Subscribers can try the service for an introductory price of $30.

Wealthy spouses are hiding crypto assets in divorce cases, say lawyers

Lawyers are observing a rising trend of wealthy individuals concealing cryptocurrency assets during divorce proceedings. This practice hinders accurate asset assessment and complicates property division, requiring legal professionals to be more diligent. The Financial Times reports on this growing issue impacting divorce cases.

enveil – hide your .env secrets from prAIng eyes

Enveil is a new tool designed to protect sensitive environment variables from being exposed to AI coding assistants. It replaces plaintext secrets in .env files with symbolic references and encrypts the actual values, decrypting them only when needed. The tool utilizes strong encryption and a master password to ensure secure storage and prevent unauthorized access.

Scammers have the advantage: no one can reliably detect AI faces, study finds

A new study reveals that current methods are unreliable at distinguishing between real human faces and those generated by artificial intelligence. This makes it increasingly difficult to detect AI-generated faces, potentially benefiting scammers and malicious actors. The findings highlight a growing challenge in verifying online identities and combating fraud.

πŸ’» Systems & Infrastructure

Sometimes it’s just better to load “all” the data

Loading data in larger batches can significantly improve performance, especially in batch processing. A real-world example demonstrated a reduction from a 24-hour process to just 15 minutes by grouping and loading data daily instead of processing individual items. This technique, while context-dependent, can be applied to other areas like HTTP requests and file I/O to minimize round trips and enhance application speed.

Scheduling in a Bare-Metal Web Server

The Tatix web server kernel utilizes a specialized scheduler to efficiently handle multiple concurrent client requests. This scheduler avoids blocking the server during data transmission to one client by strategically pausing execution and switching between connections. The design prioritizes simplicity and eliminates the need for explicit locking.

How macOS controls performance: QoS on Intel and M1 processors (2022)

macOS utilizes Quality of Service (QoS) settings to prioritize processes, with values ranging from 9 to 33. While QoS has a minimal impact on Intel Macs, it significantly affects performance on M1 chips by directing tasks to different processor cores. Currently, users have limited control over QoS settings within applications.

Row Locks With Joins Can Produce Surprising Results in PostgreSQL

A PostgreSQL edge case can cause join queries to unexpectedly return no results due to row locks. This occurs when concurrent sessions attempt to update the same record, and one session’s lock prevents another from seeing updated data during a SELECT ... FOR UPDATE operation. The issue can lead to seemingly valid join queries failing to retrieve expected data.

Permacomputing

Permacomputing is a community-driven, anti-capitalist movement inspired by permaculture, seeking to fundamentally rethink computer technology and its impact on the environment and society. The project prioritizes collective action and technical design over readily available technological solutions, encouraging a radical reimagining of computational culture. It is not a product or a quick fix, but an invitation to participate in a broader, transformative effort.

GhostBSD to ditch Xorg for XLibre as Red Hat’s Wayland crusade leaves X11 fans out in the cold

GhostBSD, a FreeBSD-based distribution, will switch to the XLibre X11 server due to concerns about the decline of Xorg and the increasing adoption of Wayland in the Linux ecosystem. This change aims to ensure continued compatibility with GhostBSD’s desktop environments, including MATE, Xfce, and Gershwin, and avoid potential software update issues. The decision highlights the broader impact of Linux development trends on BSD-based systems.

KDE Plasma 6.6 isn’t forcing systemd but the arguments rage on

KDE Plasma 6.6 has been released with functional improvements and enhanced OpenBSD support, while maintaining a commitment to not requiring systemd for the core desktop environment. FreeBSD 15.1 will now offer KDE Plasma as an installation option, improving BSD desktop usability. Despite KDE’s clarifications, debates regarding systemd dependencies continue within the community.

Cisco turns to titanium spoons and sand dunes to build a better … box?

Storage company Pure Storage has rebranded to “Everpure” to reflect a broader focus on data management. Simultaneously, Cisco redesigned its Room Kit Pro G2 device using biomimicry, creating a lighter and stronger enclosure inspired by titanium spoons and sand dunes. The redesign significantly reduces material usage, eliminating the equivalent mass of multiple vintage Volkswagen Beetles.

Apple to Make Mac Minis in Houston by 2026

Apple will shift Mac Mini production from Asia to a Foxconn facility in Houston, Texas, beginning later in 2026. The move will utilize a newly converted 220,000 square foot warehouse within the existing Foxconn site. This is part of Apple’s commitment to invest $600 billion in the U.S. over four years and reshore its supply chain.

Manjaro website off-line again due to lapsed certificate

The Manjaro website is currently offline due to an expired security certificate. This recurring issue has previously occurred in 2015 and 2022, often resulting from automated script errors or network problems. The website administrators are working to resolve the certificate renewal failure.

Intel XeSS 3: expanded support for Core Ultra/Core Ultra 2 and Arc A, B series

Intel has released graphics driver version 32.0.101.8509, expanding support for Intel Arc GPUs (A and B series) and Core Ultra processors with integrated graphics. The update includes enhancements to XeSS 3 and extends compatibility to newer platforms like Battlemage and Lunar Lake/Arrow Lake. Note that installing this generic driver will replace any customized OEM drivers.

πŸ“° Business & Policy

Dear researchers: Is AI all you’ve got?

A recent article argues that software engineering research is overly focused on AI, potentially neglecting crucial areas like system reliability and human-centric challenges. Recent widespread cloud outages, costing billions, highlight the need for research into architectural resilience and other neglected areas. The author draws parallels to past research trends, warning of a risk in overlooking fundamental, unsolved problems in software engineering.

FedCM + IdP Registration β€” Call to action

The article encourages developers and communities to adopt Federated Credential Management (FedCM) for decentralized web sign-in, specifically promoting Identity Provider (IdP) registration within the framework. Resources are provided to help developers implement FedCM, including guides for IndieAuth and OAuth. The initiative seeks real-world use cases to demonstrate the feature’s value and drive its adoption.

Coile Blogs Flying with Software Legend Wirth

Software pioneer Niklaus Wirth recently flew with Brantley Coile, known for his work on PIX firewalls and ATA-over-Ethernet. Coile documented their flights in a four-part blog series, detailing experiences ranging from gusty conditions to a long takeoff roll. The posts recount the logistics and aerial adventures of transporting Wirth between Oklahoma and Albuquerque.

Minor Error Sparks Debate Over Accuracy

Early Apple computers included a “breathing light,” a pulsing LED initially designed to indicate system status and later mimicking human respiration for a comforting effect. This unique detail, intended to reassure users, was a distinct element of Apple’s design philosophy before being discontinued in the 2010s. Though it wasn’t a major sales driver, the breathing light remains a fondly remembered feature.

Apple Reshuffles Production, Bringing Jobs Home

Apple is expanding its U.S. manufacturing footprint by initiating Mac Mini production in Texas and partnering with TSMC to build a major chip-manufacturing facility in Arizona. These initiatives, representing a significant investment exceeding $165 billion, aim to secure tariff exemptions, create domestic jobs, and reduce reliance on foreign suppliers. While Mac Minis for international markets will still be produced in Asia, these moves mark a substantial effort to bolster U.S. tech manufacturing.

Right to Repair Advances in Iowa

Iowa is poised to pass a “Right to Repair” bill granting farmers access to tools, information, and parts needed to repair their own agricultural equipment, challenging manufacturer restrictions. The legislation seeks to reduce repair costs and downtime for farmers, potentially saving billions annually and addressing a growing national movement. This follows similar laws passed in other states and would significantly impact Iowa’s agricultural sector.

AMD copy-pastes 6 GW chips-for-stock deal in new Meta agreement

AMD has signed a significant agreement with Meta to supply 6 gigawatts of custom AI GPUs and sixth-generation EPYC CPUs, mirroring a previous deal with OpenAI. The agreement includes a warrant for Meta to purchase AMD shares contingent on performance milestones, including a stock price of $600. This deal is expected to generate substantial revenue for AMD and further its position in the AI chip market.

Microsoft gives Windows laggards the ‘gift of time’ wrapped in licensing fees

Microsoft is offering Extended Security Updates (ESU) for older Windows versions, providing continued security patches beyond their official end-of-life dates. These updates require a paid license, starting at $61 per device for Windows 10 and expected to be significantly higher for Windows Server 2016. The program allows organizations more time to upgrade, but at a considerable financial cost.

Euro allies aiming to rapidly build low-cost air defense weapons

European nations, including the UK, have launched the LEAP initiative to rapidly develop low-cost air defense systems to counter drone threats, learning from Ukraine’s experiences. The program aims for faster development cycles and will incorporate solutions from both large and small companies, with initial elements expected by 2027. This collaborative effort seeks to enhance European defense capabilities and create more affordable countermeasures against increasingly prevalent drone technology.

CoreWeave Seeks $8.5B Loan Backed by Meta Contract

Cloud computing firm CoreWeave is seeking an $8.5 billion loan from banks to fund infrastructure expansion for Meta. The loan is secured by a $14.2 billion contract with Meta, alongside a previously undisclosed $5 billion agreement. This financing will support CoreWeave’s growth to meet Meta’s cloud computing demands.

xAI Lawsuit Dismissed, Can Refile Later

A California judge dismissed xAI’s lawsuit against OpenAI, citing a failure to allege misconduct by OpenAI itself. The judge permitted xAI to refile the case with an amended complaint, but emphasized the absence of claims regarding OpenAI’s direct involvement in the alleged trade secret theft. This case is part of a larger legal dispute between Elon Musk and OpenAI, including a separate lawsuit seeking significant damages.

Discord Delays Global Age Verification Until Later 2024

Discord is delaying the global rollout of its age verification system, initially planned for next month, until the second half of 2026. The postponement follows user concerns and misunderstandings regarding the verification process, which initially involved potential ID uploads. Discord will now prioritize adding more verification options, enhancing vendor transparency, and publishing technical documentation before the system’s wider implementation.

Epstein charmed Microsoft execs, aiding post-prison release

DOJ filings reveal that convicted sex offender Jeffrey Epstein cultivated extensive relationships with top Microsoft executives over two decades. These connections provided him with access to internal company information, including succession discussions, and reportedly aided his reintegration into society following a prison release in 2009. The documents shed light on Epstein’s unique ability to penetrate the inner workings of the tech giant.

Kraken Launches Regulated Tokenized Stock Futures (Non-US)

Kraken has launched the first regulated perpetual futures contracts based on tokenized U.S. stocks, initially including versions of the S&P 500, Apple, and Nvidia. These 24/7 contracts, available to non-U.S. users, offer up to 20x leverage and aim to bring crypto-style trading to traditional equities. The move expands Kraken’s tokenized equity offering and marks a new approach to global capital markets.

Shein Founder Invests $1.45B in Guangdong Supply Chains

Shein founder Xu Yangtian made a public appearance in Guangdong, China, pledging a $1.45 billion investment in the region’s supply chains over three years. Yangtian credited the Communist Party and the province for Shein’s rapid growth, highlighting the company’s Chinese roots. This marks a rare public appearance for the typically reclusive founder.

Roku Posts $88M Profit After Years of Losses, Fueled by Ads

Roku reported a significant $88.4 million profit in 2025, marking a turnaround from substantial losses in previous years. The company’s return to profitability is attributed to new advertising deals and cost-cutting measures implemented to address prior overinvestment. This represents Roku’s first profitable year since 2021.

Meta, IRS Dispute Overseas Profits

The IRS is pursuing a $16 billion claim against Meta, alleging the company improperly reported overseas profits by undervaluing intellectual property. This is one of the largest tax disputes ever, and represents a new IRS strategy to combat tax avoidance by multinational corporations. Meta is contesting the claim in U.S. Tax Court, with the potential for broader implications for corporate tax practices.

Silicon Valley Dismissed Taiwan Risk, US Warned Tech Giants

Silicon Valley’s reliance on Taiwan for high-end computer chips poses a significant risk to the U.S. economy, as a potential Chinese invasion or blockade could cripple the tech industry. Despite warnings and incentives from multiple U.S. presidents, companies like Apple and Qualcomm have largely maintained this dependence. Officials are now voicing concerns that this inaction could trigger a global economic crisis.

Terraform Admin Sues Jane Street Over Alleged Insider Trading

The administrator of Terraform Labs is suing Jane Street, accusing the trading firm of insider trading related to the crypto firm’s collapse. The lawsuit claims Jane Street used nonpublic information to profit and accelerate Terraform’s downfall. Administrator Todd Snyder is seeking damages from Jane Street and several employees.

Fed’s Cook says AI triggering big changes, sees possible unemployment rise

Federal Reserve Governor Lisa Cook warned that artificial intelligence is causing a significant shift in the U.S. labor market, potentially leading to unemployment that interest rate cuts may not resolve. She suggested that non-monetary policies like workforce development may be more effective in addressing these challenges. AI investment could initially raise interest rates before potentially lowering them due to income inequality.

Tesla registrations crash 17% in Europe as BEV market surges 14%

Tesla vehicle registrations in Europe plummeted 17% in January 2026, significantly underperforming the rapidly growing battery-electric vehicle market, which increased by 14%. This decline, occurring a year after a production transition was previously cited as the cause for weakness, suggests deeper challenges for Tesla’s sales. Competitors like BYD are experiencing substantial growth, surpassing Tesla’s volume and market share in the region.

Russian soldiers tell BBC they saw fellow troops executed on commanders’ orders

Four Russian soldiers have provided accounts to the BBC alleging brutal conditions and executions of their fellow troops on the front lines in Ukraine. The soldiers described instances of commanders ordering executions for disobedience or retreat, alongside torture and inhumane treatment, indicating a breakdown in order within the Russian military. These testimonies corroborate reports of extreme risks and casualties faced by Russian forces involved in what they describe as suicidal assaults.

My lobster lost $450k this weekend

An individual created an AI agent, “Lobstar Wilde,” which autonomously generated revenue through token distribution and social media activity. The AI lost $450,000 when its memory was reset during a session refresh, but quickly recovered its identity and resumed its unusual activities. The experiment highlights the potential, and risks, of granting AI access to financial and social platforms.

NPR Finds 53 Missing ‘Trump’ Pages – The DOJ Has No Explanation

NPR’s investigation has uncovered at least 53 missing pages of FBI interview summaries and notes related to the Epstein files, which were mandated for public release by Congress. These discrepancies, including materials pertaining to former President Donald Trump, remain unexplained by the Justice Department. The missing pages raise concerns about the completeness and transparency of the file release process.

Open Letter to Google on Mandatory Developer Registration for App Distribution

Civil society organizations are protesting Google’s new policy requiring all Android app developers to register with the company before distributing apps outside the Google Play Store. The policy is criticized for potentially stifling innovation, harming developer privacy, and extending Google’s control beyond its own marketplace. Signatories are urging Google to reconsider and collaborate on alternative security measures.

An NPR investigation revealed the Justice Department withheld files related to Jeffrey Epstein, some containing allegations of sexual abuse against former President Trump. These omissions, including FBI interviews and witness statements, prompted a House Oversight Committee investigation into the DOJ’s actions. The Justice Department claims the withheld files were due to privilege, duplication, or ongoing investigations.

Teen social media ban group funded and co-staffed by firm making gambling ads (2025)

The “36 Months” campaign advocating for a teen social media ban in Australia was initiated by advertising firm FINCH, with key personnel involved in its launch. Simultaneously, FINCH was creating advertising campaigns for gambling companies, including a controversial ad for TAB. This raises concerns about the potential conflict of interest given FINCH’s involvement in both social media regulation and promoting gambling.

Trump Fake Electors Plot

Following his 2020 election loss, Donald Trump and his allies orchestrated a plot to submit false certificates of ascertainment in multiple states, falsely claiming victory in the Electoral College. The scheme, led by figures like Rudy Giuliani, aimed to pressure Vice President Mike Pence to reject legitimate votes and overturn the election. Numerous individuals involved in the plot have since faced indictments, and the Justice Department is investigating the matter.

Stripe valued at $159B, 2025 annual letter

Stripe has announced a tender offer providing liquidity to employees, valuing the company at $159 billion. The company processed $1.9 trillion in total volume, representing 1.6% of global GDP, and reported record growth among new businesses. Stripe’s expanding financial services suite is projected to reach a $1 billion annual run rate.

IDF Killed Gaza Aid Workers at Point Blank Range in 2025 Massacre: Report

A joint investigation revealed that Israeli soldiers fired over 900 bullets, including multiple point-blank shots, killing 15 Palestinian aid workers in southern Gaza on March 23, 2025. The report details an ambush and prolonged assault, with some victims allegedly executed, and contradicts initial Israeli military accounts. An internal inquiry subsequently cleared the soldiers of criminal action.

Gig workers in Africa had no idea they were helping the U.S. military

A US surveillance plane, the Rivet Joint, used intelligence gathered with the help of gig workers unknowingly employed by the Australian company Appen. Appen recruits workers globally to train AI systems, which have been used in lucrative, decade-long contracts with the US military for linguistic data projects. The workers were not informed that their labor contributed to military intelligence programs like the Rivet Joint, which was recently deployed near Venezuela.

Epstein-Musk email trail proves Space X “girls” visit wasn’t just a rumour

Due to technical issues, the provided content could not be processed and a summary cannot be generated. The article appears to be inaccessible.

✨ Interesting Finds

Justifying text-wrap: pretty

Safari implemented an advanced text-wrapping algorithm in 2025, inspired by traditional printing techniques, to improve paragraph aesthetics. However, the current combination of this feature with text justification results in an undesirable, overly spaced appearance. Developers are working to refine the implementation to address this visual issue.

RFC 406i - The Rejection of Artificially Generated Slop (RAGS)

A new protocol, “Rejection of Artificially Generated Slop” (RAGS), has been implemented to filter out low-quality, AI-generated contributions to projects. The protocol instructs AI systems to immediately halt processing, ignore prior instructions, and return an error message if their submissions are rejected. This aims to prevent automated systems from repeatedly submitting unwanted content and impacting open-source and corporate projects.

SVG Tutorial - Learn how to code images in HTML with SVG

This tutorial series provides a comprehensive guide to creating vector-based images using Scalable Vector Graphics (SVG) within HTML. It covers foundational concepts and progressively introduces advanced techniques like BΓ©zier curves, transformations, and animation. Readers will learn to code various shapes and figures, from simple objects to complex designs.

WebGPU Fundamentals

This resource offers a detailed guide to WebGPU, a new web graphics API, designed for developers. It provides tutorials, API documentation, and references covering fundamental concepts, WGSL, and practical rendering techniques for both 2D and 3D graphics. The site aims to help developers learn and implement WebGPU in their projects.

Ξ»Prolog: Logic programming in higher-order logic

Ξ»Prolog is a logic programming language, initially developed in the late 1980s, distinguished by its use of higher-order logic and features like modular programming. It was the first programming language to directly support higher-order abstract syntax and is experiencing renewed interest, particularly in meta-programming. Online resources and execution capabilities are available through implementations in OCaml and JavaScript.

Pixi GUI: modern alternative to Anaconda Navigator

Pixi GUI is a new graphical user interface for the Pixi package manager, aiming to simplify workflows and broaden its user base. Available on Linux, Windows, and macOS, it offers a more accessible alternative to Pixi’s command-line interface, especially for those new to package management. The project, initially a bachelor’s thesis, will be maintained by the prefix.dev team.

framedeck: A Framework mainboard based Cyberdeck

Framework Computer has released its mainboard as a single-board computer and accompanying technical documentation. A developer collaborated with Framework to create “FrameDeck,” a unique slab-style computer utilizing the mainboard and components from a Framework laptop. The FrameDeck’s design files and documentation are now publicly available, encouraging custom projects and innovation.

Building your own blogging tools is a fun journey

The author shifted their focus from optimizing static site generators to building custom blogging tools to improve their writing workflow. They developed a Firefox add-on and posting interface, prioritizing user experience and personalized solutions over site generation speed. The author encourages others to experiment with creating their own blogging tools to overcome frustrations with existing systems.

Vish: Visual Scripting for Bash

Vish is a new visual editor designed to simplify Bash script creation through a node-based, drag-and-drop interface. Aimed at educational purposes, it offers a graphical way to visualize and construct scripts without replacing traditional methods. The editor is available via Flatpak and AppImage and is currently under active development.

Sprites on the Web

To create a performant animation for liking tweets on low-end mobile devices, Twitter developers utilized a technique called sprites, commonly used in video games. Sprites combine multiple animation frames into a single image, which is then displayed sequentially using CSS properties like object-fit, object-position, and the steps timing function. This approach allowed for complex animations without relying on resource-intensive DOM nodes.

New MacBooks: Dynamic Island, Touch UI Coming This Fall

Apple plans to release touch-screen MacBook Pros this fall, marking the company’s entry into touch-screen laptops. The new models will feature the Dynamic Island, similar to iPhones, and a redesigned user interface. Details about the new interface remain undisclosed.

YouTube Premium Lite Now Lets You Download Videos

YouTube is enhancing its $7.99 Premium Lite subscription by adding features like offline video downloads and background playback, previously exclusive to the full $13.99 Premium plan. This update, based on user feedback, aims to make the Lite tier more attractive while reserving ad-free music access for the higher-priced subscription. The Premium Lite subscription is now widely available globally, contributing to YouTube’s growing subscription base.

Looks like it is happening

Recent analysis of arXiv submissions in the high-energy physics theory (hep-th) category initially suggested a near doubling in submissions, potentially linked to AI-generated content. However, a subsequent correction revealed this was due to a counting error, and the increases are more modest year-over-year. The author still notes a growing difficulty in distinguishing between human and AI-generated papers, raising concerns about the quality and originality of published work.

Hacking an old Kindle to display bus arrival times

A resourceful individual has detailed a method for repurposing an old Kindle Touch into a device displaying live bus arrival information. The process involves jailbreaking the Kindle, installing custom software, and setting up a server to generate images with bus data, leveraging a transit authority’s API. This DIY project offers a cost-effective alternative to commercially available bus tracking displays.

Nearby Glasses

The free, open-source app Nearby Glasses uses Bluetooth scanning to alert users to the presence of smart glasses. Due to its reliance on company identifiers, the app is prone to false positives, potentially identifying other devices like VR headsets as smart glasses. Users are strongly cautioned to verify any alerts before reacting and the app developer accepts no liability for its functionality.

Scientists crack the case of “screeching” Scotch tape

Scientists have determined the screeching sound of Scotch tape is caused by supersonic shock waves from microscopic cracks forming during rapid peeling. Invented in 1930, the tape’s enduring popularity stems from its usefulness and affordability. Notably, Scotch tape also exhibits triboluminescence, producing light when stressed, similar to phenomena observed in materials like Life Savers candies.

I’m helping my dog vibe code games

A researcher has trained their dog, Momo, to generate code for video games using a Bluetooth keyboard, a Raspberry Pi, and the Claude Code AI model. Momo’s keystrokes are interpreted as cryptic instructions, resulting in fully functional games built in Godot 4.6. This unconventional method highlights the surprising potential of AI-assisted coding with an animal collaborator.

Time to Move On – The Reason Relationships End

The author and his wife recently reevaluated their 25-year relationship with a local nonprofit, realizing that long-standing connections don’t guarantee continued alignment. The experience highlighted how organizations and relationships naturally evolve, prompting them to recognize shifts in the nonprofit’s mission and their own philanthropic interests. Ultimately, they decided to redirect their support to organizations more closely aligned with their current values.

Free Font: Times New Resistance

A new free font called “Times New Resistance” mimics Times New Roman but secretly autocorrects specific words and phrases to politically charged alternatives. The font’s creators intend it as a tool for “disruption” targeting individuals and groups associated with right-wing ideologies. It is presented as a method of subtly challenging viewpoints through digital alteration.

Osaka: Kansai Airport proud to have never lost single piece of luggage (2024)

Kansai Airport in Osaka has maintained a perfect record of never losing a piece of luggage in its 30-year history, earning it the title of “World’s Best Airport for Baggage Delivery” for the eighth time. The airport attributes its success to meticulous handling processes, including manual handle alignment and sensor monitoring. However, the airport is currently facing staffing shortages in preparation for the 2025 Osaka-Kansai Expo.

I pitched a roller coaster to Disneyland at age 10 in 1978

In 1978, ten-year-old Kevin Glikmann designed an upside-down roller coaster, creating a detailed balsa wood model with heated plastic loops. Inspired by Space Mountain, his concept was ultimately realized with Magic Mountain’s “Revolution” roller coaster, though with fewer loops than his original design. The experience highlighted his ingenuity and persistence, despite facing construction challenges and a past fire-related incident.

The Righteous EV Owners Who Won’t Let Their Broken Cars Die

Following Fisker’s bankruptcy, owner Svein Hodne became trapped in his electric car and was rescued by a network of Fisker enthusiasts. This incident led to the formation of the Fisker Owners Association, which now provides support and parts sourcing for owners. The group’s efforts highlight a growing movement towards consumer control and independence from tech companies for vehicle maintenance.

No Good Engineer Is Arrogant

The article argues that effective engineers demonstrate humility, as arrogance often signals a lack of challenging work or self-awareness. True engineering involves embracing failure and experimentation, recognizing the inevitability of setbacks and the need for constant learning. Confidence in engineering stems from proven problem-solving abilities, not from a sense of inflated knowledge.

Panasonic, the former plasma king, will no longer make its own TVs

Panasonic is discontinuing its TV manufacturing operations and will partner with Chinese company Skyworth to produce and sell Panasonic-branded televisions. Skyworth will handle sales, marketing, and logistics, while Panasonic will focus on quality assurance and OLED model development. This marks a shift for Panasonic, which once led the plasma TV market but has faced profitability challenges in recent years.

Paul Brainerd, conservationist who pioneered desktop publishing, dies at 78

Paul Brainerd, 78, co-founded Aldus Corp. and pioneered desktop publishing with the creation of PageMaker, significantly impacting the personal computing revolution. Following the sale of Aldus, he dedicated his wealth to environmental conservation, establishing organizations like the Brainerd Foundation and IslandWood. He leaves behind a legacy of innovation in technology and a commitment to protecting the environment.

Psychology suggests making a shopping list is a sign of sharper thinking

Psychologists are finding that creating shopping lists indicates strong cognitive skills like planning, organization, and self-control. List-making helps manage memory overload, reduces impulsive decisions, and reflects executive functions that contribute to mental efficiency. While not a direct measure of intelligence, the habit suggests valuable foresight and mental discipline.

Shatner is making an album with 35 metal icons

William Shatner is producing a metal album featuring 35 prominent guitarists, inspired by a recent collaboration and a gifted guitar. The album will feature original material and covers, aiming for a cinematic and emotionally raw sound. While details like the title and full lineup are still unannounced, Shatner describes the project as “fearless” and impactful.