How a blind man made it possible for others with low vision to build Lego sets (apnews.com)
An AP profile describes how a blind Lego fan used audio and braille instructions to help people with low vision build Lego sets. Inspired by his own experience learning the sets, he worked to make the instructions easier to follow without sight-based steps. The story highlights how adaptive formats can broaden access to popular toys.
What we learned about TEE security from auditing WhatsApp's Private Inference (blog.trailofbits.com) AI
Trail of Bits reports findings from an audit of Meta’s WhatsApp “Private Inference,” which uses TEEs to run AI message summarization without exposing plaintext to Meta. The review found 28 issues, including high-severity problems that could undermine the privacy model, and describes fixes focused on correctly measuring and validating inputs, verifying firmware patch levels, and ensuring attestations can’t be replayed. The authors argue TEEs can support privacy-preserving AI features, but security depends on many deployment details—such as input validation, attestation freshness, and negative testing—not just the underlying TEE isolation.
S3 Files (aws.amazon.com)
AWS announced S3 Files, which exposes Amazon S3 buckets through a NFS v4.1+ file system interface. The service automatically syncs file changes back to the underlying S3 objects and lets multiple compute resources access the same data without duplication, using features like ~1 ms latency for active data via EFS-based architecture. AWS says access control, encryption, and monitoring are handled through existing IAM, TLS/S3 encryption options, CloudWatch, and CloudTrail, and the blog includes steps to mount an S3 bucket on EC2.
S3 Files and the changing face of S3 (allthingsdistributed.com)
The article traces how early experience with moving large genomics datasets highlighted “data friction” between S3-style object storage and tools that expect POSIX-like filesystems. It then explains how the S3 team has responded by adding higher-level storage primitives—S3 Tables for structured data using managed Iceberg, and S3 Vectors for similarity search—before announcing S3 Files, which integrates Amazon EFS so existing S3 data can be accessed directly as a network-attached filesystem.
Linux kernel maintainers are following through on removing Intel 486 support (arstechnica.com)
Linux kernel maintainers are progressing toward dropping built-in support for Intel’s 486 CPUs, with commits indicating Linux 7.1 will be the first release that can’t be built with 486 support. The change is intended to remove aging x86 emulation/compatibility code that benefits very few users and can cause maintenance problems. Users who still rely on 486-era systems will need to stay on older kernels or use vintage/vendor-supported operating environments, while most modern Linux distributions won’t be affected.
Show HN: Gemma 4 Multimodal Fine-Tuner for Apple Silicon (github.com) AI
The GitHub project “gemma-tuner-multimodal” describes a PyTorch/LoRA fine-tuning toolkit for Gemma 4 and Gemma 3n that targets multimodal data (text, images, and audio) on Apple Silicon using MPS/Metal, without requiring NVIDIA GPUs. It supports local CSV-based training (with streaming from cloud stores mentioned as an option) and exports fine-tuned adapters for use with HF/SafeTensors and related inference tooling. The repo also includes a CLI “wizard” for configuring datasets and launching training, plus installation guidance including a separate dependency path for Gemma 4.
College instructor turns to typewriters to curb AI-written work (apnews.com) AI
An instructor is requiring students to write assignments on typewriters to discourage the use of AI tools that can generate text, arguing that the method reduces AI-written cheating.
Why IPv6 is the only way forward (ankshilp.in)
The post argues that IPv4 address shortages have forced ISPs to rely heavily on NAT, creating long-term limitations—especially for countries like India where IPv4 allocation is low. It presents a comparison of IPv4 address allocation versus population and claims IPv6 adoption is both feasible and already widely supported by devices. The author urges a faster, more decisive transition to IPv6, suggesting that ongoing debate about IPv6 is outdated.
US Labor Force Participation Continues to Slide (restaurant.org)
The National Restaurant Association reports that the U.S. labor force participation rate has continued to decline, with the civilian labor force shrinking and participation falling to the lowest level since late 2021. The declines are broad-based but steepest among younger people, while prime-age participation has stayed relatively steady. The article links the trend to ongoing staffing challenges for restaurants, which rely heavily on entry-level and early-career workers.
Boneyard: Generate pixel-perfect skeleton screens from your real DOM (github.com)
Boneyard is an open-source tool that automatically generates pixel-perfect skeleton loading screens by capturing the real layout of your app’s UI (from the DOM or React Native’s view tree). It provides framework-specific components for React, Svelte 5, and React Native, plus a CLI that snapshots skeleton layouts across configurable breakpoints and outputs a shared “.bones.json” format. The project aims to remove manual measurement and hand-tuned placeholders while keeping production overhead low.
ICE arrested more than 800 people after tips from US airport security agency (reuters.com)
U.S. Immigration and Customs Enforcement arrested more than 800 people following tips from a U.S. airport security agency, according to a Reuters report, as authorities coordinated enforcement actions based on the provided information.
LED bulbs can damage paintings (vrt.be)
A doctoral thesis reports that some museum lighting—particularly LED bulbs emitting high levels of blue light—can alter certain chrome-yellow pigments used by artists such as Vincent van Gogh and Paul Cézanne. Using X-ray analysis, researchers found that unstable forms of chrome-yellow in works like “Sunflowers” and “Portrait of Gauguin” can change color within days under such light. The study advises museums to reconsider using LEDs over paintings to avoid damaging color changes.
The Image Boards of Hayao Miyazaki (animationobsessive.substack.com)
The article explains how Hayao Miyazaki uses “image boards”—loose concept sketches rather than storyboards—to explore visual ideas and set the tone for projects. Tracing his practice from early Toei days through major works like Totoro, it argues that many iconic scenes and motifs emerged from these drawings, which he repeatedly revisited, refined, and recombined over decades.
Apple Is Reportedly Facing a 'Massive Dilemma' with the MacBook Neo (macrumors.com)
Apple may face supply constraints for the MacBook Neo as strong sales could exhaust stocks of “binned” A18 Pro chips with a 5-core GPU before a next-generation model is ready. Because TSMC’s N3E production is reportedly at maximum capacity, Apple might need to pay more to restart or reallocate chip production, potentially squeezing profit margins. The report also suggests Apple could limit specific lower-storage configurations or shift timing to newer A19 Pro chips, though those options may be costly.
Show HN: The King James Bible deserved a better website (officialkingjamesbible.com)
The “Show HN” post links to OfficialKingJamesBible.com, a web app for reading the 1611 King James Bible with verse-by-verse commentary from multiple scholars. The site emphasizes customizable typography and themes, on-device notes and highlights, and tools like cross references, reading plans, and a “verse of the day.”
40 Years Later: Is Milk Still Radioactive? (birdsoutsidemywindow.org)
A blog post marks 40 years since the Chernobyl disaster and reviews evidence from a 2018 study showing that cows far from the plant were still producing radioactive milk in the 2000s/2010s, driven by long-lived cesium-137 contamination in soil. It notes that while air levels had improved, plants and then cattle could still take up cesium, and that mitigating the problem would require adding an expensive binding agent to feed. The author suggests the issue likely persists without treatment.
Wait Out AI's Super-Spending False Start (bloomberg.com) AI
Bloomberg discusses how AI-related spending has seen an early, uneven push, and suggests investors may need to wait for more durable signals of where money is actually working.
Your boss wants to use surveillance data to cut your wages (pluralistic.net)
The article argues that so-called “personalized pricing” is really “surveillance pricing,” where companies use data about individuals to set different prices—potentially including wages—based on predicted willingness to accept less pay. It describes “algorithmic wage discrimination” as a growing practice in sectors like ride-hailing, nursing staffing, and retail, citing how external data brokers and surveillance tools can depress wages by targeting people’s financial precarity. The author also claims weak privacy and competition enforcement has enabled the practice, and points to emerging state efforts to restrict it, such as disclosure and proposed bans.
Donald Trump is threatening the extinction of an 'entire civilization' tonight (isdonaldtrumpalive.com)
The article alleges that Donald Trump threatened to destroy an “entire civilization” over Iran’s response regarding the Strait of Hormuz, framing the remarks as potentially escalating a conflict. It also argues that any claim of regime change is unproven and urges diplomatic solutions rather than threats. The piece further calls for removal of Trump from power if he were to pursue extreme military actions, though it provides no independent verification of the claims.
The Blueprint of a North Korean Attack on Open-Source (casco.com)
The article describes how North Korean attackers used a supply-chain attack against JavaScript open-source projects by smuggling obfuscated malware into a seemingly legitimate pull request that modified build configuration files. The malicious code runs during npm build/dev/CI, decodes multi-stage payloads, pulls additional stages from blockchain “dead drops,” and then establishes command-and-control to execute arbitrary commands via a “zombie” process. It also discusses how GitHub’s default PR views can hide dangerous changes after merging, and provides a technical breakdown of the malware’s obfuscation and execution flow.
Testing suggests Google's AI Overviews tells lies per hour (arstechnica.com) AI
A test analysis (via Oumi) that benchmarks Google’s AI Overviews against thousands of fact-checkable questions found it answers correctly about 90% of the time, implying large numbers of incorrect summaries across all searches. Examples cited include confident factual errors about dates and institutions. Google disputes the benchmark’s relevance, saying the test includes problematic questions and that it uses different models per query to improve accuracy.
CIA used "long-range quantum magnetometry" called "Ghost Murmur" in Iran (nypost.com)
The article says the CIA used a previously untested system called “Ghost Murmur,” described as long-range quantum magnetometry that can detect a human heartbeat’s electromagnetic signature and use AI to distinguish it from background noise. It reports the tool was deployed during a two-day search for a downed U.S. airman in southern Iran and helped confirm his location in a remote mountain crevice. The piece adds that the technology was developed by Lockheed Martin’s Skunk Works and has been tested on helicopters, while noting its broader potential uses are unclear.
Hemingway's 4 Fast Rules for Effective Writing (wordsthatsing.com.au)
The article outlines Ernest Hemingway’s four “fast rules” for effective writing: use short sentences, short first paragraphs, vigorous (purposeful and precise) language, and be positive rather than negative. It argues these principles create clearer, more engaging prose—especially for readers who skim online—and suggests they can be applied through editing even if they aren’t achieved on the first draft.
Assessing Claude Mythos Preview's cybersecurity capabilities (red.anthropic.com) AI
Anthropic says its Claude Mythos Preview model shows “next-generation” strength in cybersecurity research, including finding and exploiting zero-day vulnerabilities across major operating systems and browsers. In testing under Project Glasswing, the company reports Mythos Preview can construct complex exploits (including sandbox-escaping and privilege-escalation chains) and turn known or newly discovered vulnerabilities into working attacks. The post details their evaluation approach and notes that most reported findings remain unpatched, so they provide limited disclosure while urging coordinated defensive action from the industry.
A truck driver spent 20 years making a scale model of every building in NYC (smithsonianmag.com)
Truck driver Joe Macken spent more than 20 years building a 1,350-square-foot, 1:2,400-scale model of New York City’s buildings, working from his home and adding hundreds of sections representing roughly one square mile each. The finished model—made from materials like balsa wood, paint, and glue—now has a museum exhibition at the Museum of the City of New York, where visitors can use binoculars to inspect specific neighborhoods. The project began in 2004 with a replica of 30 Rockefeller Plaza and later expanded to include parts of New Jersey and Nassau County.