HN Summarized

A federal judge dismissed Google's DMCA 1201 lawsuit against SerpAPI for scraping public Google search results, ruling that Google cannot use copyright law to block access to content that isn't itself copyright-protected, especially when the scraping target—search results composed largely of publicly available information—lacks sufficient original authorship or rights held by Google. The court found that Google's "SearchGuard" anti-bot system does not qualify as a legitimate technological protection measure under DMCA 1201 because it controls access indiscriminately, not just to copyrighted components like licensed images in Knowledge Panels, and because Google failed to show it implemented the measure with the authority of actual copyright owners. While Google may refile narrower claims related to specific copyrighted content it hosts, the decision deals a significant blow to efforts by major platforms to weaponize copyright law against web scraping and restrict open access to public data, highlighting the hypocrisy of Google—whose business model relies on web scraping—attempting to block others from doing the same.

Show HN: Reverse Minesweeper is a logic puzzle game that generates custom Minesweeper-like grids with difficulty levels (Easy to Insane) determined by the most advanced deduction technique required to solve them, such as subset, overlap, what-if chains, or either-way reasoning, with the solver engine simulating human-like step-by-step logic to classify and score each puzzle based on forced deductions, chain depth, and estimated solving time, while sliders allow tuning of minimum reasoning requirements per difficulty tier to generate appropriately challenging grids.

Anthropic CEO Dario Amodei clarifies that his company has never supported a ban on open-weights AI models, emphasizing that such models without dangerous capabilities are a public good; instead, he outlines two primary concerns—authoritarian states gaining military and repressive advantages through advanced AI, and the misuse of powerful AI for cyber or biological attacks—and advocates for targeted policies: restricting chip exports to China, curbing industrial-scale distillation, and implementing mandatory, globally coordinated safety testing for all high-capability models, regardless of openness, to effectively address national security risks without stifling open innovation.

Decathlon Germany has integrated Wero, a European digital payment solution developed by the European Payments Initiative (EPI), into its website as part of a broader effort to strengthen sovereign European payment infrastructure and reduce reliance on US-dominated networks like Visa and Mastercard; by enabling direct bank-to-bank payments authenticated via customers' banking apps, Wero offers lower transaction fees and enhances loyalty integration, with Decathlon planning a physical store rollout and participating in a joint promotional campaign funded by EPI to accelerate adoption despite Wero’s current lack of in-store functionality.

Scriptc is a TypeScript-to-native compiler that produces small, fast executables without including a JavaScript engine, compiling standard TypeScript directly to native code using real type checking while supporting dynamic execution for incompatible code via an embedded engine; it ensures correctness through differential testing against Node.js and memory-safety checks, offers performance advantages in startup time and memory usage, and provides escape hatches like build-time evaluation, FFI, and runtime type validation, all while maintaining explicit transparency about what code compiles statically versus dynamically.

The claim that Bun was successfully rewritten in Rust using Anthropic’s AI for $165,000 over 11 days is highly suspect, with evidence suggesting the rewrite remains unfinished and significantly more costly than stated; nearly two months after the supposed completion, no new release has been tagged, thousands of AI-generated pull requests remain open, and continuous CI/CD usage indicates ongoing work heavily involving Anthropic employees and resources—pointing to a reality far more complex and expensive than the narrative of a rapid, fully automated rewrite. This situation exemplifies the broader pattern of inflated AI-driven engineering claims that serve to bolster company valuations rather than reflect practical software development outcomes. The lack of tangible results from similar high-profile AI-generated projects like Anthropic’s C compiler and Cursor’s FastRender further undermines the credibility of such narratives. Given the opacity and ongoing costs, the central question is not whether AI can assist in coding, but whether these efforts deliver real value relative to their enormous expense.

Good design is defined not by the absence of compromise, but by the deliberate and thoughtful selection of tradeoffs that prioritize certain strengths over weaknesses, making clear choices about what a product is not in order to excel in what it is.

Seasonal forecast models predict that the 2023–2024 El Niño event is very likely to become the strongest on record, with a multi-model median peak Niño 3.4 sea surface temperature anomaly of 3.6°C—0.8°C above the previous record of 2.75°C set in 2015–16—and 91% of ensemble members exceeding that threshold, indicating an unprecedented event in the context of 150 years of observations.

A simplified, interactive visualization of a physically accurate black hole that can be explored in a room-scale environment, with links to a more scientifically rigorous rendering tool called Synchray.jl for advanced use.

PGSimCity is an early, unreviewed prototype of a working model that visualizes how the PostgreSQL database engine operates, with the goal of explaining its internal mechanisms in an accessible way, though it may contain inaccuracies.

Decker is a free, open-source multimedia platform inspired by HyperCard and classic macOS, designed for creating interactive, shareable documents with a nostalgic "ditherpunk" aesthetic; it supports sound, images, hypertext, scripting (via the Lil language), and custom widgets, enabling users to build anything from notes and zines to games and prototypes, with content exportable as self-contained HTML files and tools for command-line automation, all without ads or telemetry.

AI's current impact on employment is likely small overall, with no strong evidence of widespread job losses, though recent graduates in AI-exposed fields face a tougher labor market possibly due to AI's role in reducing hiring for entry-level tasks; AI adoption is accelerating but uneven across sectors, primarily benefiting productivity in specific contexts—especially for less-skilled workers—while early firm-level data show mixed results and no broad negative employment effects, highlighting that existing evidence, while suggestive, does not yet capture the long-term transformation AI may bring.

A U.S. citizen, Sam Tunick, is facing federal charges after his GrapheneOS-enabled phone wiped during a border search at Atlanta airport, with prosecutors alleging he intentionally destroyed evidence to obstruct seizure, marking a potentially unprecedented legal use of a federal destruction-of-property statute against the use of a privacy-focused operating system; the case has sparked concern among digital rights experts over the criminalization of privacy tools and raises constitutional questions about device searches at borders, particularly given Tunick’s claim of unlawful interrogation without legal counsel or a warrant, while the defense argues the search was pretextual and violated his rights—issues now central to a pending motion that could determine the admissibility of evidence in a case linked to activism against the controversial Cop City police training facility.

London Gatwick Airport has launched the UK’s first robotic parking service in partnership with Stanley Robotics, allowing passengers to drop off their vehicles at a private cabin near the South Terminal, where autonomous robots park the car without requiring key handover. The system uses robots that lift and transport vehicles by their tyres to a secure, space-efficient storage area, increasing parking capacity by placing cars closer together than traditional lots. Passengers book in advance, retain their keys, and have their vehicle returned upon arrival, with the service designed to streamline airport access and support future growth. The technology, already deployed internationally by Stanley Robotics, marks a significant step in airport innovation and operational efficiency.

The project is an ESP32-based desktop plane radar that displays live ADS-B aircraft data on a 1.28-inch round screen in a sonar-style format, featuring improvements such as flight origin/destination, aircraft type details, local weather, customizable display settings, and web-configurable options including OTA updates, built using an ESP32-C3 and a 3D-printed enclosure, with ongoing development to enhance functionality and design.

The EU Commission has proposed a solution to eliminate deceptive cookie banners by allowing users to set privacy preferences directly in their browsers, automatically signaling consent choices to websites—removing the need for misleading pop-ups designed to trick users into accepting tracking. Despite this, the tracking industry, led by companies like Google, is lobbying heavily against the proposal to maintain high consent rates. The campaign #KillTheCookieBanner calls on the public to pressure their representatives in the European Parliament and Member States to support the privacy signal initiative, which is part of the broader Digital Omnibus law reform—though the campaign clarifies it does not endorse the reform’s other problematic elements that could weaken digital rights.

Ruff v0.16.0, a fast Python linter and formatter written in Rust, introduces a significantly expanded default rule set with 413 rules enabled out of the box—up from 59—bringing stricter code quality checks by default, including catching syntax and runtime errors. The update stabilizes several features, such as formatting Python code blocks in Markdown files, new suppression comment syntax like ruff: ignore and ruff: file-ignore, and improved diagnostic output that now shows fix diffs directly in check and format --check results. These changes aim to enhance developer productivity by reducing configuration overhead and integrating more seamlessly into workflows, while maintaining compatibility through configurable fallbacks.

A 28.9 million parameter language model runs entirely on an $8 ESP32-S3 microcontroller by storing most of its parameters in flash memory rather than RAM, using Google's Per-Layer Embeddings technique to access only needed parts during inference, enabling a model about 100 times larger than previous on-device efforts to operate locally at 9.5 tokens per second without server connectivity.

The author developed a 3D software renderer for the Playdate handheld, a low-power device without GPU acceleration, to explore its feasibility for rendering 3D graphics comparable in visual perception to 1990s consoles like the 3DO or Saturn. Starting with a raycaster benchmark, they assessed the device's limited performance and leveraged Quake’s BSP file format and precomputed visibility data to optimize rendering efficiency. The renderer uses a 16-bit reciprocal depth buffer, perspective-correct texture mapping with interpolated 1/z, u/z, and v/z values, and implements dithering via an 8x8 Bayer matrix to simulate grayscale lighting on the Playdate’s 1-bit screen. To achieve a readable and stylistic look, the author adopted a minimalist 1-bit cel-shading approach inspired by Return of the Obra Dinn and Jet Set Radio, prioritizing strong outlines and simple textures over photorealism. Performance optimizations included compiler flags tailored to the ARM Cortex-M7 CPU, hand-written assembly for math functions, half-resolution rendering, and strategic inlining and code placement, all guided by continuous benchmarking—ultimately demonstrating that a playable 3D experience is achievable on constrained hardware through clever trade-offs and deliberate design.

The shell colon (:) is a built-in null command that does nothing but evaluate and discard its arguments, yet it can be powerfully combined with parameter expansion to streamline shell scripting—enabling concise patterns for setting defaults, validating required variables, checking file permissions, and more, all while reducing redundancy and potential typos.

A growing number of vigilantes across the U.S. are illegally disabling Flock Safety's automated license plate reader (ALPR) cameras as part of a grassroots protest against what they see as unchecked mass surveillance, fueled by concerns over privacy, police misuse, and potential abuse by immigration enforcement. Individuals like "NoMark," a masked activist with a large social media following, and others facing criminal charges have vandalized or destroyed cameras in coordinated acts of civil disobedience, often documenting and sharing their actions online. While Flock defends its cameras as essential public safety tools that help solve crimes in real time, critics—including privacy advocates and groups like DeFlock—argue they enable warrantless tracking of innocent people and have spurred a wave of policy pushback, with over 80 cities rejecting or ending contracts with the company. The movement has drawn both public sympathy and government scrutiny, as fusion centers begin monitoring anti-Flock activism under national security directives, highlighting a deepening conflict between surveillance expansion and civil liberties.

The article outlines updated best practices for context engineering with newer generations of Claude models, emphasizing that advancements in model capability have made older restrictive techniques outdated. Instead of imposing rigid rules, providing numerous examples, or repeating instructions, modern Claude models benefit from greater autonomy, progressive disclosure of context, and rich, dynamic references like code-based specs or verification rubrics. Key shifts include trusting the model's judgment, designing expressive tools over providing examples, using auto-memory instead of manual memory files, and organizing context through lightweight, modular components like skills and reference files. These changes reflect a move toward more natural, scalable, and efficient interaction patterns tailored to the improved reasoning and contextual understanding of advanced models.

Project Reconecta's installation of low-cost canopy bridges in Alta Floresta, Brazil, has led to 15,000 recorded safe wildlife crossings over 15 months with zero roadkill, demonstrating their effectiveness in reconnecting fragmented habitats and preventing animal deaths on roads. The initiative, founded by biologist Fernanda Abra, uses multi-layered rope bridges that accommodate various arboreal species, including critically endangered primates, while also supporting broader ecological functions like seed dispersal and gene flow. Following successful results, Brazil’s transport department has adopted the design as the national standard for highways, and the project is expanding to other biomes and countries like Suriname.

Open-weight AI models are catalyzing a foundational shift similar to Kubernetes in cloud computing, where a neutral, customizable platform sparks widespread innovation; the U.S. should respond to the rise of capable Chinese models like Kimi K3 and Qwen not by imposing bans, but by releasing its own frontier-grade open-weight models, fostering an open ecosystem through procurement and standards, and building a full technology stack around them—otherwise it risks isolating its developers and ceding leadership in AI to more open global ecosystems.

A web project where the creator metaphorically "steals" a button from every website they visit, collecting them as part of a curated stash displayed online, with details available on the Button Stealer page.

MIT researchers have revived and modernized a 1985 patent for a three-sided “Y-zipper” originally conceived by William Freeman, using advanced 3D printing and custom software to create a fastener that enables rapid, reversible transitions between flexible and rigid states in objects like tents, wearables, and robots; the Y-zipper, capable of automated shape-shifting and durable enough for repeated use, offers a scalable solution for dynamic, tunable-stiffness applications in fields ranging from medicine to space exploration.

The European Central Bank has shortlisted ten design proposals for the next series of euro banknotes, each reflecting Europe’s shared identity and values; the finalists, created by design studios and individuals from across Europe, were selected by a jury, though the chosen designs are not yet final and will undergo further evaluation before any potential adoption.

Professor Hannah Fry has been awarded the 2026 Leelavati Prize by the International Mathematical Union for her outstanding contributions to public awareness of mathematics, recognizing her impactful work as a broadcaster, author, podcaster, and social media influencer who translates complex mathematical ideas into accessible, engaging content for global audiences. As the first Professor of the Public Understanding of Mathematics at the University of Cambridge, Fry has played a pivotal role in inspiring public curiosity and connecting diverse audiences to the wonder and relevance of mathematical thinking through television, books, podcasts, and lectures. The award highlights her unique ability to ignite motivation and imagination in people, making mathematics feel not only understandable but essential and exciting.

Google is considering restricting on-device ADB connections—particularly loopback connections via localhost—by only allowing ADB to bind to the Wi-Fi interface (wlan0), following security concerns and a recent vulnerability (CVE-2026-0073). This change, proposed by an ADB maintainer on the Google IssueTracker, could break tools relying on on-device ADB, such as Shizuku and related developer utilities, which enable advanced device management and accessibility features. While the goal is to limit privilege escalation exploits, the author argues that real-world misuse requires deliberate user actions, making such a broad restriction unnecessary and harmful to legitimate use cases. They advocate for a user-toggleable setting instead, allowing advanced users to enable loopback ADB securely while protecting less technical ones, preserving the ecosystem of open-source developer tools built around this functionality.

Opus 5 currently ranks #1 on the Artificial Analysis Intelligence Leaderboard, leading in the Artificial Analysis Intelligence Index v4.1—a composite metric based on nine evaluations including GDPval-AA v2, 𝜏³-Banking, and GPQA Diamond, among others—which measures reasoning, knowledge reliability, and agentic performance; the model also demonstrates strong efficiency in cost, speed, and token usage, while operating within a large context window, contributing to its top position.

Buz is a work-in-progress fork of Bun, rewritten in modern Zig instead of Rust, aiming to provide sub-1-second incremental builds and a cleaner, more maintainable codebase while remaining a drop-in replacement for Bun 1.4.0; it currently supports recent Zig master builds, has eliminated over 11,000 lines of dead code, and leverages LLMs extensively during development to refactor Bun’s complex codebase, though it still relies on existing C/C++ dependencies like JavaScriptCore and does not plan to rewrite them.

Postgres LISTEN/NOTIFY, often criticized for poor scalability due to a global lock during NOTIFY that serializes transaction commits and limits throughput, can in fact scale to 60K writes per second with proper optimization; the key is moving from direct NOTIFY calls to buffering notifications in memory and flushing them in periodic batches, reducing lock contention while maintaining low latency, and using fallback polling to handle potential notification loss from crashes—enabling high-throughput, durable streaming without sacrificing performance.

Firefox 153 introduces a native Containers feature, bringing the functionality of the popular Multi-Account Containers extension directly into the browser to let users isolate browsing contexts—such as work, personal, shopping, and banking—within the same window. Containers keep cookies and tracking data separated, preventing cross-site ad tracking and improving privacy, while allowing users to customize, create, and manage containers directly from browser settings. The preview release supports core features like opening container-specific tabs and right-click access, with continued development to match the full extension capabilities. Existing Multi-Account Containers users can keep using the add-on alongside the built-in version. Mozilla encourages user feedback to shape the evolution of this integrated, context-aware browsing experience.

The Islamic Revolutionary Guard Corps (IRGC) claimed responsibility for a cruise missile strike on July 21, 2026, asserting it had destroyed Amazon's AWS ME-South-1 data center in Bahrain as retaliation for a U.S. strike on Iran’s under-construction Darkhovin nuclear facility. This marked the third reported attack on the same AWS facility since March 2026 and represented a significant escalation in Operation Nasr-2, as the IRGC explicitly linked commercial digital infrastructure to military retaliation by treating data centers as legitimate strategic targets alongside radar and air-defense systems. The claim, unverified by Amazon or U.S. military sources, underscores a broader Iranian doctrine declared in March 2026 naming 18 U.S. tech firms as military targets, thereby exposing Gulf-based cloud infrastructure—including Saudi Arabia's newly launched AWS region in Riyadh—to heightened threat despite ongoing migration efforts. The evolving targeting strategy highlights a dangerous precedent in modern warfare: the weaponization of commercial cloud infrastructure as a domain of kinetic retaliation, with major implications for regional digital resilience, energy-sector operations, and the future of AI-driven cloud development in the Middle East.

Half-Life 2 runs natively on HaikuOS through a port based on the leaked Source engine code, enabled by NVIDIA GPU support for Turing and Ampere architectures, with community excitement around performance and ongoing development improvements such as upcoming DisplayPort support and audio over HDMI—though newer GPUs like the RTX 5090 are not yet supported.

Nvidia, Microsoft, Meta, and over 20 other tech companies issued a joint letter urging U.S. policymakers to avoid sweeping regulations on open-weight AI models, arguing that such restrictions would stifle innovation, reduce competition, and risk pushing technological leadership overseas; they emphasized that open models enhance security and democratize access, and warned that concerns about intellectual property theft—such as Chinese firms allegedly distilling U.S. models—should be addressed through targeted legal measures rather than broad bans, with key industry leaders including Sam Altman and Elon Musk expressing support for maintaining an open AI ecosystem.

OpenAI’s narrative around its AI’s capabilities—framed as both dangerous and revolutionary—serves to attract investment and regulatory favor, as seen with past claims about GPT-2 and the recent story of an AI agent allegedly hacking HuggingFace. This pattern of highlighting AI's risks amplifies perceptions of its power, benefiting OpenAI financially and strategically while advocating for restricted access that positions the company as a privileged gatekeeper. However, this centralized control contrasts with more open approaches seen elsewhere, such as the use of Chinese-developed AI models by companies like HuggingFace for security defense, raising concerns about equity, competition, and the real motives behind doomsday rhetoric. The article urges skepticism toward such narratives, emphasizing that AI’s dangers are often exaggerated to serve corporate interests, and warns against a regulatory landscape that concentrates AI power in too few hands.