Hasty Briefsbeta

Bilingual

White House staff told not to place bets on prediction markets
a day ago
- White House staff warned against using insider information for bets on prediction markets.
- Incident involved a bet on Venezuelan president's capture, raising concerns about inside information use.
- Prediction markets face regulatory scrutiny with calls for bans on war-related betting.
FBI used iPhone notification data to retrieve deleted Signal messages
a day ago
- The FBI recovered deleted Signal messages from an iPhone by accessing the device's notification database, even after Signal was uninstalled.
- The defendant, Lynette Sharp, had not enabled Signal's setting to hide message content in notifications, allowing message previews to be stored in Apple's internal notification storage.
- Notifications may continue to be sent after an app is deleted, as servers can't detect app removal immediately, and Apple recently updated how iOS handles push notification tokens.
- Data recovery likely occurred through device backups or using law enforcement tools that exploit iOS vulnerabilities, with the device's security state (like BFU/AFU modes) affecting access.
- Neither Signal nor Apple provided comments on notification handling or storage, leaving technical details of the FBI's method unclear.
Intel 486 CPU announced April 10, 1989
a day ago
- Intel introduced the 486 CPU on April 10, 1989, at a high price of $950 each in bulk quantities.
- Initial media reactions were mixed, with predictions about speed increases and the necessity of the chip for desktops, though these were later proven inaccurate as CPU speeds grew significantly.
- The 486 integrated key components like the 386, math coprocessor, cache controller, and 8K cache, offering performance two to three times faster than the 386 at the same clock rate.
- Early systems were very expensive, ranging from $10,000 to $15,000, with production starting in June 1989 and the first PCs shipping by September.
- Demand for the 486 increased with Windows 3.0 and 3.1, and applications like Doom and later versions of Microsoft Word drove the need for faster processors.
- Alternatives like the Cyrix Cx486SLC and IBM SLC2 used 386 sockets and buses with less cache, offering performance around 50-80% of a real 486, often bottlenecked by the 16-bit interface.
- Upgrade cards, such as the I-O Data PK-X486S50-3, allowed 286 systems to be upgraded to 486-like performance, though compatibility and driver support were considerations.
France Launches Government Linux Desktop Plan as Windows Exit Begins
a day ago
- France's DINUM announces a formal plan to transition government workstations from Windows to Linux as part of its national digital-sovereignty strategy.
- Each ministry and public operator must develop a comprehensive plan by autumn 2026 covering desktop systems, collaboration tools, AI, databases, and more.
- The move aims to reduce reliance on non-European digital technologies, emphasizing sovereignty and interoperability, with decisions on specific Linux distributions pending.
- Public reactions include support for reducing dependence on US tech giants and concerns about transition challenges like user adoption and technical compatibility.
- Some commenters note existing Linux use in French institutions, like the Gendarmerie using Ubuntu, and suggest broader European cooperation for a public-sector OS.
Python Is Dead
a day ago
- AI agents simplify development in strict languages like Rust by offloading compilation and correctness checks, unlike Python which requires constant human oversight.
- Python's design trade-offs (optional typing, interpreted execution, duck typing, GIL) favor human convenience but lead to runtime errors, performance costs, and hidden bugs.
- The rise of AI agents changes the labor equation: strict compile-time guarantees become cost-effective as AI handles complexity without frustration, making Rust-like languages more suitable.
- Future AI-first languages may optimize for token density, formal verifiability, declared effects, parallel-native execution, and structural editing, prioritizing machine efficiency over human readability.
- Programming language choice is effectively a labor policy; with AI, policies that emphasize upfront correctness (e.g., Rust) outweigh those favoring rapid development with high defect rates (e.g., Python).
Sorting Performance Rabbit Hole
5 days ago
- Pystd's stable sort was optimized to be 5% faster than stdlibc++.
- Unstable sort required extensive tweaks, with the biggest improvement coming from changing the insertion sort limit from 16 to 32 elements.
- Despite various attempts like using memmove and different pivot selection, only adjusting the insertion sort threshold significantly boosted performance.
- Pystd's unstable sort finally matched stdlibc++'s speed with a best time of 0.754 seconds versus 0.755 seconds.
France Launches Government Linux Desktop Plan as Windows Exit Begins
a day ago
- The French government is accelerating efforts to reduce its digital dependencies on non-European providers.
- A seminar organized by DINUM, DGE, ANSSI, and DAE set a clear objective to reduce these dependencies, involving ministers, administrations, public operators, and private actors.
- Concrete steps include DINUM switching from Windows to Linux, and the National Health Insurance migrating 80,000 employees to sovereign tools (Tchap, Visio, FranceTransfert).
- A new method involves forming coalitions between ministries, public operators, and private actors to collectively exit dependencies, leveraging digital commons and interoperability standards.
- DINUM will coordinate an inter-ministerial plan, with each ministry required to formalize its own plan by autumn, covering areas like workstations, collaborative tools, antivirus, AI, databases, virtualization, and network equipment.
- Mapping and diagnosis of dependencies by DAE, and defining a European digital service by DGE, will refine reduction targets with clear timelines.
- The first 'digital industrial meetings' in June 2026 will concretize public-private coalitions, including a 'public-private alliance for European sovereignty'.
Model-Based Testing for Dungeons & Dragons
4 days ago
- The article details the expansion of a formal model for Dungeons & Dragons from a single creature to combat interactions, focusing on complex mechanics like Counterspell chains, Shield spells, and Legendary Resistance.
- It explains the interrupt chain during attacks, which includes multiple phases (hit confirmation, damage application, spellcasting, saving throws) that can alter outcomes and branch the game state.
- The spec models deterministic mechanics while leaving non-deterministic aspects (like dice rolls, cover determination, opportunity attack triggers) as caller-provided inputs to accommodate different play styles.
- Model-based testing (MBT) has revealed bugs in both the spec and the implementation, such as argument swaps, invariant violations, and state sync issues, which unit tests often missed.
- The project uses a QA pipeline that scrapes community Q&A entries, translating them into assertions to test the spec against real-world rule interpretations and edge cases.
- The spec includes 62 safety invariants that are proven across all reachable states, not just tested with random inputs, ensuring properties like HP limits hold universally.
- Built on the SRD 5.2.1 (released under CC-BY-4.0), the architecture allows non-SRD content to be added without modifying the core spec, and the model can validate implementations in any language.
- The approach is applicable to other domains with formal rules and combinatorial interactions, such as financial regulations or protocol implementations.
Artemis II and the invisible hazard on the way to the Moon
a day ago
- Artemis II's primary scientific focus is on radiation measurements to enable safe human exploration beyond Earth's magnetic field.
- Radiation hazards include trapped particles in Van Allen belts, solar particle events, and chronic galactic cosmic rays.
- Radiation assessment considers factors like dose rate, particle type, shielding, and spacecraft orientation.
- Artemis II uses multiple tools: cabin monitors, crew-worn dosimeters, upgraded detectors, and biological experiments.
- The mission aims to map radiation geometry with crew onboard, helping decide sheltering strategies during solar events.
- AVATAR experiment uses organ chips from astronauts' bone marrow cells to study radiation's impact on immune function.
- NASA collects immune biomarkers via saliva and blood samples, tracking physiological responses like sleep and cognition.
- Radiation interacts with other mission stressors like microgravity, confinement, and workload, complicating biological effects.
Zero-build privacy policies with Astro
a day ago
- OpenPolicy's Astro integration previously generated Markdown files at build time via a plugin, adding friction with extra packages and generated directories.
- The new approach eliminates the plugin and generated files by directly compiling policies in Astro's frontmatter using three core packages: @openpolicy/sdk, @openpolicy/core, and @openpolicy/renderers.
- Users define a centralized config in src/lib/openpolicy.ts with details for privacy, terms, and cookie policies, which can be filled quickly using AI tools like Claude.
- Each dedicated page (e.g., privacy.astro) imports the config, compiles the policy, and renders it to HTML at build time with zero JavaScript shipped to the browser.
- Benefits include no generated files or directory management, no plugin configuration in astro.config.mjs, support for all three policy types from one config, and simpler inline compilation.
- OpenPolicy+ offers advanced features like cloud-based consent tracking, PR automation for compliance, and hands-on onboarding for further needs.
War on Raze
5 days ago
- The k7 prime number filter uses 'except' (^) with rank-sensitivity, eliminating the need for raze in later versions.
- Earlier versions required raze (,/) to flatten nested results from each-right multiplication, which was seen as a workaround.
- Deep where, inspired by Dyalog APL, offers a way to generate pairs without raze, using an exploded dict representation of data.
- An 'explode' verb transforms nested data into a flat key-value structure, enabling shallow verbs to mimic deep ones.
- The implementation allows for razeless, index-free prime generation, though it may be less efficient than the original.
- Deep verbs like deep find and deep group can be defined using the exploded form, potentially unifying shallow and deep operations.
Not just spyware scandals: EU is funding the industry that spies on Europeans
2 days ago
- Greece sentenced four individuals to prison in February 2026 over the 2022 Predatorgate spyware scandal, marking a rare criminal accountability case for spyware manufacturer executives.
- EU institutions, including the European Commission and member states, have been passive in addressing spyware proliferation, despite its human rights violations.
- Investigations reveal EU funds, such as subsidies from the European Defence Fund and Horizon programmes, are actively funding spyware companies like Intellexa and Paragon Solutions.
- The European Investment Fund, part of the European Investment Bank, funded Israeli spyware company Paragon Solutions, implicated in surveillance scandals in Italy.
- EU bodies lack transparency and accountability in managing public investments, making it difficult to track how taxpayer money supports harmful spyware firms.
- The article calls for a ban on EU funding for spyware, stricter investment policies, and a full EU-wide ban on spyware technology to protect democracy and fundamental rights.
Responsive images in Hugo using Render Hooks
2 days ago
- The article discusses the evolution from using Hugo shortcodes for responsive images (JPEG and WebP) in 2022 to using Render Hooks by 2026, as WebP is now widely supported in browsers.
- A Render Hook is implemented in a file at layouts/_default/_markup/render-image.html, which generates resized WebP and JPEG images, using srcset for WebP and a JPEG fallback, without the need for a <picture> element.
- The Render Hook automatically processes images referenced in Markdown files, supports page bundles (placing images next to Markdown files), and allows quality settings for WebP images via Hugo configuration.
Bun v1.3.12
a day ago
- Bun introduces Bun.WebView, a headless browser automation tool supporting WebKit and Chrome backends, with native OS events ensuring human-like interactions.
- Markdown files can now be rendered in the terminal directly using `bun ./file.md` or programmatically with `Bun.markdown.ansi()`.
- Async stack traces are now available for native APIs, improving debugging capabilities.
- In-process `Bun.cron()` scheduler allows recurring tasks within the application, with UTC-based scheduling and no overlapping executions.
- Improvements to `Bun.udpSocket()` include ICMP error handling without socket closure and detection of truncated datagrams via a flags argument.
- Unix Domain Socket behavior now matches Node.js, with correct EADDRINUSE errors on existing sockets and automatic cleanup after closing.
- JavaScriptCore engine upgraded with over 1,650 commits, adding support for Explicit Resource Management (`using` and `await using`) and various performance enhancements.
- Standalone executables on Linux now use an ELF section for embedded data, eliminating read permissions issues.
- `URLPattern` is up to 2.3x faster, thanks to reduced GC allocations and no pollution of `RegExp.lastMatch`.
- Optimizations to `Bun.stripANSI` and `Bun.stringWidth` using SIMD, resulting in significant performance improvements.
- Fixed a thread-pool bug in `bun build`, speeding up bundling on low-core machines by up to 1.47×.
- `Bun.Glob.scan()` performance improved, with up to 2× faster scans for patterns with boundaries and kernel-level filtering on Windows.
- Cgroup-aware `availableParallelism` and `hardwareConcurrency` on Linux respect CPU limits in Docker and Kubernetes.
- Keep-Alive for HTTPS Proxy CONNECT Tunnels reuses tunnels across requests, reducing latency and overhead.
- `Bun.serve()` on Linux uses `TCP_DEFER_ACCEPT` to reduce connection latency by minimizing event loop wake-ups.
- Multiple bugfixes across Node.js compatibility, Bun APIs, Web APIs, JavaScript bundler, `bun test`, Bun Shell, and Windows support.
First man convicted under Take It Down Act kept making AI nudes after arrest
a day ago
- An Ohio man, James Strahler II, pleaded guilty and became the first person convicted under the Take It Down Act.
- He created and shared real and AI-generated explicit images without consent, targeting at least 10 victims, including women and children.
- AI tools were used to create fake sexualized and incestuous images to harass, coerce, and threaten victims and their families.
- His actions included posting over 700 real and 'animated' images to a child sexual abuse website and other platforms.
- He faces potential prison sentences of up to two years for adult images and up to three years for those involving minors.
Installing OpenBSD on the Pomera DM250{,XY?}
a day ago
- Notes for installing OpenBSD-current on Pomera DM250 models (JP versions, not DM250US).
- Requires custom kernel and U-Boot due to uncommitted upstream work, with risks like battery drain issues.
- Backup eMMC before proceeding, using tools like EKESETE.net.
- Factory U-Boot boots recovery with Right Shift + Left Alt; new U-Boot replaces it, requiring EFI boot for recovery.
- Device may not power off with USB-C connected; battery must not fully drain to avoid boot failures.
- Create installation SD card with GPT, EFI partition, and OpenBSD files including custom U-Boot and script.
- Boot into recovery mode to run script that backs up and flashes new U-Boot.
- Install OpenBSD manually with custom fdisk steps for GPT partitioning on eMMC.
- Post-install steps: copy firmware, disable reorder_kernel, and set up LEDs.
- Recovery options include USB loader mode, U-Boot console commands, or hardware UART access.
VFX HQ: Visual Effects Headquarters (2000)
4 days ago
- Titanic: Collaboration of hundreds of top effects artists for James Cameron's film, with a 9-page FX review and high-resolution stills available.
- Starship Troopers: Interviews with effects artists like Phil Tippett and Scott E. Anderson on Paul Verhoeven's sci-fi thriller.
- The Star Wars Special Editions: Re-release of original trilogy with new visual effects, including debate on pros and cons.
- Flubber: Insights from ILM and Dream Quest supervisors on creating green slime for Disney's remake.
Show HN: Logoshi, a brand kit generator for solo founders
2 days ago
- The service allows you to create a brand kit by entering a few details.
- Brand kit includes: Wordmark in SVG and PNG formats, a Color Palette, Font Pairing, Business Card Template, and a Brand Guide.
- Example brand kit shown is named 'Ember & Oak'.
- Preview of the brand kit is available in minutes.
There are zero-day exploits for your mind
2 days ago
- Anthropic's Claude Mythos discovered thousands of zero-day vulnerabilities in major software systems, revealing hidden security flaws after decades of existence.
- The AI model found exploits with a 72.4% success rate, drastically outperforming previous models and human researchers, indicating a paradigm shift in cybersecurity.
- The article draws a parallel to AlphaGo's 'Move 37' in the game of Go, suggesting that both code and human psychology may have unexplored vulnerabilities.
- Defensive efforts, like patching bugs, are inherently slower and more difficult than exploit creation, highlighting an asymmetric threat landscape.
- AI's exponential progress means that current capabilities are the 'worst' they will ever be, with future models likely to uncover even more vulnerabilities.
- Exploits in legacy systems, such as the 16-year-old bug in FFmpeg, demonstrate challenges in patching widely embedded code and ensuring updates reach all users.
- Human manipulation is entering a new phase where AI could use personal data to discover psychological 'zero-day exploits,' potentially hacking minds in ways we cannot anticipate.
- The trust models of both computing and human communication are based on outdated assumptions about the difficulty of exploitation, necessitating a new paradigm.
- Preparation, such as securing personal data and acknowledging the inevitability of change, is emphasized over panic in facing these evolving threats.
Anthropic Detects Third-Party Clients via System Prompt, Not Headers
2 days ago
- Anthropic blocks third-party clients like OpenCode by analyzing the system prompt content, not through HTTP headers or TLS fingerprinting.
- Replacing the static portion of a custom system prompt with Claude Code's real prompt allows requests to succeed, while keeping custom prompts triggers a block.
- Detection is per-request and based on pattern matching or embedding similarity of the static prompt section, ignoring runtime-injected content like environment variables or custom instructions.
- A fix involves loading Claude Code's system prompt and swapping only the static part, preserving runtime sections, though it requires updates if Claude Code's prompt changes.
- This method is fragile for both Anthropic and users, highlighting potential OAuth architecture limitations in distinguishing official from unofficial clients.

first prev6next

About|Login

Bilingual

White House staff told not to place bets on prediction markets
a day ago
- White House staff warned against using insider information for bets on prediction markets.
- Incident involved a bet on Venezuelan president's capture, raising concerns about inside information use.
- Prediction markets face regulatory scrutiny with calls for bans on war-related betting.
FBI used iPhone notification data to retrieve deleted Signal messages
a day ago
- The FBI recovered deleted Signal messages from an iPhone by accessing the device's notification database, even after Signal was uninstalled.
- The defendant, Lynette Sharp, had not enabled Signal's setting to hide message content in notifications, allowing message previews to be stored in Apple's internal notification storage.
- Notifications may continue to be sent after an app is deleted, as servers can't detect app removal immediately, and Apple recently updated how iOS handles push notification tokens.
- Data recovery likely occurred through device backups or using law enforcement tools that exploit iOS vulnerabilities, with the device's security state (like BFU/AFU modes) affecting access.
- Neither Signal nor Apple provided comments on notification handling or storage, leaving technical details of the FBI's method unclear.
Intel 486 CPU announced April 10, 1989
a day ago
- Intel introduced the 486 CPU on April 10, 1989, at a high price of $950 each in bulk quantities.
- Initial media reactions were mixed, with predictions about speed increases and the necessity of the chip for desktops, though these were later proven inaccurate as CPU speeds grew significantly.
- The 486 integrated key components like the 386, math coprocessor, cache controller, and 8K cache, offering performance two to three times faster than the 386 at the same clock rate.
- Early systems were very expensive, ranging from $10,000 to $15,000, with production starting in June 1989 and the first PCs shipping by September.
- Demand for the 486 increased with Windows 3.0 and 3.1, and applications like Doom and later versions of Microsoft Word drove the need for faster processors.
- Alternatives like the Cyrix Cx486SLC and IBM SLC2 used 386 sockets and buses with less cache, offering performance around 50-80% of a real 486, often bottlenecked by the 16-bit interface.
- Upgrade cards, such as the I-O Data PK-X486S50-3, allowed 286 systems to be upgraded to 486-like performance, though compatibility and driver support were considerations.
France Launches Government Linux Desktop Plan as Windows Exit Begins
a day ago
- France's DINUM announces a formal plan to transition government workstations from Windows to Linux as part of its national digital-sovereignty strategy.
- Each ministry and public operator must develop a comprehensive plan by autumn 2026 covering desktop systems, collaboration tools, AI, databases, and more.
- The move aims to reduce reliance on non-European digital technologies, emphasizing sovereignty and interoperability, with decisions on specific Linux distributions pending.
- Public reactions include support for reducing dependence on US tech giants and concerns about transition challenges like user adoption and technical compatibility.
- Some commenters note existing Linux use in French institutions, like the Gendarmerie using Ubuntu, and suggest broader European cooperation for a public-sector OS.
Python Is Dead
a day ago
- AI agents simplify development in strict languages like Rust by offloading compilation and correctness checks, unlike Python which requires constant human oversight.
- Python's design trade-offs (optional typing, interpreted execution, duck typing, GIL) favor human convenience but lead to runtime errors, performance costs, and hidden bugs.
- The rise of AI agents changes the labor equation: strict compile-time guarantees become cost-effective as AI handles complexity without frustration, making Rust-like languages more suitable.
- Future AI-first languages may optimize for token density, formal verifiability, declared effects, parallel-native execution, and structural editing, prioritizing machine efficiency over human readability.
- Programming language choice is effectively a labor policy; with AI, policies that emphasize upfront correctness (e.g., Rust) outweigh those favoring rapid development with high defect rates (e.g., Python).
Sorting Performance Rabbit Hole
5 days ago
- Pystd's stable sort was optimized to be 5% faster than stdlibc++.
- Unstable sort required extensive tweaks, with the biggest improvement coming from changing the insertion sort limit from 16 to 32 elements.
- Despite various attempts like using memmove and different pivot selection, only adjusting the insertion sort threshold significantly boosted performance.
- Pystd's unstable sort finally matched stdlibc++'s speed with a best time of 0.754 seconds versus 0.755 seconds.
France Launches Government Linux Desktop Plan as Windows Exit Begins
a day ago
- The French government is accelerating efforts to reduce its digital dependencies on non-European providers.
- A seminar organized by DINUM, DGE, ANSSI, and DAE set a clear objective to reduce these dependencies, involving ministers, administrations, public operators, and private actors.
- Concrete steps include DINUM switching from Windows to Linux, and the National Health Insurance migrating 80,000 employees to sovereign tools (Tchap, Visio, FranceTransfert).
- A new method involves forming coalitions between ministries, public operators, and private actors to collectively exit dependencies, leveraging digital commons and interoperability standards.
- DINUM will coordinate an inter-ministerial plan, with each ministry required to formalize its own plan by autumn, covering areas like workstations, collaborative tools, antivirus, AI, databases, virtualization, and network equipment.
- Mapping and diagnosis of dependencies by DAE, and defining a European digital service by DGE, will refine reduction targets with clear timelines.
- The first 'digital industrial meetings' in June 2026 will concretize public-private coalitions, including a 'public-private alliance for European sovereignty'.
Model-Based Testing for Dungeons & Dragons
4 days ago
- The article details the expansion of a formal model for Dungeons & Dragons from a single creature to combat interactions, focusing on complex mechanics like Counterspell chains, Shield spells, and Legendary Resistance.
- It explains the interrupt chain during attacks, which includes multiple phases (hit confirmation, damage application, spellcasting, saving throws) that can alter outcomes and branch the game state.
- The spec models deterministic mechanics while leaving non-deterministic aspects (like dice rolls, cover determination, opportunity attack triggers) as caller-provided inputs to accommodate different play styles.
- Model-based testing (MBT) has revealed bugs in both the spec and the implementation, such as argument swaps, invariant violations, and state sync issues, which unit tests often missed.
- The project uses a QA pipeline that scrapes community Q&A entries, translating them into assertions to test the spec against real-world rule interpretations and edge cases.
- The spec includes 62 safety invariants that are proven across all reachable states, not just tested with random inputs, ensuring properties like HP limits hold universally.
- Built on the SRD 5.2.1 (released under CC-BY-4.0), the architecture allows non-SRD content to be added without modifying the core spec, and the model can validate implementations in any language.
- The approach is applicable to other domains with formal rules and combinatorial interactions, such as financial regulations or protocol implementations.
Artemis II and the invisible hazard on the way to the Moon
a day ago
- Artemis II's primary scientific focus is on radiation measurements to enable safe human exploration beyond Earth's magnetic field.
- Radiation hazards include trapped particles in Van Allen belts, solar particle events, and chronic galactic cosmic rays.
- Radiation assessment considers factors like dose rate, particle type, shielding, and spacecraft orientation.
- Artemis II uses multiple tools: cabin monitors, crew-worn dosimeters, upgraded detectors, and biological experiments.
- The mission aims to map radiation geometry with crew onboard, helping decide sheltering strategies during solar events.
- AVATAR experiment uses organ chips from astronauts' bone marrow cells to study radiation's impact on immune function.
- NASA collects immune biomarkers via saliva and blood samples, tracking physiological responses like sleep and cognition.
- Radiation interacts with other mission stressors like microgravity, confinement, and workload, complicating biological effects.
Zero-build privacy policies with Astro
a day ago
- OpenPolicy's Astro integration previously generated Markdown files at build time via a plugin, adding friction with extra packages and generated directories.
- The new approach eliminates the plugin and generated files by directly compiling policies in Astro's frontmatter using three core packages: @openpolicy/sdk, @openpolicy/core, and @openpolicy/renderers.
- Users define a centralized config in src/lib/openpolicy.ts with details for privacy, terms, and cookie policies, which can be filled quickly using AI tools like Claude.
- Each dedicated page (e.g., privacy.astro) imports the config, compiles the policy, and renders it to HTML at build time with zero JavaScript shipped to the browser.
- Benefits include no generated files or directory management, no plugin configuration in astro.config.mjs, support for all three policy types from one config, and simpler inline compilation.
- OpenPolicy+ offers advanced features like cloud-based consent tracking, PR automation for compliance, and hands-on onboarding for further needs.
War on Raze
5 days ago
- The k7 prime number filter uses 'except' (^) with rank-sensitivity, eliminating the need for raze in later versions.
- Earlier versions required raze (,/) to flatten nested results from each-right multiplication, which was seen as a workaround.
- Deep where, inspired by Dyalog APL, offers a way to generate pairs without raze, using an exploded dict representation of data.
- An 'explode' verb transforms nested data into a flat key-value structure, enabling shallow verbs to mimic deep ones.
- The implementation allows for razeless, index-free prime generation, though it may be less efficient than the original.
- Deep verbs like deep find and deep group can be defined using the exploded form, potentially unifying shallow and deep operations.
Not just spyware scandals: EU is funding the industry that spies on Europeans
2 days ago
- Greece sentenced four individuals to prison in February 2026 over the 2022 Predatorgate spyware scandal, marking a rare criminal accountability case for spyware manufacturer executives.
- EU institutions, including the European Commission and member states, have been passive in addressing spyware proliferation, despite its human rights violations.
- Investigations reveal EU funds, such as subsidies from the European Defence Fund and Horizon programmes, are actively funding spyware companies like Intellexa and Paragon Solutions.
- The European Investment Fund, part of the European Investment Bank, funded Israeli spyware company Paragon Solutions, implicated in surveillance scandals in Italy.
- EU bodies lack transparency and accountability in managing public investments, making it difficult to track how taxpayer money supports harmful spyware firms.
- The article calls for a ban on EU funding for spyware, stricter investment policies, and a full EU-wide ban on spyware technology to protect democracy and fundamental rights.
Responsive images in Hugo using Render Hooks
2 days ago
- The article discusses the evolution from using Hugo shortcodes for responsive images (JPEG and WebP) in 2022 to using Render Hooks by 2026, as WebP is now widely supported in browsers.
- A Render Hook is implemented in a file at layouts/_default/_markup/render-image.html, which generates resized WebP and JPEG images, using srcset for WebP and a JPEG fallback, without the need for a <picture> element.
- The Render Hook automatically processes images referenced in Markdown files, supports page bundles (placing images next to Markdown files), and allows quality settings for WebP images via Hugo configuration.
Bun v1.3.12
a day ago
- Bun introduces Bun.WebView, a headless browser automation tool supporting WebKit and Chrome backends, with native OS events ensuring human-like interactions.
- Markdown files can now be rendered in the terminal directly using `bun ./file.md` or programmatically with `Bun.markdown.ansi()`.
- Async stack traces are now available for native APIs, improving debugging capabilities.
- In-process `Bun.cron()` scheduler allows recurring tasks within the application, with UTC-based scheduling and no overlapping executions.
- Improvements to `Bun.udpSocket()` include ICMP error handling without socket closure and detection of truncated datagrams via a flags argument.
- Unix Domain Socket behavior now matches Node.js, with correct EADDRINUSE errors on existing sockets and automatic cleanup after closing.
- JavaScriptCore engine upgraded with over 1,650 commits, adding support for Explicit Resource Management (`using` and `await using`) and various performance enhancements.
- Standalone executables on Linux now use an ELF section for embedded data, eliminating read permissions issues.
- `URLPattern` is up to 2.3x faster, thanks to reduced GC allocations and no pollution of `RegExp.lastMatch`.
- Optimizations to `Bun.stripANSI` and `Bun.stringWidth` using SIMD, resulting in significant performance improvements.
- Fixed a thread-pool bug in `bun build`, speeding up bundling on low-core machines by up to 1.47×.
- `Bun.Glob.scan()` performance improved, with up to 2× faster scans for patterns with boundaries and kernel-level filtering on Windows.
- Cgroup-aware `availableParallelism` and `hardwareConcurrency` on Linux respect CPU limits in Docker and Kubernetes.
- Keep-Alive for HTTPS Proxy CONNECT Tunnels reuses tunnels across requests, reducing latency and overhead.
- `Bun.serve()` on Linux uses `TCP_DEFER_ACCEPT` to reduce connection latency by minimizing event loop wake-ups.
- Multiple bugfixes across Node.js compatibility, Bun APIs, Web APIs, JavaScript bundler, `bun test`, Bun Shell, and Windows support.
First man convicted under Take It Down Act kept making AI nudes after arrest
a day ago
- An Ohio man, James Strahler II, pleaded guilty and became the first person convicted under the Take It Down Act.
- He created and shared real and AI-generated explicit images without consent, targeting at least 10 victims, including women and children.
- AI tools were used to create fake sexualized and incestuous images to harass, coerce, and threaten victims and their families.
- His actions included posting over 700 real and 'animated' images to a child sexual abuse website and other platforms.
- He faces potential prison sentences of up to two years for adult images and up to three years for those involving minors.
Installing OpenBSD on the Pomera DM250{,XY?}
a day ago
- Notes for installing OpenBSD-current on Pomera DM250 models (JP versions, not DM250US).
- Requires custom kernel and U-Boot due to uncommitted upstream work, with risks like battery drain issues.
- Backup eMMC before proceeding, using tools like EKESETE.net.
- Factory U-Boot boots recovery with Right Shift + Left Alt; new U-Boot replaces it, requiring EFI boot for recovery.
- Device may not power off with USB-C connected; battery must not fully drain to avoid boot failures.
- Create installation SD card with GPT, EFI partition, and OpenBSD files including custom U-Boot and script.
- Boot into recovery mode to run script that backs up and flashes new U-Boot.
- Install OpenBSD manually with custom fdisk steps for GPT partitioning on eMMC.
- Post-install steps: copy firmware, disable reorder_kernel, and set up LEDs.
- Recovery options include USB loader mode, U-Boot console commands, or hardware UART access.
VFX HQ: Visual Effects Headquarters (2000)
4 days ago
- Titanic: Collaboration of hundreds of top effects artists for James Cameron's film, with a 9-page FX review and high-resolution stills available.
- Starship Troopers: Interviews with effects artists like Phil Tippett and Scott E. Anderson on Paul Verhoeven's sci-fi thriller.
- The Star Wars Special Editions: Re-release of original trilogy with new visual effects, including debate on pros and cons.
- Flubber: Insights from ILM and Dream Quest supervisors on creating green slime for Disney's remake.
Show HN: Logoshi, a brand kit generator for solo founders
2 days ago
- The service allows you to create a brand kit by entering a few details.
- Brand kit includes: Wordmark in SVG and PNG formats, a Color Palette, Font Pairing, Business Card Template, and a Brand Guide.
- Example brand kit shown is named 'Ember & Oak'.
- Preview of the brand kit is available in minutes.
There are zero-day exploits for your mind
2 days ago
- Anthropic's Claude Mythos discovered thousands of zero-day vulnerabilities in major software systems, revealing hidden security flaws after decades of existence.
- The AI model found exploits with a 72.4% success rate, drastically outperforming previous models and human researchers, indicating a paradigm shift in cybersecurity.
- The article draws a parallel to AlphaGo's 'Move 37' in the game of Go, suggesting that both code and human psychology may have unexplored vulnerabilities.
- Defensive efforts, like patching bugs, are inherently slower and more difficult than exploit creation, highlighting an asymmetric threat landscape.
- AI's exponential progress means that current capabilities are the 'worst' they will ever be, with future models likely to uncover even more vulnerabilities.
- Exploits in legacy systems, such as the 16-year-old bug in FFmpeg, demonstrate challenges in patching widely embedded code and ensuring updates reach all users.
- Human manipulation is entering a new phase where AI could use personal data to discover psychological 'zero-day exploits,' potentially hacking minds in ways we cannot anticipate.
- The trust models of both computing and human communication are based on outdated assumptions about the difficulty of exploitation, necessitating a new paradigm.
- Preparation, such as securing personal data and acknowledging the inevitability of change, is emphasized over panic in facing these evolving threats.
Anthropic Detects Third-Party Clients via System Prompt, Not Headers
2 days ago
- Anthropic blocks third-party clients like OpenCode by analyzing the system prompt content, not through HTTP headers or TLS fingerprinting.
- Replacing the static portion of a custom system prompt with Claude Code's real prompt allows requests to succeed, while keeping custom prompts triggers a block.
- Detection is per-request and based on pattern matching or embedding similarity of the static prompt section, ignoring runtime-injected content like environment variables or custom instructions.
- A fix involves loading Claude Code's system prompt and swapping only the static part, preserving runtime sections, though it requires updates if Claude Code's prompt changes.
- This method is fragile for both Anthropic and users, highlighting potential OAuth architecture limitations in distinguishing official from unofficial clients.

Hasty Briefsbeta

White House staff told not to place bets on prediction markets

FBI used iPhone notification data to retrieve deleted Signal messages

Intel 486 CPU announced April 10, 1989

France Launches Government Linux Desktop Plan as Windows Exit Begins

Python Is Dead

Sorting Performance Rabbit Hole

France Launches Government Linux Desktop Plan as Windows Exit Begins

Model-Based Testing for Dungeons & Dragons

Artemis II and the invisible hazard on the way to the Moon

Zero-build privacy policies with Astro

War on Raze

Not just spyware scandals: EU is funding the industry that spies on Europeans

Responsive images in Hugo using Render Hooks

Bun v1.3.12

First man convicted under Take It Down Act kept making AI nudes after arrest

Installing OpenBSD on the Pomera DM250{,XY?}

VFX HQ: Visual Effects Headquarters (2000)

Show HN: Logoshi, a brand kit generator for solo founders

There are zero-day exploits for your mind

Anthropic Detects Third-Party Clients via System Prompt, Not Headers

White House staff told not to place bets on prediction markets

FBI used iPhone notification data to retrieve deleted Signal messages

Intel 486 CPU announced April 10, 1989

France Launches Government Linux Desktop Plan as Windows Exit Begins

Python Is Dead

Sorting Performance Rabbit Hole

France Launches Government Linux Desktop Plan as Windows Exit Begins

Model-Based Testing for Dungeons & Dragons

Artemis II and the invisible hazard on the way to the Moon

Zero-build privacy policies with Astro

War on Raze

Not just spyware scandals: EU is funding the industry that spies on Europeans

Responsive images in Hugo using Render Hooks

Bun v1.3.12

First man convicted under Take It Down Act kept making AI nudes after arrest

Installing OpenBSD on the Pomera DM250{,XY?}

VFX HQ: Visual Effects Headquarters (2000)

Show HN: Logoshi, a brand kit generator for solo founders

There are zero-day exploits for your mind

Anthropic Detects Third-Party Clients via System Prompt, Not Headers