Hasty Briefsbeta

Bilingual

"Memflation": Cheaper RAM not expected until 2028, says Gartner
a day ago
- DRAM and NAND flash chip prices are expected to rise sharply, with DRAM up 125% and NAND flash up 234% this year.
- Gartner forecasts worldwide semiconductor revenue to grow 64% to $1.3 trillion in 2026, driven largely by high memory chip prices, not increased production volumes.
- 'Memflation' is predicted to weaken or delay non-AI hardware demand until 2028, as expensive memory chips stifle markets outside of AI applications.
- AI chips are expected to make up 30% of all devices sold in 2026, but rising memory prices may also dampen demand for some AI chips due to high DRAM and flash requirements.
- Significant price reductions for RAM and flash memory are not anticipated until late 2027, after a period of sharp price increases.
Artemis II Flight Day 10: Crew Sets for Final Burn, Splashdown
a day ago
- The Artemis II crew (Reid Wiseman, Victor Glover, Christina Koch, Jeremy Hansen) began final return preparations with music and a scheduled correction burn.
- Splashdown targeted for 8:07 p.m. EDT off San Diego, with NASA and U.S. military recovery teams on standby.
- Final return burn at 2:53 p.m. to refine Orion's path for atmospheric entry.
- Key splashdown timeline includes module separation, heat shield exposure, atmospheric entry at 400,000 feet, parachute deployments, and splashdown at 20 mph.
- Crew extraction within two hours post-splashdown via helicopter to USS John P. Murtha, followed by medical evaluations and return to Johnson Space Center.
- Mission launched on April 1 from Kennedy Space Center, featuring a historic lunar flyby and system tests in deep space.
Atanasoff-Berry Computer
a day ago
- The Atanasoff-Berry Computer (ABC) was an electronic computing machine conceived by John Atanasoff in 1937 and built with Clifford Berry from 1939 to 1942 to solve linear equations automatically, though it was not programmable.
- A working reconstruction of the ABC was built between 1994 and 1997, revealing design details like hand-selected resistors and components such as add-subtract circuits, a card reader, binary punch, and memory drums storing 30 numbers each.
- The ENIAC patent lawsuit in 1967 highlighted the ABC's prior work, leading to the invalidation of ENIAC patents after it was ruled that John Mauchly derived ideas from Atanasoff, effectively making the invention of the electronic computer unpatented.
Microsoft Upgrades Its WSL2 Kernel Against Linux 6.18 LTS
5 hours ago
- Microsoft released linux-msft-wsl-6.18.20.1, upgrading the WSL2 kernel to Linux 6.18 LTS series from the previous 6.6 LTS.
- The upgrade allows removal of out-of-tree patches, including for VirtIO PMEM support, compared to the Linux 6.6 base.
- New kernel configuration enables support for F2FS and ExFAT file-systems, ANON_VMA_NAME, CAN options, joystick interface, USB monitor, and FAT support on ARM64 builds.
Borges' cartographers and the tacit skill of reading LM output
6 hours ago
- Borges' story illustrates that maps are useful because they are reductions; when they lack compression, they lose their purpose.
- Language models (LMs) are compared to maps, having evolved beyond mere representations to actively reshape the territory they describe, requiring users to develop skills to engage critically with them.
- Baudrillard's four stages of representation apply to LMs: from faithful copies to distorting reality, masking its absence, and potentially detaching completely, with transitions between stages being slippery and hard to detect.
- Unlike traditional maps, LMs are interactive and personalized, with outputs varying based on prompts and user backgrounds, offering flexibility but also challenging shared understanding.
- LMs serve as useful tools for summarization and reflection, clearing clutter and aiding in tasks like research or coding, but they also become objects of study themselves, shifting focus from the territory to the map.
- Effective use of LMs involves developing tacit knowledge—similar to clinical intuition or code smell—to sense when outputs are unverified, too smooth, or in need of verification, a skill learned through practice rather than rules.
- The analogy of LMs as maps is limited; they are also tools, economic forces, and research subjects, with future improvements making the skill of critical engagement even more important.
- A balanced approach is needed between obsession with LMs and indifference, emphasizing the delicate practice of using them well while staying honest about their evolving role in our lives.
The Problem That Built an Industry
5 hours ago
- In the mid-1950s, American Airlines faced a reservation crisis, managing bookings on index cards and taking up to 90 minutes for transatlantic confirmations, leading to the creation of SABRE in 1964 after a 1953 conversation between airline and IBM executives.
- TPF (Transaction Processing Facility), an IBM mainframe operating system derived from American Airlines' ACP, became the dominant runtime for Global Distribution Systems (GDS), handling up to 10,000 transactions per second with sub-100ms latency, due to its specialized design for high-volume, low-latency transaction processing.
- Modern airline bookings, like those made through MakeMyTrip using Amadeus GDS, rely on decades-old infrastructure and protocols, such as cryptic command language, with data models and conventions established in the 1960s, ensuring consistency across systems.
- Airlines like Air India use Amadeus Altéa for passenger services, while low-cost carriers like IndiGo opt for Navitaire's NewSkies platform, leading to interoperability challenges in re-accommodation during disruptions due to system divergences.
- A single booking triggers a complex chain of synchronous and asynchronous calls across multiple systems, with a PNR (Passenger Name Record) serving as the key identifier holding all components together.
- Key takeaways include: fitness for purpose over trendy architecture, as seen with TPF's enduring efficiency; convergent evolution in GDS platforms; and the high cost and complexity of migrations, exemplified by Air India's 2023 move to Amadeus Altéa.
The Future of Everything Is Lies, I Guess: Annoyances
4 hours ago
- Companies will use machine learning, especially LLMs, for customer service, making it harder to reach humans and leading to frustrating interactions.
- LLMs will be deployed in various 'fuzzy' tasks like insurance, pricing, and compliance, creating new drudgery as people must argue with or manipulate these systems.
- ML models will diffuse responsibility, making it difficult to hold individuals accountable for errors, such as in misidentifications or biased decisions.
- 'Agentic commerce' will involve LLMs handling purchases, leading to new advertising tactics, dark patterns, and potential fraud, with financial risks distributed across transactions.
- The deployment of LLMs may create an arms race between companies and consumers, forcing widespread adoption and resulting in an obnoxious equilibrium with increased complexity and bias.
Cooperative Vectors Introduction
a day ago
- Neural Networks (NN) usage in rendering engine began in 2021, initially for Neural Materials (NM) with offline training and hardware-accelerated inference via early Vulkan extensions.
- Framework expanded to support runtime training for Neural Radiance Caching (NRC), handling small NNs with features like 16, 32, or 64 widths, despite platform-specific hardware acceleration challenges.
- Cooperative Matrix extension (SPV_NV_cooperative_matrix, now KHR) abstracted matrix operations for hardware acceleration, while DirectX's WaveMatrix (Shader Model 6.8) was introduced but not fully released.
- Cooperative Vector (VK_NV_cooperative_vector) addressed divergent data challenges in Neural Materials and Neural Texture Compression (NTC), enabling vector-matrix operations for varied weights per pixel.
- Long vectors in HLSL are transparent data structures stored in VGPRs, accessible across shader stages, with each thread owning its vector for efficient handling of divergent inputs like in NRC.
- Matrices come in plain layouts (row/column major) and optimal layouts (MulOptimal for inference, OuterProductOptimal for training), with optimal layouts being opaque and requiring conversion for accessibility.
- Inference uses MulOptimal layouts via matrix conversions and operations like MatMul or MatMulAdd, while training involves gradients computed with cooperative vectors for input, weights, and bias updates.
- Future developments include DirectX's shift to linear algebra features combining Cooperative Vector and Matrix, and Vulkan's promotion of Cooperative Matrix to KHR with Cooperative Vector remaining NVIDIA-specific.
The mystery of the lost Roman herb (2017)
7 hours ago
- Silphium was a highly prized herb in ancient times, valued for culinary, medicinal, and aphrodisiac uses, and was worth its weight in gold.
- It grew only in a narrow region of Cyrene, Libya, and could not be cultivated, leading to overharvesting and eventual extinction due to overgrazing and smuggling.
- Silphium's identity remains a mystery; it may have been a hybrid plant from the Ferula family, related to fennel, and might still exist undetected in Libya.
- Similar modern examples, like huckleberries and bluefin tuna, face challenges in cultivation and conservation, highlighting ongoing issues with overexploitation.
- The story of silphium underscores broader ecological lessons about sustainable harvesting and the threats to medicinal herbs from climate change and development.
Researchers used AI to analyze 400k Reddit posts, revealing GLP-1 side effects
7 hours ago
- Penn researchers used AI to analyze over 400,000 Reddit posts to identify potential side effects of GLP-1 drugs, such as semaglutide and tirzepatide, that are not fully captured in clinical trials or regulatory documents.
- The study, published in Nature Health, highlighted two main classes of underreported symptoms: reproductive symptoms (e.g., irregular menstrual cycles) and temperature-related complaints (e.g., chills, hot flashes).
- While some known side effects like nausea were detected, confirming the method's reliability, the research found that nearly 4% of users reported menstrual irregularities, suggesting a significant signal worth further investigation.
- The researchers employed large language models (e.g., GPT, Gemini) to standardize and scale the analysis of social media data, mapping posts to the Medical Dictionary for Regulatory Activities (MedDRA) for efficient symptom identification.
- Despite limitations, such as the non-representative sample of Reddit users, the findings indicate that AI-assisted social media analysis can offer rapid insights into patient-reported side effects, complementing slower clinical trials.
- Future directions include expanding the research to other platforms and languages to validate findings globally and using this approach for early detection of side effects in emerging drugs and wellness trends.
Show HN: Hormuz Havoc, a satirical game that got overrun by AI bots in 24 hours
8 hours ago
- The text presents a game interface titled 'Presidential Panic: Hormuz Havoc', indicating a scenario involving crisis management in the Hormuz region.
- Gameplay elements include tracking oil price limits, presidential approval ratings, personal enrichment, and a scoring system, with a weekly progression structure.
- The interface includes options for player actions, a leaderboard comparing human and AI-assisted performances, and features for sharing scores or replaying.
- A breaking news prompt asks the player, addressed as 'Mr President', to confirm their name, adding an interactive, role-playing dimension to the experience.
Bitcoin miners are losing $19,000 on every BTC produced as difficulty drops 7.8%
5 hours ago
- Bitcoin miners are facing significant losses, with production costs around $88,000 per coin against a market price near $69,200, leading to a loss of about $19,000 per bitcoin mined.
- Geopolitical tensions, including high oil prices above $100 and disruptions like the closure of the Strait of Hormuz, are increasing electricity costs and contributing to a decline in hashrate and network difficulty.
- The Bitcoin network difficulty dropped 7.76% to 133.79 trillion, with hashrate falling to roughly 920 EH/s and average block times exceeding the 10-minute target, indicating network stress.
- Miners are selling bitcoin to cover costs and diversifying into AI and high-performance computing for steadier revenue, adding supply pressure to a market with significant underwater holdings.
- Further difficulty declines are expected if bitcoin prices remain below production costs, potentially leading to a continued miner exodus and forced selling, impacting both miners and the spot market.
Cirrus Labs to join OpenAI shut down Circus CI on Monday, June 1, 2026
6 hours ago
- Cirrus Labs, founded in 2017 to create tooling for engineers in the cloud computing era, is joining OpenAI's Agent Infrastructure team.
- The company innovated in CI/CD systems and virtualization, notably with Tart for Apple Silicon, and operated without outside capital.
- Existing products like Tart, Vetu, and Orchard will be relicensed under a permissive license with no fees; Cirrus CI will shut down on June 1, 2026.
- Cirrus Runners will not accept new customers but will support existing ones through their contracts.
- The move aims to extend Cirrus Labs' mission to build tooling for both human and agentic engineers, aligning with the agentic engineering era.
How to breathe in fewer microplastics in your home
6 hours ago
- Microplastics from synthetic fabrics, furnishings, and carpets concentrate in indoor air, potentially leading to inhalation of hundreds of thousands to millions of particles annually.
- Inhalation may be the primary route of microplastic exposure, with indoor concentrations up to eight times higher than outdoors, partly due to people spending about 90% of their time inside.
- Reducing exposure involves behavioral changes, such as washing clothes less frequently, using washing machine filters, drying clothes outside, and opting for natural-fiber clothing and furnishings.
- Vacuuming with HEPA filters and wet cleaning can help remove settled microplastics, but may resuspend particles temporarily; wearing masks during cleaning and ensuring good ventilation is advised.
- Air filters with multiple stages can effectively capture microplastics, while air conditioning may increase their distribution; health impacts include inflammation and organ accumulation, particularly from fibrous particles.
Doctor behind vaccine-autism link loses license (2010)
9 hours ago
- Dr. Andrew Wakefield lost his UK medical license after the GMC ruled he acted dishonestly and irresponsibly in autism-vaccine research.
- His 1998 Lancet paper linked MMR vaccines to autism in eight children, but subsequent studies failed to confirm his claims, and the Lancet retracted it.
- Wakefield faced misconduct charges, including conducting studies without ethical approval and paying children for blood samples, but he maintains vaccines cause autism and plans to appeal.
’Abhorrent’: the inside story of the Polymarket gamblers betting millions on war
6 hours ago
- Anonymous gamblers are betting heavily on war-related events through prediction markets like Polymarket, with over $500,000 wagered on whether Russia will capture Kostyantynivka this year.
- Polymarket users sometimes collaborate to influence events or how they are recorded, leading to ethical concerns and calls for regulation from critics like US senators.
- The platform has grown rapidly, trading over $400 million in 2024 and now handling similar amounts daily, positioning itself as a source of truth for future predictions.
- Some users compare Polymarket to a casino, where even tragic events like war are monetized, and there are instances of gamblers threatening journalists to sway bet outcomes.
- Prediction Hunt, a Discord community, helps users share tips for betting on Polymarket, including tracking 'fade' bets, insider actions, and arbitrage opportunities.
- Disputes over bet outcomes are settled by anonymous holders of the UMA crypto token, raising transparency issues and potential corruption risks.
- Major financial institutions like the Intercontinental Exchange and Goldman Sachs are investing in or using Polymarket's data, integrating its predictions into broader markets.
- Despite moral dilemmas, supporters argue prediction markets offer clarity in uncertain situations, such as during war, where traditional information sources may be unreliable.
- The platform's influence could lead to manipulation of larger financial markets, as small bets might skew perceived odds and affect investment decisions.
Splitting the Web (2023)
13 hours ago
- The modern web is splitting into two extremes: commercial, ad-driven sites vs. minimalist, tech-savvy alternatives.
- Tech-savvy users block ads, avoid JavaScript-heavy pages, and use privacy-focused tools like adblockers, Mastodon, or Gemini.
- This divide forces users to choose sides, with the commercial web requiring disabling blockers for access, leading to disconnection.
- Personal strategies include using privacy search engines like Kagi.com, avoiding analytics, and opting for phone/email over complex sites.
- Despite rejecting mainstream web practices, engagement via blogs, Mastodon, RSS feeds, and minimalist sites remains high and appreciated.
- The author suggests the web isn't dying but splitting, with 'dark web' referring to non-commercial, ad-free spaces for discussion.
- A call to embrace the simple web, rejecting ad-driven models seen as destructive and intrusive.
Gen Z workers who fear AI will take their job actively sabotaging its rollout
10 hours ago
- A significant share of employees, 29% overall and 44% of Gen Z, admit to sabotaging their company's AI strategy through actions like entering proprietary information into public tools or refusing to use AI.
- Fear of job loss (FOBO) is a major driver, with 30% of saboteurs citing it, yet workers resisting AI adoption are more vulnerable to layoffs; 60% of executives consider cutting such employees.
- AI "super-users" who actively integrate AI into workflows are rewarded, being three times more likely to receive promotions and pay raises, and save more time compared to laggards.
- Many workers hold negative views of AI, with 46% of U.S. voters having a negative perception, fueled by warnings from experts about AI automating white-collar jobs across fields like law and finance.
- Successful companies focus on redesigning operations with human-agent collaboration rather than layoffs, as AI adoption gaps often stem from organizational learning issues rather than technology quality.
Artemis II is competency porn
10 hours ago
- The Artemis II mission evoked strong emotional reactions from millions, offering a sense of relief by showcasing competence and success, contrasting with years of anxiety and dread.
- Crew members like Reid Wiseman, Christina Koch, and Victor Glover brought personal and historic significance, with gestures like naming a crater after Wiseman's late wife highlighting human connection and inclusion.
- Psychological effects such as awe and the 'overview effect' from the mission reduced fear, broadened thinking, and provided perspective, emphasizing humanity's resilience and capacity for wonder beyond current crises.
BlueHammer abuses Windows Defender's update process to gain SYSTEM access
9 hours ago
- BlueHammer is a zero-day exploit targeting Windows Defender, chaining five legitimate Windows components to achieve privilege escalation from low-privileged accounts to SYSTEM level.
- The exploit was publicly disclosed by a researcher under the alias Chaotic Eclipse, with full source code on GitHub, after Microsoft allegedly broke an agreement.
- BlueHammer requires a pending Defender signature update to trigger; it leverages Defender's update process, Volume Shadow Copy, Cloud Files API, opportunistic locks, and Defender's RPC interface.
- It works by freezing Defender during a signature update, accessing the SAM database via a shadow copy, decrypting NTLM hashes, changing an admin password, and spawning a SYSTEM command prompt before restoring the original password hash.
- Microsoft has only released a signature update that detects the original compiled binary, but the technique remains undetected; there is no patch or CVE, and the exploit is confirmed working on fully updated Windows 10 and 11.
- Security measures to detect BlueHammer include monitoring for VSS enumeration by user processes, Cloud Files sync root registration by unknown processes, and alerts on local admin password changes or service creation.

first prev0next

About|Login

Bilingual

"Memflation": Cheaper RAM not expected until 2028, says Gartner
a day ago
- DRAM and NAND flash chip prices are expected to rise sharply, with DRAM up 125% and NAND flash up 234% this year.
- Gartner forecasts worldwide semiconductor revenue to grow 64% to $1.3 trillion in 2026, driven largely by high memory chip prices, not increased production volumes.
- 'Memflation' is predicted to weaken or delay non-AI hardware demand until 2028, as expensive memory chips stifle markets outside of AI applications.
- AI chips are expected to make up 30% of all devices sold in 2026, but rising memory prices may also dampen demand for some AI chips due to high DRAM and flash requirements.
- Significant price reductions for RAM and flash memory are not anticipated until late 2027, after a period of sharp price increases.
Artemis II Flight Day 10: Crew Sets for Final Burn, Splashdown
a day ago
- The Artemis II crew (Reid Wiseman, Victor Glover, Christina Koch, Jeremy Hansen) began final return preparations with music and a scheduled correction burn.
- Splashdown targeted for 8:07 p.m. EDT off San Diego, with NASA and U.S. military recovery teams on standby.
- Final return burn at 2:53 p.m. to refine Orion's path for atmospheric entry.
- Key splashdown timeline includes module separation, heat shield exposure, atmospheric entry at 400,000 feet, parachute deployments, and splashdown at 20 mph.
- Crew extraction within two hours post-splashdown via helicopter to USS John P. Murtha, followed by medical evaluations and return to Johnson Space Center.
- Mission launched on April 1 from Kennedy Space Center, featuring a historic lunar flyby and system tests in deep space.
Atanasoff-Berry Computer
a day ago
- The Atanasoff-Berry Computer (ABC) was an electronic computing machine conceived by John Atanasoff in 1937 and built with Clifford Berry from 1939 to 1942 to solve linear equations automatically, though it was not programmable.
- A working reconstruction of the ABC was built between 1994 and 1997, revealing design details like hand-selected resistors and components such as add-subtract circuits, a card reader, binary punch, and memory drums storing 30 numbers each.
- The ENIAC patent lawsuit in 1967 highlighted the ABC's prior work, leading to the invalidation of ENIAC patents after it was ruled that John Mauchly derived ideas from Atanasoff, effectively making the invention of the electronic computer unpatented.
Microsoft Upgrades Its WSL2 Kernel Against Linux 6.18 LTS
5 hours ago
- Microsoft released linux-msft-wsl-6.18.20.1, upgrading the WSL2 kernel to Linux 6.18 LTS series from the previous 6.6 LTS.
- The upgrade allows removal of out-of-tree patches, including for VirtIO PMEM support, compared to the Linux 6.6 base.
- New kernel configuration enables support for F2FS and ExFAT file-systems, ANON_VMA_NAME, CAN options, joystick interface, USB monitor, and FAT support on ARM64 builds.
Borges' cartographers and the tacit skill of reading LM output
6 hours ago
- Borges' story illustrates that maps are useful because they are reductions; when they lack compression, they lose their purpose.
- Language models (LMs) are compared to maps, having evolved beyond mere representations to actively reshape the territory they describe, requiring users to develop skills to engage critically with them.
- Baudrillard's four stages of representation apply to LMs: from faithful copies to distorting reality, masking its absence, and potentially detaching completely, with transitions between stages being slippery and hard to detect.
- Unlike traditional maps, LMs are interactive and personalized, with outputs varying based on prompts and user backgrounds, offering flexibility but also challenging shared understanding.
- LMs serve as useful tools for summarization and reflection, clearing clutter and aiding in tasks like research or coding, but they also become objects of study themselves, shifting focus from the territory to the map.
- Effective use of LMs involves developing tacit knowledge—similar to clinical intuition or code smell—to sense when outputs are unverified, too smooth, or in need of verification, a skill learned through practice rather than rules.
- The analogy of LMs as maps is limited; they are also tools, economic forces, and research subjects, with future improvements making the skill of critical engagement even more important.
- A balanced approach is needed between obsession with LMs and indifference, emphasizing the delicate practice of using them well while staying honest about their evolving role in our lives.
The Problem That Built an Industry
5 hours ago
- In the mid-1950s, American Airlines faced a reservation crisis, managing bookings on index cards and taking up to 90 minutes for transatlantic confirmations, leading to the creation of SABRE in 1964 after a 1953 conversation between airline and IBM executives.
- TPF (Transaction Processing Facility), an IBM mainframe operating system derived from American Airlines' ACP, became the dominant runtime for Global Distribution Systems (GDS), handling up to 10,000 transactions per second with sub-100ms latency, due to its specialized design for high-volume, low-latency transaction processing.
- Modern airline bookings, like those made through MakeMyTrip using Amadeus GDS, rely on decades-old infrastructure and protocols, such as cryptic command language, with data models and conventions established in the 1960s, ensuring consistency across systems.
- Airlines like Air India use Amadeus Altéa for passenger services, while low-cost carriers like IndiGo opt for Navitaire's NewSkies platform, leading to interoperability challenges in re-accommodation during disruptions due to system divergences.
- A single booking triggers a complex chain of synchronous and asynchronous calls across multiple systems, with a PNR (Passenger Name Record) serving as the key identifier holding all components together.
- Key takeaways include: fitness for purpose over trendy architecture, as seen with TPF's enduring efficiency; convergent evolution in GDS platforms; and the high cost and complexity of migrations, exemplified by Air India's 2023 move to Amadeus Altéa.
The Future of Everything Is Lies, I Guess: Annoyances
4 hours ago
- Companies will use machine learning, especially LLMs, for customer service, making it harder to reach humans and leading to frustrating interactions.
- LLMs will be deployed in various 'fuzzy' tasks like insurance, pricing, and compliance, creating new drudgery as people must argue with or manipulate these systems.
- ML models will diffuse responsibility, making it difficult to hold individuals accountable for errors, such as in misidentifications or biased decisions.
- 'Agentic commerce' will involve LLMs handling purchases, leading to new advertising tactics, dark patterns, and potential fraud, with financial risks distributed across transactions.
- The deployment of LLMs may create an arms race between companies and consumers, forcing widespread adoption and resulting in an obnoxious equilibrium with increased complexity and bias.
Cooperative Vectors Introduction
a day ago
- Neural Networks (NN) usage in rendering engine began in 2021, initially for Neural Materials (NM) with offline training and hardware-accelerated inference via early Vulkan extensions.
- Framework expanded to support runtime training for Neural Radiance Caching (NRC), handling small NNs with features like 16, 32, or 64 widths, despite platform-specific hardware acceleration challenges.
- Cooperative Matrix extension (SPV_NV_cooperative_matrix, now KHR) abstracted matrix operations for hardware acceleration, while DirectX's WaveMatrix (Shader Model 6.8) was introduced but not fully released.
- Cooperative Vector (VK_NV_cooperative_vector) addressed divergent data challenges in Neural Materials and Neural Texture Compression (NTC), enabling vector-matrix operations for varied weights per pixel.
- Long vectors in HLSL are transparent data structures stored in VGPRs, accessible across shader stages, with each thread owning its vector for efficient handling of divergent inputs like in NRC.
- Matrices come in plain layouts (row/column major) and optimal layouts (MulOptimal for inference, OuterProductOptimal for training), with optimal layouts being opaque and requiring conversion for accessibility.
- Inference uses MulOptimal layouts via matrix conversions and operations like MatMul or MatMulAdd, while training involves gradients computed with cooperative vectors for input, weights, and bias updates.
- Future developments include DirectX's shift to linear algebra features combining Cooperative Vector and Matrix, and Vulkan's promotion of Cooperative Matrix to KHR with Cooperative Vector remaining NVIDIA-specific.
The mystery of the lost Roman herb (2017)
7 hours ago
- Silphium was a highly prized herb in ancient times, valued for culinary, medicinal, and aphrodisiac uses, and was worth its weight in gold.
- It grew only in a narrow region of Cyrene, Libya, and could not be cultivated, leading to overharvesting and eventual extinction due to overgrazing and smuggling.
- Silphium's identity remains a mystery; it may have been a hybrid plant from the Ferula family, related to fennel, and might still exist undetected in Libya.
- Similar modern examples, like huckleberries and bluefin tuna, face challenges in cultivation and conservation, highlighting ongoing issues with overexploitation.
- The story of silphium underscores broader ecological lessons about sustainable harvesting and the threats to medicinal herbs from climate change and development.
Researchers used AI to analyze 400k Reddit posts, revealing GLP-1 side effects
7 hours ago
- Penn researchers used AI to analyze over 400,000 Reddit posts to identify potential side effects of GLP-1 drugs, such as semaglutide and tirzepatide, that are not fully captured in clinical trials or regulatory documents.
- The study, published in Nature Health, highlighted two main classes of underreported symptoms: reproductive symptoms (e.g., irregular menstrual cycles) and temperature-related complaints (e.g., chills, hot flashes).
- While some known side effects like nausea were detected, confirming the method's reliability, the research found that nearly 4% of users reported menstrual irregularities, suggesting a significant signal worth further investigation.
- The researchers employed large language models (e.g., GPT, Gemini) to standardize and scale the analysis of social media data, mapping posts to the Medical Dictionary for Regulatory Activities (MedDRA) for efficient symptom identification.
- Despite limitations, such as the non-representative sample of Reddit users, the findings indicate that AI-assisted social media analysis can offer rapid insights into patient-reported side effects, complementing slower clinical trials.
- Future directions include expanding the research to other platforms and languages to validate findings globally and using this approach for early detection of side effects in emerging drugs and wellness trends.
Show HN: Hormuz Havoc, a satirical game that got overrun by AI bots in 24 hours
8 hours ago
- The text presents a game interface titled 'Presidential Panic: Hormuz Havoc', indicating a scenario involving crisis management in the Hormuz region.
- Gameplay elements include tracking oil price limits, presidential approval ratings, personal enrichment, and a scoring system, with a weekly progression structure.
- The interface includes options for player actions, a leaderboard comparing human and AI-assisted performances, and features for sharing scores or replaying.
- A breaking news prompt asks the player, addressed as 'Mr President', to confirm their name, adding an interactive, role-playing dimension to the experience.
Bitcoin miners are losing $19,000 on every BTC produced as difficulty drops 7.8%
5 hours ago
- Bitcoin miners are facing significant losses, with production costs around $88,000 per coin against a market price near $69,200, leading to a loss of about $19,000 per bitcoin mined.
- Geopolitical tensions, including high oil prices above $100 and disruptions like the closure of the Strait of Hormuz, are increasing electricity costs and contributing to a decline in hashrate and network difficulty.
- The Bitcoin network difficulty dropped 7.76% to 133.79 trillion, with hashrate falling to roughly 920 EH/s and average block times exceeding the 10-minute target, indicating network stress.
- Miners are selling bitcoin to cover costs and diversifying into AI and high-performance computing for steadier revenue, adding supply pressure to a market with significant underwater holdings.
- Further difficulty declines are expected if bitcoin prices remain below production costs, potentially leading to a continued miner exodus and forced selling, impacting both miners and the spot market.
Cirrus Labs to join OpenAI shut down Circus CI on Monday, June 1, 2026
6 hours ago
- Cirrus Labs, founded in 2017 to create tooling for engineers in the cloud computing era, is joining OpenAI's Agent Infrastructure team.
- The company innovated in CI/CD systems and virtualization, notably with Tart for Apple Silicon, and operated without outside capital.
- Existing products like Tart, Vetu, and Orchard will be relicensed under a permissive license with no fees; Cirrus CI will shut down on June 1, 2026.
- Cirrus Runners will not accept new customers but will support existing ones through their contracts.
- The move aims to extend Cirrus Labs' mission to build tooling for both human and agentic engineers, aligning with the agentic engineering era.
How to breathe in fewer microplastics in your home
6 hours ago
- Microplastics from synthetic fabrics, furnishings, and carpets concentrate in indoor air, potentially leading to inhalation of hundreds of thousands to millions of particles annually.
- Inhalation may be the primary route of microplastic exposure, with indoor concentrations up to eight times higher than outdoors, partly due to people spending about 90% of their time inside.
- Reducing exposure involves behavioral changes, such as washing clothes less frequently, using washing machine filters, drying clothes outside, and opting for natural-fiber clothing and furnishings.
- Vacuuming with HEPA filters and wet cleaning can help remove settled microplastics, but may resuspend particles temporarily; wearing masks during cleaning and ensuring good ventilation is advised.
- Air filters with multiple stages can effectively capture microplastics, while air conditioning may increase their distribution; health impacts include inflammation and organ accumulation, particularly from fibrous particles.
Doctor behind vaccine-autism link loses license (2010)
9 hours ago
- Dr. Andrew Wakefield lost his UK medical license after the GMC ruled he acted dishonestly and irresponsibly in autism-vaccine research.
- His 1998 Lancet paper linked MMR vaccines to autism in eight children, but subsequent studies failed to confirm his claims, and the Lancet retracted it.
- Wakefield faced misconduct charges, including conducting studies without ethical approval and paying children for blood samples, but he maintains vaccines cause autism and plans to appeal.
’Abhorrent’: the inside story of the Polymarket gamblers betting millions on war
6 hours ago
- Anonymous gamblers are betting heavily on war-related events through prediction markets like Polymarket, with over $500,000 wagered on whether Russia will capture Kostyantynivka this year.
- Polymarket users sometimes collaborate to influence events or how they are recorded, leading to ethical concerns and calls for regulation from critics like US senators.
- The platform has grown rapidly, trading over $400 million in 2024 and now handling similar amounts daily, positioning itself as a source of truth for future predictions.
- Some users compare Polymarket to a casino, where even tragic events like war are monetized, and there are instances of gamblers threatening journalists to sway bet outcomes.
- Prediction Hunt, a Discord community, helps users share tips for betting on Polymarket, including tracking 'fade' bets, insider actions, and arbitrage opportunities.
- Disputes over bet outcomes are settled by anonymous holders of the UMA crypto token, raising transparency issues and potential corruption risks.
- Major financial institutions like the Intercontinental Exchange and Goldman Sachs are investing in or using Polymarket's data, integrating its predictions into broader markets.
- Despite moral dilemmas, supporters argue prediction markets offer clarity in uncertain situations, such as during war, where traditional information sources may be unreliable.
- The platform's influence could lead to manipulation of larger financial markets, as small bets might skew perceived odds and affect investment decisions.
Splitting the Web (2023)
13 hours ago
- The modern web is splitting into two extremes: commercial, ad-driven sites vs. minimalist, tech-savvy alternatives.
- Tech-savvy users block ads, avoid JavaScript-heavy pages, and use privacy-focused tools like adblockers, Mastodon, or Gemini.
- This divide forces users to choose sides, with the commercial web requiring disabling blockers for access, leading to disconnection.
- Personal strategies include using privacy search engines like Kagi.com, avoiding analytics, and opting for phone/email over complex sites.
- Despite rejecting mainstream web practices, engagement via blogs, Mastodon, RSS feeds, and minimalist sites remains high and appreciated.
- The author suggests the web isn't dying but splitting, with 'dark web' referring to non-commercial, ad-free spaces for discussion.
- A call to embrace the simple web, rejecting ad-driven models seen as destructive and intrusive.
Gen Z workers who fear AI will take their job actively sabotaging its rollout
10 hours ago
- A significant share of employees, 29% overall and 44% of Gen Z, admit to sabotaging their company's AI strategy through actions like entering proprietary information into public tools or refusing to use AI.
- Fear of job loss (FOBO) is a major driver, with 30% of saboteurs citing it, yet workers resisting AI adoption are more vulnerable to layoffs; 60% of executives consider cutting such employees.
- AI "super-users" who actively integrate AI into workflows are rewarded, being three times more likely to receive promotions and pay raises, and save more time compared to laggards.
- Many workers hold negative views of AI, with 46% of U.S. voters having a negative perception, fueled by warnings from experts about AI automating white-collar jobs across fields like law and finance.
- Successful companies focus on redesigning operations with human-agent collaboration rather than layoffs, as AI adoption gaps often stem from organizational learning issues rather than technology quality.
Artemis II is competency porn
10 hours ago
- The Artemis II mission evoked strong emotional reactions from millions, offering a sense of relief by showcasing competence and success, contrasting with years of anxiety and dread.
- Crew members like Reid Wiseman, Christina Koch, and Victor Glover brought personal and historic significance, with gestures like naming a crater after Wiseman's late wife highlighting human connection and inclusion.
- Psychological effects such as awe and the 'overview effect' from the mission reduced fear, broadened thinking, and provided perspective, emphasizing humanity's resilience and capacity for wonder beyond current crises.
BlueHammer abuses Windows Defender's update process to gain SYSTEM access
9 hours ago
- BlueHammer is a zero-day exploit targeting Windows Defender, chaining five legitimate Windows components to achieve privilege escalation from low-privileged accounts to SYSTEM level.
- The exploit was publicly disclosed by a researcher under the alias Chaotic Eclipse, with full source code on GitHub, after Microsoft allegedly broke an agreement.
- BlueHammer requires a pending Defender signature update to trigger; it leverages Defender's update process, Volume Shadow Copy, Cloud Files API, opportunistic locks, and Defender's RPC interface.
- It works by freezing Defender during a signature update, accessing the SAM database via a shadow copy, decrypting NTLM hashes, changing an admin password, and spawning a SYSTEM command prompt before restoring the original password hash.
- Microsoft has only released a signature update that detects the original compiled binary, but the technique remains undetected; there is no patch or CVE, and the exploit is confirmed working on fully updated Windows 10 and 11.
- Security measures to detect BlueHammer include monitoring for VSS enumeration by user processes, Cloud Files sync root registration by unknown processes, and alerts on local admin password changes or service creation.

Hasty Briefsbeta

"Memflation": Cheaper RAM not expected until 2028, says Gartner

Artemis II Flight Day 10: Crew Sets for Final Burn, Splashdown

Atanasoff-Berry Computer

Microsoft Upgrades Its WSL2 Kernel Against Linux 6.18 LTS

Borges' cartographers and the tacit skill of reading LM output

The Problem That Built an Industry

The Future of Everything Is Lies, I Guess: Annoyances

Cooperative Vectors Introduction

The mystery of the lost Roman herb (2017)

Researchers used AI to analyze 400k Reddit posts, revealing GLP-1 side effects

Show HN: Hormuz Havoc, a satirical game that got overrun by AI bots in 24 hours

Bitcoin miners are losing $19,000 on every BTC produced as difficulty drops 7.8%

Cirrus Labs to join OpenAI shut down Circus CI on Monday, June 1, 2026

How to breathe in fewer microplastics in your home

Doctor behind vaccine-autism link loses license (2010)

’Abhorrent’: the inside story of the Polymarket gamblers betting millions on war

Splitting the Web (2023)

Gen Z workers who fear AI will take their job actively sabotaging its rollout

Artemis II is competency porn

BlueHammer abuses Windows Defender's update process to gain SYSTEM access

"Memflation": Cheaper RAM not expected until 2028, says Gartner

Artemis II Flight Day 10: Crew Sets for Final Burn, Splashdown

Atanasoff-Berry Computer

Microsoft Upgrades Its WSL2 Kernel Against Linux 6.18 LTS

Borges' cartographers and the tacit skill of reading LM output

The Problem That Built an Industry

The Future of Everything Is Lies, I Guess: Annoyances

Cooperative Vectors Introduction

The mystery of the lost Roman herb (2017)

Researchers used AI to analyze 400k Reddit posts, revealing GLP-1 side effects

Show HN: Hormuz Havoc, a satirical game that got overrun by AI bots in 24 hours

Bitcoin miners are losing $19,000 on every BTC produced as difficulty drops 7.8%

Cirrus Labs to join OpenAI shut down Circus CI on Monday, June 1, 2026

How to breathe in fewer microplastics in your home

Doctor behind vaccine-autism link loses license (2010)

’Abhorrent’: the inside story of the Polymarket gamblers betting millions on war

Splitting the Web (2023)

Gen Z workers who fear AI will take their job actively sabotaging its rollout

Artemis II is competency porn

BlueHammer abuses Windows Defender's update process to gain SYSTEM access