Popular：
Virtualization DNS security formal verification reachability analysis compiler errors macro conflict web extension development framework Bitmap Graphics API inconsistencies All Tags

Python Core Devs Summit: JITs, Virtual Threads, and the Future

2025-06-15
Python Core Devs Summit: JITs, Virtual Threads, and the Future

The annual Python core developers' summit showcased exciting discussions. Meta engineers explored pluggable JIT compilers, aiming to simplify development via new APIs. Insights from Java spurred discussions on virtual threads for Python, boosting concurrency. The summit also featured debates on null-coalescing operators, AI-assisted coding tools, and the 'worse is better' philosophy. Finally, developers called for memory benchmark focus and delved into the future evolution of T-strings' type system.

Read more
(pyfound.blogspot.com)
Development

Restore Oddly Shaped App Icons in macOS 26 Tahoe

2025-06-15
Restore Oddly Shaped App Icons in macOS 26 Tahoe

macOS 26 Tahoe replaces the unique, oddly shaped app icons in the Dock with iOS-style squarcles, a change many users dislike. This article provides a solution for both users and developers to restore custom icon shapes. Users can replace the .icns file within the application package; developers can use NSApplication.shared.dockTile.contentView to change the icon at runtime. Get your Dock back to its former glory!

Read more
(simonbs.dev)
Development App Icons

The Scalability Challenge of Reinforcement Learning: Can Q-Learning Handle Long Horizons?

2025-06-15

Recent years have witnessed the scalability of many machine learning objectives, such as next-token prediction, denoising diffusion, and contrastive learning. However, reinforcement learning (RL), particularly off-policy RL based on Q-learning, faces challenges in scaling to complex, long-horizon problems. This article argues that existing Q-learning algorithms struggle with problems requiring more than 100 semantic decision steps due to accumulating bias in prediction targets. Experiments show that even with abundant data and controlled variables, standard off-policy RL algorithms fail to solve complex tasks. However, horizon reduction significantly improves scalability, suggesting the need for better algorithms that directly address the horizon problem rather than solely relying on increased data and compute.

Read more
(seohong.me)
AI Q-learning Scalability

AMD Doubles Down on AI with Instinct MI350 Series and ROCm 7

2025-06-15
AMD Doubles Down on AI with Instinct MI350 Series and ROCm 7

AMD unveiled its next-generation Instinct MI350 series AI accelerators, boasting double the AI performance of its predecessor, the MI300X, thanks to the new CDNA 4 architecture. The MI350 series supports FP6 and FP4 formats for increased throughput and features 288GB of HBM3E memory with 8TB/s bandwidth. Complementing the hardware is ROCm 7, delivering performance improvements and day-0 support. AMD also announced turnkey rack-scale AI solutions integrating AMD CPUs, GPUs, and networking, and laid out a roadmap targeting a 20x increase in rack-scale energy efficiency by 2030. The MI355X, the flagship model, offers up to 5 PFLOPS of FP16 performance.

Read more
(morethanmoore.substack.com)
Hardware AI Accelerators

From SQL Server to Key-Value Stores: A Postmortem of a Database Rewrite Gone Wrong

2025-06-15

A senior developer recounts their experience with a database rewrite at a previous company. The original system, using SQL Server, suffered from performance bottlenecks and frequent outages due to complex stored procedures. The rewrite opted for simple key-value stores, but due to data model mismatch and lack of transaction support, I/O operations skyrocketed, performance worsened, and a complex checkpointing system was introduced. The rewrite ultimately failed to solve the original problems and created new challenges. This humorous account reflects on the importance of technology selection and architectural design, and the negative impact of oversimplification.

Read more
(www.luu.io)
Development database rewrite

Critical Vulnerability in YubiKey 5's Secure Element: Key Extraction Achieved

2025-06-15
Critical Vulnerability in YubiKey 5's Secure Element: Key Extraction Achieved

Security researcher Thomas Roche uncovered a critical vulnerability in the Infineon SLE78 secure element used in YubiKey 5. Through side-channel attacks, he successfully extracted the ECDSA secret key. This vulnerability affects multiple versions of Infineon's cryptographic library, impacting security across automotive, medical, industrial control, and avionics sectors. The researcher urges industries to address this vulnerability immediately.

Read more
(it4sec.substack.com)
Tech side-channel attack

AI-Powered Student Loan Fraud Explodes: Colleges and Students Targeted

2025-06-15
AI-Powered Student Loan Fraud Explodes: Colleges and Students Targeted

The rise of AI and online classes has fueled a surge in student loan fraud. Criminal rings are deploying AI chatbots as "ghost students," enrolling in online courses and collecting financial aid. California colleges reported 1.2 million fraudulent applications in 2024, resulting in 223,000 suspected fake enrollments and at least $11.1 million in losses. Victims face not only significant debt but also potential inability to enroll in needed courses due to bots filling class rosters. The US Department of Education has implemented temporary measures requiring government-issued ID, but more robust long-term solutions are under development.

Read more
(apnews.com)
Tech student loans

AI-Powered Art Restoration: 600-Year-Old Painting Restored in 3.5 Hours

2025-06-15
AI-Powered Art Restoration: 600-Year-Old Painting Restored in 3.5 Hours

MIT researchers have developed a new method using AI and 3D printing to apply digital restorations directly to original paintings. This method automatically identifies and repairs thousands of damaged areas, 66 times faster than traditional methods. A 15th-century painting, traditionally requiring years to restore, was completed in 3.5 hours. While ethical considerations exist, this technology promises to accelerate the restoration of numerous damaged artworks, bringing more art to the public eye.

Read more
(news.mit.edu)
Tech Art Restoration

Chess Elo Surge: A Simple Technique Changed Everything

2025-06-15
Chess Elo Surge: A Simple Technique Changed Everything

A chess player consistently performed better in blitz than bullet chess. Switching from drag-and-drop to click-and-click piece movement shaved 0.25 seconds off his average move time. This seemingly minor change resulted in a significant Elo gain in bullet chess, halving the gap between his blitz and bullet ratings. He attributes the improvement to having more time to think and less pre-moving, reducing blunders.

Read more
(jacobbrazeal.wordpress.com)
Game game improvement

Ethanol Expansion: Hidden Costs and Growing Inequality

2025-06-14
Ethanol Expansion: Hidden Costs and Growing Inequality

A new report concludes that expanding ethanol production not only increases greenhouse gas emissions but also fails to deliver promised social and economic benefits to Midwestern communities. Subsidies primarily benefit large agribusinesses, leading to land consolidation and hindering smaller farmers. Proposed policies could exacerbate these issues, increasing land conversion and emissions. Trump's tax cuts further complicate the problem by excluding land conversion emissions from low-emission fuel calculations. Biofuel industry groups haven't responded, but some argue the report overstates carbon emissions and highlight the economic contributions and job creation from soy-based fuels.

Read more
(arstechnica.com)
Tech ethanol greenhouse gas emissions

A Hacker's Guide to Compiling C Programs on macOS

2025-06-14

This post details the author's journey compiling C/C++ programs on macOS, highlighting the challenges and workarounds encountered. From installing compilers and managing dependencies (using tools like `apt-get` and `brew`), to navigating `Makefile`s and `./configure` scripts, the author provides a practical guide. Key takeaways include handling compiler warnings, resolving linker errors using environment variables like `CPPFLAGS` and `LDLIBS`, and selectively compiling targets with `make`. The author also emphasizes the importance of understanding C compilation, even for non-C programmers, due to its prevalence in system tools and libraries. The post concludes with the author's successful contribution of a compiled package to Homebrew, showcasing the practical benefits of mastering the compilation process.

Read more
(jvns.ca)
Development

Amsterdam's Fair Fraud Detection Model: A Case Study in Algorithmic Bias

2025-06-14

Amsterdam attempted to build a 'fair' AI model for fraud detection in its welfare system, aiming to reduce investigations while improving efficiency and avoiding discrimination against vulnerable groups. The initial model showed bias against non-Dutch and non-Western applicants. While reweighting the training data mitigated some bias, real-world deployment revealed new biases in the opposite direction, along with significant performance degradation. The project was ultimately shelved, highlighting the inherent trade-offs between different fairness definitions in AI. Attempts to reduce bias in one group can inadvertently increase it in others, demonstrating the complexities of achieving fairness in algorithmic decision-making.

Read more
(www.lighthousereports.com)
AI AI fairness fraud detection

Decoupling Low-Level Programming from Systems Design: Rethinking "Systems Programming"

2025-06-14

This article explores the evolution of the term "systems programming." The author argues that it conflates two distinct ideas: low-level programming (dealing with machine implementation details) and systems design (creating and managing complex interacting components). From the 1970s improvements on assembly to the rise of scripting languages in the 1990s and the performance advancements of today's languages, the boundaries of systems programming have blurred. The author proposes redefining "systems programming" as "low-level programming," leaving systems design as a separate field. He argues that functional programming principles are valuable in systems design and suggests separating low-level programming and systems design instruction in computer science education to foster cross-pollination of ideas.

Read more
(willcrichton.net)
Development systems design

Chicken Eyeglasses: A Surprisingly Long History

2025-06-14
Chicken Eyeglasses: A Surprisingly Long History

Believe it or not, eyeglasses for chickens were once a common sight in American poultry farms. Developed in the early 20th century to curb feather pecking and cannibalism, these small spectacles, made of materials ranging from celluloid to aluminum, came in both clear and, interestingly, red-tinted versions. The red tint was supposedly to mask the color of blood, though later it was admitted chickens are colorblind. Despite this, millions of chicken eyeglasses were sold, even featured on a popular TV show. Now collector's items, their story reflects the ongoing evolution of animal welfare practices.

Read more
(en.wikipedia.org)
Misc chicken eyeglasses

Rocky Linux 10 Released: Divergence Widens Among RHEL Alternatives

2025-06-14
Rocky Linux 10 Released: Divergence Widens Among RHEL Alternatives

Rocky Linux 10, "Red Quartz," has reached general availability, adding RISC-V support but dropping older Raspberry Pi models. Compared to AlmaLinux 10 and RHEL 10, released earlier this year, subtle differences emerge in both hardware and software. Most notably, RHEL 10 and Rocky Linux 10 require x86-64-v3 CPUs, while AlmaLinux 10 uniquely supports x86-64-v2. Furthermore, RHEL 10's AI assistant, "Lightspeed," is absent from Rocky Linux 10. While functionally similar, Rocky Linux 10 is subtly diverging from its RHEL alternatives in hardware compatibility, AI features, and commercial support, carving its own niche in the market.

Read more
(www.theregister.com)
Development

Apple Paper Exposes Limits of Scaling in Large Language Models

2025-06-14
Apple Paper Exposes Limits of Scaling in Large Language Models

An Apple paper highlighting limitations in the reasoning capabilities of large language models (LLMs) has sparked a heated debate in the AI community. The paper demonstrates that even massive models struggle with seemingly simple reasoning tasks, challenging the prevalent 'scaling solves all' hypothesis for achieving Artificial General Intelligence (AGI). While some attempted rebuttals emerged, none proved compelling. The core issue, the article argues, is LLMs' unreliability in executing complex algorithms due to output length limitations and over-reliance on training data. True AGI, the author suggests, requires superior models and a hybrid approach combining neural networks with symbolic algorithms. The paper's significance lies in its prompting a critical reassessment of AGI's development path, revealing that scaling alone is insufficient.

Read more
(garymarcus.substack.com)
AI

AI Bots Are Replacing Google Search: A New Kind of Web Traffic

2025-06-14
AI Bots Are Replacing Google Search: A New Kind of Web Traffic

AI tools like ChatGPT are replacing Google search, creating a massive shift in web traffic. Companies like OpenAI and Anthropic are deploying bots to summarize web content in real-time, feeding far more data into AI models than any human could. Data from TollBit reveals a 49% surge in retrieval bot traffic in Q1 2025 compared to Q4 2024, an exponential growth. While AI tools decrease human website traffic, overall content access explodes. Websites must adapt to AI visitors, but negotiating payment for content scraped by AI companies will be challenging, amidst ongoing copyright debates. The modern internet, optimized for Google and social media, needs restructuring to accommodate the rise of AI answers.

Read more
(www.yahoo.com)
Tech Web Traffic

AI + SQL: The Future of Information Retrieval

2025-06-14
AI + SQL: The Future of Information Retrieval

This article proposes a revolutionary approach to information retrieval by leveraging the synergy between AI and advanced SQL systems. Large Language Models (LLMs) are used to interpret human intent, translating natural language queries into precise SQL queries to access massive, distributed object-relational databases. This overcomes the limitations of LLMs relying solely on pattern learning, enabling the handling of diverse data types (geographic, image, video, etc.) and ensuring speed and reliability through distributed systems. The ultimate goal is to empower users to access complex databases using natural language without needing SQL expertise.

Read more
(thenewstack.io)
AI

Debunking HDR: The Hype vs. Reality

2025-06-14
Debunking HDR: The Hype vs. Reality

This article dives deep into the technical details and artistic implications of HDR (High Dynamic Range) imaging, arguing that it's not always superior to SDR (Standard Dynamic Range). It debunks common HDR misconceptions, such as increased efficiency, wider gamut, and improved artistic expression. The author highlights flaws in HDR conversion processes, arguing that forced automated conversions often damage the artist's intent and degrade image quality. The article emphasizes the relative nature of tonal perception and prioritizes artistic expression over technical specifications. Ultimately, it advocates for making informed choices between HDR and SDR, stressing the importance of the creator's vision.

Read more
(yedlin.net)
Design

Dia: The AI-Powered Browser Challenging the Status Quo

2025-06-14
Dia: The AI-Powered Browser Challenging the Status Quo

Traditional web browsers face a challenge from AI. The Browser Company's Dia browser integrates AI deeply, aiming to redefine how users interact with the internet. Built on Chromium, Dia boasts a clean interface and features an AI chatbot that searches the web, summarizes files, generates content based on open tabs, and leverages browsing history for context. Its 'Skills' feature allows users to create code snippets as shortcuts, such as automatically generating reading layouts. While browser AI isn't new, Dia's ease of use and powerful features position it as a strong contender against established browsers.

Read more
(techcrunch.com)
Tech

libc-less Programming: Mastering Linux Syscalls with strace

2025-06-14

The author recently embarked on building software without libc to gain a deeper understanding of Linux syscalls and internals. This involved creating a minimal shell, a Snake game, a pure ARM64 assembly HTTP server, and a threads implementation. Debugging heavily relied on strace, and the article details numerous useful strace options and flags. These range from tracing child processes and printing verbose struct information to selectively tracing syscalls and even injecting syscall errors for debugging purposes. This provides valuable insights into advanced Linux system programming and debugging techniques.

Read more
(rrampage.github.io)
Development Syscalls

Argparse's Mutually Exclusive Group Nesting Limitation: A Frustrating Conundrum

2025-06-14

Python's argparse module, while offering convenient features for handling command-line arguments, including mutually exclusive groups, has a frustrating limitation when it comes to nesting. Consider a program with multiple timeout settings where users can either adjust individual timeouts or disable them entirely. Argparse doesn't support nesting a 'no-timeout' option within a group of individual timeout options, making configuration cumbersome. While you can nest a mutually exclusive group inside a regular group, the reverse isn't supported, and the official documentation explicitly states this limitation. This forces developers to manually check if specific switches were used, adding complexity.

Read more
(utcc.utoronto.ca)
Development

Reverse Engineering the Commodore 64 Freezer Cartridge: A Deep Dive

2025-06-14

This article delves into the reverse engineering of Commodore 64 freezer cartridges, such as the Final Cartridge III. These cartridges leverage the C64's Ultimax mode and NMI interrupts to achieve functionalities like freezing programs, applying cheat codes, and saving game states. The article meticulously explains the technical challenges of the freezing process, such as coordinating 6502 CPU instruction cycles with Ultimax mode activation, and how limited memory resources are utilized for displaying menus and managing state backups. The author also analyzes the cartridge's backup mechanisms and game trainer functionality, praising the developers' deep understanding of the C64 hardware and their masterful coding skills.

Read more
(www.pagetable.com)
Hardware Freezer Cartridge

Cheap Drones Reshape Global Power Dynamics

2025-06-14
Cheap Drones Reshape Global Power Dynamics

Recent successful drone strikes by Israel and Ukraine against enemy targets have sparked concerns about the future of warfare and the global balance of power. The article analyzes the disruptive impact of inexpensive drones: their ability to inflict significant economic damage by targeting critical infrastructure (highways, bridges, power plants, etc.) at low cost represents a stark contrast to traditional large-scale military engagements. This asymmetric warfare capability renders even large nations vulnerable and could lead to a reshaping of the global order, ultimately moving toward a more decentralized and fragmented power structure.

Read more
(blog.exitgroup.us)
Tech drone warfare global politics power balance

Waymo's Meteoric Rise: Could it Topple Uber in San Francisco?

2025-06-14
Waymo's Meteoric Rise: Could it Topple Uber in San Francisco?

Waymo's autonomous ride-hailing service is experiencing explosive growth in San Francisco. Data from YipitData reveals Waymo has surpassed Lyft in market share and is on track to become the city's largest ride-hailing service by the end of the year, potentially even overtaking Uber. This rapid ascent raises concerns about the displacement of tens of thousands of rideshare drivers in San Francisco, highlighting the disruptive potential of AI in the gig economy.

Read more
(underscoresf.com)
Tech ride-hailing

Beyond Silicon: Penn State Creates CMOS Computer from 2D Materials

2025-06-14
Beyond Silicon: Penn State Creates CMOS Computer from 2D Materials

Researchers at Penn State have achieved a breakthrough, creating a functional CMOS computer using two-dimensional (2D) materials—just one atom thick—instead of silicon. Published in Nature, this development represents a significant leap towards thinner, faster, and more energy-efficient electronics. Using molybdenum disulfide and tungsten diselenide for n-type and p-type transistors respectively, the team overcame previous challenges in scaling 2D materials to complex computer architectures. While the operating frequency is currently low, this research paves the way for moving beyond silicon-based technology, promising a new era in electronics.

Read more
(www.sciencedaily.com)
Tech CMOS computer

Europe's First Exascale Supercomputer, Jupiter: An Nvidia-Powered Beast with Arm Aspirations

2025-06-14
Europe's First Exascale Supercomputer, Jupiter: An Nvidia-Powered Beast with Arm Aspirations

The Forschungszentrum Jülich's long-awaited exascale supercomputer, Jupiter, has finally debuted on the Top500 list. This hybrid CPU-GPU machine, built by Eviden and ParTec, boasts a GPU booster module that ranked fourth in the June HPL benchmark. While heavily reliant on Nvidia GPUs and interconnect technology, Jupiter incorporates a Universal Cluster module based on SiPearl's Rhea1 Arm CPU, signifying a move towards European HPC independence. However, reaching the full exascale FP64 performance goal requires further expansion of GPU nodes. The €500 million project highlights the substantial investment in high-performance computing, with a significant portion allocated to hardware and software.

Read more
(www.nextplatform.com)
Tech

Apollo's '8-Ball': Dissecting the Lunar Module's Flight Director/Attitude Indicator

2025-06-14
Apollo's '8-Ball': Dissecting the Lunar Module's Flight Director/Attitude Indicator

This article delves into the Apollo lunar missions' Flight Director/Attitude Indicator (FDAI), a unique instrument featuring a rotating black ball nicknamed the '8-ball'. It meticulously explains the ingenious mechanism allowing the '8-ball' to rotate around three axes (roll, pitch, yaw), and the complex servo-control system within the FDAI, including synchros, servo loops, motor/tachometers, and amplifiers. The author traces the FDAI's history from its use in the X-15 rocket plane and F-4 fighter to its role in the Apollo lunar module and Space Shuttle simulator, highlighting its significance in aerospace history. Comparisons are drawn between the Apollo FDAI and the F-4's ARU/11-A indicator, revealing similarities and differences.

Read more
(www.righto.com)
Tech instrumentation

Lisp Truth Oracle: A Curious Tale of Type Theory, Curry-Howard Isomorphism, and call/cc

2025-06-14

This post attempts to write a "truth oracle" in Lisp—a program that determines the truth or falsehood of arbitrary mathematical statements. The author introduces the Curry-Howard isomorphism, explaining how logical proofs correspond to expressions in typed functional programming. Using Racket's call/cc function (isomorphic to Peirce's law), an attempt is made to implement a program isomorphic to the law of the excluded middle. Unexpectedly, the oracle always returns false until attempting to access an impossible type value, revealing the differences between classical and constructive logic, and the non-standard control flow of call/cc. Finally, the author uses a metaphor of a "devil's bargain" to explain this strange behavior, showcasing the time-travel-like mechanism behind call/cc.

Read more
(lambda-cove.net)
Development type theory
1 2 … 165 166 167 169 171 172 173 … 596 597