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

arXivLabs: Experimental Projects with Community Collaborators

2025-09-04
arXivLabs: Experimental Projects with Community Collaborators

arXivLabs is a framework enabling collaborators to develop and share new arXiv features directly on the arXiv website. Individuals and organizations working with arXivLabs embrace our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only partners with those who share them. Have an idea for a project that will benefit the arXiv community? Learn more about arXivLabs.

Read more
(arxiv.org)
Development

Dolby Vision 2: AI-Powered HDR Gets a Major Upgrade

2025-09-04
Dolby Vision 2: AI-Powered HDR Gets a Major Upgrade

Dolby has unveiled Dolby Vision 2, an evolution of its HDR format. Beyond fine-tuning picture settings, Dolby Vision 2 introduces "Content Intelligence," leveraging AI and TV sensors to dynamically adjust brightness, addressing common complaints about overly dark scenes (think *Game of Thrones*' infamous 'Battle of Winterfell'). A new "Authentic Motion" feature aims to optimize motion handling across various viewing environments, though this may prove controversial among purists.

Read more
(arstechnica.com)
Tech Dolby Vision

Garmin's Fenix 8 Pro: A Satellite-Connected Smartwatch Beats Apple to the Punch

2025-09-04
Garmin's Fenix 8 Pro: A Satellite-Connected Smartwatch Beats Apple to the Punch

Just days before Apple's anticipated unveiling of the satellite-capable Apple Watch Ultra, Garmin launched its own satellite-connected smartwatch, the Fenix 8 Pro. Featuring Garmin's inReach technology, the Fenix 8 Pro enables satellite-based location check-ins and text messaging, and also boasts cellular connectivity for calls, voice messages, LiveTrack, and weather forecasts. An SOS emergency feature, a high-brightness microLED display, durable construction, and comprehensive health tracking round out the features. Available in 47mm and 51mm sizes with AMOLED and microLED display options, the Fenix 8 Pro, priced from $1200 and $2000 respectively, launches September 8th, directly challenging Apple's upcoming Apple Watch Ultra 3. Note that Garmin's satellite services come with a subscription fee, unlike Apple's rumored offering.

Read more
(www.macrumors.com)
Hardware

Building Effective AI Agent Evaluation: From E2E Tests to N-1 Evaluations

2025-09-04

This article explores building efficient AI agent evaluation systems. The author stresses that while models constantly improve, evaluation remains crucial. It advocates starting with end-to-end (E2E) evaluations, defining success criteria and outputting simple yes/no results to quickly identify problems, refine prompts, and compare different model performances. Next, "N-1" evaluations, simulating previous user interactions, can directly pinpoint issues, but require maintaining updated "N-1" interactions. Checkpoints within prompts are also suggested to verify LLM adherence to desired conversation patterns. Finally, the author notes that external tools simplify setup, but custom evaluations tailored to the specific use case are still necessary.

Read more
(aunhumano.com)
AI agent evaluation

Dissecting a Minimalist Transformer: Unveiling the Inner Workings of LLMs with 10k Parameters

2025-09-04
Dissecting a Minimalist Transformer: Unveiling the Inner Workings of LLMs with 10k Parameters

This paper presents a radically simplified Transformer model with only ~10,000 parameters, offering a clear window into the inner workings of large language models (LLMs). Using a minimal dataset focused on fruit and taste relationships, the authors achieve surprisingly strong performance. Visualizations reveal how word embeddings and the attention mechanism function. Crucially, the model generalizes beyond memorization, correctly predicting "chili" when prompted with "I like spicy so I like", demonstrating the core principles of LLM operation in a highly accessible manner.

Read more
(rti.github.io)
AI

Depot Seeks First Solutions Engineer: Accelerating Software Builds, Reshaping the Development Process

2025-09-04
Depot Seeks First Solutions Engineer: Accelerating Software Builds, Reshaping the Development Process

Rapidly growing software build platform Depot is seeking its first dedicated Solutions Engineer. This role requires an experienced developer who can help other developers dramatically improve their day-to-day efficiency. The ideal candidate will be a Depot user and comfortable working independently in a fast-paced startup environment, solving customers' most challenging build performance issues. The position involves close collaboration with customer engineering teams, providing technical guidance, analyzing build logs, and conducting technical demos. Candidates need experience with Docker, Kubernetes, and CI/CD pipelines and the ability to clearly explain complex technical concepts.

Read more
(www.ycombinator.com)
Development software builds solutions engineer

FreeBSD Says No to LLM-Generated Code

2025-09-04
FreeBSD Says No to LLM-Generated Code

The FreeBSD Project's latest quarterly report reveals a new policy in the works: a ban on code and documentation generated by large language models (LLMs). This aligns with similar stances taken by NetBSD and Gentoo Linux, reflecting concerns about the reliability and security of AI-generated code. The report also details other FreeBSD projects underway, including improvements to Wi-Fi, graphics, sound, and power management; restructuring the OS into pkg packages; and developing a web-based GUI for virtualization. Despite competition from Linux, FreeBSD remains actively developed, continuously improving its features and user experience.

Read more
(www.theregister.com)
Development

The Limits of Empathy: What Is It Like to Be a Bat?

2025-09-04
The Limits of Empathy: What Is It Like to Be a Bat?

Thomas Nagel's 1974 philosophical paper, "What Is It Like to Be a Bat?", explores the nature of consciousness. He argues that while we can imagine a bat's perspective, we can never truly know "what it is like" to be a bat. The paper challenges reductive materialism, asserting that subjective experience cannot be fully explained by objective physical processes. Nagel's bat analogy has become a classic in consciousness studies, sparking ongoing debates about subjective experience, objective observation, and the mind-body problem.

Read more
(en.wikipedia.org)
Misc mind-body problem

Neovim Package Management: Unveiling the Secrets of `start` and `opt`

2025-09-04

Neovim's package management system allows developers to organize plugins into two types of packages: `start` and `opt`. `start` packages are automatically loaded on Neovim startup, while `opt` packages require manual loading with the `:packadd` command. This article details how to create and use both types of packages, including installation from zip archives, updating using git repositories, and handling dependencies between plugins. Additionally, it introduces Neovim's built-in `vim.pack` plugin manager, which leverages git for plugin management, providing version control and streamlined updates, significantly simplifying the plugin management workflow.

Read more
(neovim.io)
Development

Wordsworth's Open Letter Supporting Copyright Reform

2025-09-04
Wordsworth's Open Letter Supporting Copyright Reform

In 1838, William Wordsworth penned a powerful letter to Serjeant Talfourd, MP, voicing his staunch support for a bill aimed at reforming copyright law. Faced with significant opposition from printers and publishers, Wordsworth declined to petition Parliament, instead choosing to publicly declare his belief in authors' inherent right to perpetual ownership of their works—a right far exceeding the bill's proposed term. He argued this right stemmed from common law and criticized opponents for hiding behind existing statutes, avoiding a defense of this fundamental right. The letter also touches upon his concerns for literary giants like Coleridge, Scott, and Southey, expressing his deep respect and gratitude for Talfourd's efforts.

Read more
(gutenberg.org)
Misc Literary History

Solving Minesweeper with the Boltzmann Distribution

2025-09-04

This article explores using the Boltzmann distribution from statistical mechanics to improve Minesweeper win rates. It uses a challenging Minesweeper scenario to illustrate the limitations of traditional probability calculations. The author proposes a Boltzmann distribution-based improvement, treating the number of remaining mines as 'energy' to calculate the probability of each possibility, thus more accurately judging the likelihood of each cell containing a mine. While the approximation deviates somewhat from the true value, the article highlights the potential of statistical mechanics models in solving seemingly simple combinatorial problems, especially with larger scales where the method's accuracy significantly improves.

Read more
(oscarcunningham.com)
Game Boltzmann Distribution

Notepad++ Gets a Rogue-lite Plugin: Adventure in Your Text Editor

2025-09-04
Notepad++ Gets a Rogue-lite Plugin: Adventure in Your Text Editor

A new Notepad++ plugin brings rogue-lite gameplay to your text editor! This 64-bit Windows-only plugin features six levels of turn-based combat, powerful relic collection, boss battles, and trap avoidance. It includes a storyline and audio, but play at your own risk—data and settings loss is possible. Installation is easy: unzip, install the font, drag and drop the theme and plugin files into their respective Notepad++ folders. Ready for your Notepad++ adventure?

Read more
(github.com)
Game Game Plugin

Six Months on Alpine: The Musl Trade-off

2025-09-04
Six Months on Alpine: The Musl Trade-off

The author spent six months using Alpine Linux as their daily driver. Alpine is praised for its speed, excellent package management, and lightweight nature. However, the author encountered compatibility issues due to Alpine's use of the musl libc instead of glibc, particularly with experimental software requiring glibc. While workarounds like gcompat, self-compilation, or Flatpak exist, they add friction. Ultimately, the author decided to explore other distributions like Void Linux or Debian for better compatibility and stability.

Read more
(blog.jutty.dev)
Development

Zed Integrates Claude Code: AI-Powered Coding in Your Editor

2025-09-04
Zed Integrates Claude Code: AI-Powered Coding in Your Editor

Zed editor has released a public beta integration with Claude Code, running natively through the new Agent Client Protocol (ACP). This represents a significant step forward in AI-powered code assistance. Developers can now run Claude Code directly within Zed's high-performance editor, view real-time multi-file edits, and review code changes granularly. ACP, an open standard, enables any agent to connect to Zed and other editors, making this Claude Code integration possible. The future will see more AI agents integrated into Zed via ACP, offering developers greater choice.

Read more
(zed.dev)
Development

The AI Coding Lie: Productivity Hype or Bust?

2025-09-04
The AI Coding Lie: Productivity Hype or Bust?

A seasoned programmer with 28 years of experience challenges the claims of AI coding tools, revealing a six-week experiment that showed no significant productivity gains, and potentially even a slowdown. The author argues that the industry's hype around AI-driven productivity increases is vastly overblown, unsupported by real-world data. Using extensive data, the article demonstrates the lack of an expected surge in software development output, debunking the myth of the '10x engineer'. The author urges developers to approach AI tools critically, avoid blind adoption, and resist unrealistic marketing claims.

Read more
(mikelovesrobots.substack.com)
Development

Cursor for Xcode Joins OpenAI's Codex Team

2025-09-04

The Cursor for Xcode team, creators of the popular AI coding assistant Alex for iOS and macOS developers, announced they're joining OpenAI's Codex team. While new downloads of the Alex app will cease on October 1st, existing users can continue using it, though no new features will be added. The team will continue their mission of helping people create, now at a larger scale with OpenAI.

Read more
(www.alexcodes.app)
Development

Fluid Simulation Meets Reality: A 2025 Demo

2025-09-04

This demo showcases a fascinating fluid simulation interacting with real-world objects. The creator cleverly uses a webcam to capture object shapes, a polarization filter to prevent screen feedback, and aligns the feed with the simulation for real-time interaction. Hands are surprisingly recognized as obstacles, adding a fun, unexpected element. The simulation itself is a wind-tunnel style model, drawing inspiration from Tidepodious' work. It's a visually impressive blend of virtual and real.

Read more
(danybittel.ch)
Development real-time interaction

Tesco Sues Broadcom Over VMware Licensing Dispute, Threatens Supply Chain

2025-09-04
Tesco Sues Broadcom Over VMware Licensing Dispute, Threatens Supply Chain

Tesco, a UK supermarket giant, is suing Broadcom for breach of contract regarding its VMware licenses, also naming Computacenter as a co-defendant. Broadcom's refusal to provide support services for perpetually licensed VMware software after its acquisition threatens Tesco's operations and could disrupt grocery supply. Tesco argues Broadcom's subscription model is excessively expensive and prevents necessary software updates. The lawsuit highlights the broader issue of perpetual license support after acquisitions and the potential for significant financial damages. Other companies have filed similar lawsuits against Broadcom.

Read more
(www.theregister.com)
Tech Licensing Dispute

Record-Breaking 11.5 Tbps DDoS Attack Successfully Mitigated

2025-09-04
Record-Breaking 11.5 Tbps DDoS Attack Successfully Mitigated

Over the Labor Day weekend, Cloudflare successfully mitigated a record-breaking 11.5 Tbps distributed denial-of-service (DDoS) attack lasting approximately 35 seconds and peaking at over 5.1 billion packets per second. The attack leveraged a simple UDP flood, originating from various IoT devices and cloud providers, including compromised Google Cloud accounts. While simple in nature, the scale and frequency of such attacks are rapidly increasing, with Cloudflare blocking over 6,500 similar attacks in Q2 2025. This highlights the importance of modern internet security defenses and serves as a warning for businesses to implement robust DDoS protection.

Read more
(www.zdnet.com)
Tech

C Compiler in 500 Lines of Python: A Single-Pass Approach

2025-09-04

The author undertook the challenge of writing a C compiler in just 500 lines of Python, targeting WebAssembly. To achieve this, a single-pass compilation strategy was employed, foregoing an Abstract Syntax Tree (AST) and generating code directly during parsing. This resulted in a concise compiler, but at the cost of several features, including floating-point numbers and structs. The resulting compiler successfully compiles and runs simple C programs, demonstrating the feasibility of a single-pass approach.

Read more
(vgel.me)
Development

Century-Old Math Conjecture Overturned: New Knot Theory Discovery

2025-09-04
Century-Old Math Conjecture Overturned: New Knot Theory Discovery

Mathematicians have overturned a long-held conjecture in knot theory. It was believed that connecting two different knots would result in a new knot with complexity equal to the sum of the individual knots' complexities. However, researchers recently found a knot simpler than the sum of its parts. This discovery challenges our understanding of knot complexity and offers new insights into fields like protein folding and molecular stability.

Read more
(www.scientificamerican.com)
Tech knot theory

Leaked: Microsoft's 1978 BASIC 1.1 for 6502!

2025-09-04
Leaked: Microsoft's 1978 BASIC 1.1 for 6502!

The source code for Microsoft BASIC 1.1 for the 6502 microprocessor, dating back to 1978, has surfaced! This 6,955-line assembly language program was foundational to the personal computer revolution, powering early machines like the Apple II and Commodore PET. Its release offers a glimpse into Microsoft's early successes, showcasing its cross-platform compatibility and efficient memory management, leaving an indelible mark on the software industry.

Read more
(github.com)
Tech 6502 Assembly

Instagram Finally Launches Dedicated iPad App

2025-09-04
Instagram Finally Launches Dedicated iPad App

After years of user requests, Instagram has finally released a dedicated app for iPads! Launching September 3rd, the app prioritizes Reels, Instagram's TikTok competitor, reflecting the dominance of short-form video. This move, possibly spurred by TikTok's regulatory challenges, solidifies Instagram's short-video strategy. The iPad app retains Stories and a "Following" tab, offering a more familiar experience, but optimized for the larger screen with features like comments directly beside Reels. Meta confirms similar updates are coming soon to Android tablets.

Read more
(www.theverge.com)
Tech iPad App

PyTorch Model with Metal Acceleration: Performance and Correctness

2025-09-04
PyTorch Model with Metal Acceleration: Performance and Correctness

This article presents a PyTorch-based model that attempts to leverage Metal for accelerated computation while providing pure PyTorch fallbacks to guarantee correctness. The model's core involves complex calculations including matrix multiplications, cumulative sums, and exponentiation. To enhance performance, the authors attempt to use Metal for custom kernels, but fall back to a pure PyTorch implementation if the Metal extension isn't available. This design ensures compatibility and reliability across different hardware platforms, offering developers a solution that balances performance and correctness.

Read more
(gimletlabs.ai)
Development Model Acceleration

Random Walks in 10 Dimensions: Defying Intuition in High-Dimensional Spaces

2025-09-04
Random Walks in 10 Dimensions: Defying Intuition in High-Dimensional Spaces

High-dimensional physics is the norm in modern dynamics, from string theory's ten dimensions to complex systems. However, high dimensions present the 'curse of dimensionality': visualization is impossible, overfitting is rampant, and intuition fails. This article uses a 10-dimensional random walk to illustrate high-dimensional space characteristics. In high dimensions, mountain ridges are far more common than peaks, profoundly impacting evolution, complex systems, and machine learning. Random walks efficiently explore high-dimensional spaces, even maximally rough landscapes, potentially traversing the entire space. This helps understand the evolution of complex structures in life and how to avoid local minima in deep learning.

Read more
(galileo-unbound.blog)
AI high-dimensional physics curse of dimensionality random walk

TPDE-LLVM: A 10-20x Faster LLVM Back-end

2025-09-03
TPDE-LLVM: A 10-20x Faster LLVM Back-end

The TPDE project has open-sourced its fast LLVM back-end, TPDE-LLVM, achieving a 10-20x speedup in compilation compared to the LLVM -O0 back-end on SPEC CPU 2017 benchmarks, with similar runtime performance and a 10-30% increase in code size. TPDE-LLVM currently supports x86-64 and AArch64 architectures and incorporates optimizations to LLVM-IR, such as removing constant expressions inside functions and limiting struct/array sizes. Future plans include broader LLVM-IR feature support, DWARF debug info, and improved register allocation.

Read more
(discourse.llvm.org)
Development

Server CPU Utilization: Don't Be Fooled by the Numbers!

2025-09-03
Server CPU Utilization: Don't Be Fooled by the Numbers!

Do you rely on server CPU utilization to assess server performance? This author ran a series of stress tests and discovered that CPU utilization isn't linearly correlated with actual work efficiency. Especially above 50% utilization, hyperthreading and Turbo Boost significantly skew the relationship, leading to much higher actual throughput than the reported utilization suggests. Instead of relying on CPU utilization, the author recommends benchmarking to measure actual work completed for a more accurate assessment of server performance.

Read more
(www.brendanlong.com)
Development server performance CPU utilization hyperthreading

Data, Not Compute: The Next AI Bottleneck

2025-09-03
Data, Not Compute: The Next AI Bottleneck

For years, we've misinterpreted the Bitter Lesson; it's not about compute, but data. Increasing GPUs requires a 40% data increase, otherwise it's wasted resources. The internet's data is nearing saturation. The future lies in 'alchemists' (high-risk, high-reward data generation) and 'architects' (steadily improving model architecture), not just compute. The article analyzes the pros, cons, and risks of both paths, concluding that solving data scarcity in 2025 will determine AI company survival in 2026.

Read more
(obviouslywrong.substack.com)
AI data bottleneck model architecture

Tencent's HunyuanWorld-Voyager: World-Consistent 3D Video Generation from a Single Image

2025-09-03
Tencent's HunyuanWorld-Voyager: World-Consistent 3D Video Generation from a Single Image

Tencent's AI team introduces HunyuanWorld-Voyager, a novel video diffusion framework generating world-consistent 3D point cloud sequences from a single image with user-defined camera paths. Voyager produces 3D-consistent scene videos for exploring virtual worlds along custom trajectories, also generating aligned depth and RGB video for efficient 3D reconstruction. Trained on over 100,000 video clips combining real-world and Unreal Engine synthetic data, Voyager achieves state-of-the-art results on the WorldScore benchmark. Code and pre-trained models are publicly available.

Read more
(github.com)
AI 3D video generation video diffusion
2 4 5 6 7 8 9 … 560 561