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

A5: A Global, Millimeter-Accurate Geospatial Index

2025-05-13
A5: A Global, Millimeter-Accurate Geospatial Index

A5 is a geospatial index partitioning the world into pentagonal cells at 32 resolution levels, with the smallest cell under 30mm² and near-equal area across levels. It simplifies spatial data representation and analysis, enabling calculations of correlations between variables (e.g., elevation and crop yield) and aggregation of point data to understand spatial distribution (e.g., holiday rental density). Implemented in TypeScript, A5 is open-source under the Apache 2.0 license. Compared to other DGGS systems, A5 boasts uniform cell sizes, extremely high resolution (30mm²), and minimal global cell area distortion. This stems from its unique pentagonal tiling of a dodecahedron, minimizing geometric distortion during projection.

Read more
(github.com)
Tech Geospatial Index

German Startup Unveils Continuously Operating Fusion Power Plant Design

2025-02-27
German Startup Unveils Continuously Operating Fusion Power Plant Design

Two-year-old German nuclear fusion startup Proxima Fusion published its "Stellaris" fusion power plant design in a peer-reviewed journal. This stellarator-based design aims for continuous, reliable operation, addressing instability issues plaguing tokamak approaches. Proxima Fusion, backed by €65 million in funding, plans to build a fully operational reactor by 2031. This breakthrough marks a significant step forward in the race for clean energy.

Read more
(techcrunch.com)
Tech nuclear fusion stellarator

LHC Ultraperipheral Collisions Unravel the Mystery of Gluon Saturation

2025-02-11
LHC Ultraperipheral Collisions Unravel the Mystery of Gluon Saturation

Ultraperipheral collisions (UPCs) at the Large Hadron Collider (LHC) offer a unique window into gluon dynamics. Experiments use photons to probe gluons within protons and nuclei, investigating gluon saturation—a dynamic equilibrium between gluon splitting and recombination. Researchers found that as energy decreases, the number of gluons in hadrons increases, forming 'gluonic hotspots' that overlap in the gluon saturation regime. The experiments also observed nuclear shadowing, where nuclei contain fewer gluons than expected. These findings shed light on the origin of 99% of the visible universe's mass and the nature of the strong interaction. Future LHC runs and the Electron-Ion Collider (EIC) will further explore gluon dynamics, uncovering more mysteries.

Read more
(cerncourier.com)
Tech gluon saturation hadron physics

OP-1 Field Price: A Bold Experiment in Crowdsourced Pricing

2025-06-04
OP-1 Field Price: A Bold Experiment in Crowdsourced Pricing

Teenage Engineering is conducting a bold experiment: letting users decide the price of their OP-1 Field synthesizer. They aim to better understand the market and explore how to truly satisfy customer needs. This reflects the current uncertain world, encouraging trying new things and engaging customers in pricing in a novel way. The final price will be revealed later this year, or until the world is a bit more stable.

Read more
(teenage.engineering)
Hardware product pricing

Meta Prompting: Revolutionizing LLM Prompt Engineering

2025-06-03
Meta Prompting: Revolutionizing LLM Prompt Engineering

This article explores meta prompting, a technique using Large Language Models (LLMs) to create and refine prompts. It details various meta-prompting methods, including the Stanford and OpenAI collaboration's method using a 'conductor' LLM to orchestrate expert LLMs; Amazon's Learning from Contrastive Prompts (LCP), which improves prompts by comparing good and bad ones; Automatic Prompt Engineer (APE), Prompt Agent, Conversational Prompt Engineering (CPE), DSPy, and TEXTGRAD. The article compares their strengths and weaknesses, highlighting how these methods significantly improve prompt engineering efficiency. Finally, it showcases prompt generation tools from platforms like PromptHub, Anthropic, and OpenAI, simplifying meta-prompting implementation and unlocking the full potential of LLMs.

Read more
(www.prompthub.us)
Development meta prompting

Zed Code Editor Shifts AI Pricing to Token-Based Model

2025-09-24
Zed Code Editor Shifts AI Pricing to Token-Based Model

Code editor Zed is changing its AI pricing from prompt-based to token-based, better reflecting the actual cost of AI services. This allows Zed to sustainably invest in editor features and enterprise offerings. The new pricing includes more AI models like GPT-5 and Gemini 2.5, and offers users flexibility with options like bringing their own API keys or using local models. Existing users have a three-month transition period with support. The change also simplifies pricing and reduces costs.

Read more
(zed.dev)
Development

The Modern Guide to OAuth 2.0: Beyond the Specs

2025-06-09
The Modern Guide to OAuth 2.0: Beyond the Specs

This isn't just another OAuth 2.0 guide; it's a deep dive into real-world OAuth usage based on the experience of building FusionAuth, an OAuth server with over a million downloads. The guide details eight common OAuth modes, including local login, third-party login, enterprise login, service authorization, and machine-to-machine authentication, explaining each mode's workflow and security considerations. It also delves into the authorization code grant, PKCE, JWTs, token refresh, and user info retrieval, offering practical implementation advice.

Read more
(fusionauth.io)
Development

Russia Automates Disinformation to Game AI Chatbots

2025-04-19
Russia Automates Disinformation to Game AI Chatbots

Russia is automating the spread of disinformation to manipulate AI chatbots, influencing responses on key topics like the war in Ukraine. Researchers found that leading chatbots repeated Russian lies, highlighting a vulnerability in AI's reliance on data. Russia created a network of websites (Pravda network) designed to be picked up by AI crawlers, saturating the internet with false narratives. This low-cost, highly effective tactic undermines information integrity, exacerbated by reduced government oversight and the rapid deployment of chatbots. The lack of effective response mechanisms poses a significant threat.

Read more
(www.detroitnews.com)
Tech

Tesla's Next-Gen Vehicle Network and 4680 Battery: A Technological Leap

2025-04-21
Tesla's Next-Gen Vehicle Network and 4680 Battery: A Technological Leap

Tesla is undergoing a significant vehicle architecture upgrade. They're replacing the legacy CAN bus with a next-generation network based on TDMA, enabling more efficient data transfer for high-resolution infotainment, OTA updates, and autonomous driving. Simultaneously, Tesla's 4680 battery, particularly its second-generation "Cybercell," is improving production efficiency, lowering costs, and enhancing vehicle performance. However, the launch of a cheaper Model Y has been pushed back to Q3 2025 or early 2026, suggesting Tesla is prioritizing its technological advancements and production optimization.

Read more
(www.notateslaapp.com)
Tech Vehicle Network 4680 Battery

The End of SEO? Generative Engine Optimization (GEO) is Here

2025-06-01
The End of SEO? Generative Engine Optimization (GEO) is Here

The rise of Large Language Model (LLM) platforms is signaling the end of traditional Search Engine Optimization (SEO). A new paradigm, Generative Engine Optimization (GEO), is emerging, focusing not on page rank but on how frequently content appears in LLM-generated answers. Brands need to optimize content for easy parsing and citation by LLMs, using new GEO tools to monitor brand presence in AI-generated results, gaining a competitive edge. GEO is not just a tooling shift, but a platform opportunity. The company that integrates insights, creative input, feedback, and iteration, and masters first- and third-party data will dominate.

Read more
(a16z.com)
Tech Generative Engine Optimization AI Marketing

Open-Source Tool LVTShift: Model Your City's Land Value Tax

2025-06-05
Open-Source Tool LVTShift: Model Your City's Land Value Tax

This blog post details using the open-source tool LVTShift to model the impacts of a land value tax (LVT). The author showcases analyses of South Bend and Syracuse, demonstrating how LVTShift simulates various LVT policies (revenue-neutral, different tax burden shifts, etc.) and their effects on city residents and economies. The post thoroughly explains data acquisition, processing, model building, and analysis, including code examples and data sources. Readers are encouraged to model their city's LVT using LVTShift and share their results.

Read more
(progressandpoverty.substack.com)
Development

Over 3.1 Million Fake GitHub Stars Used to Promote Malware

2024-12-31
Over 3.1 Million Fake GitHub Stars Used to Promote Malware

A recent study revealed over 3.1 million fake "stars" on GitHub, used to artificially inflate the popularity of scam and malware repositories. Researchers used a tool called StarScout to analyze massive datasets, identifying 278,000 accounts responsible for these fake stars across 15,835 repositories. This deceptive practice, particularly rampant in 2024, allows malicious projects to appear legitimate and reach unsuspecting users. While GitHub has removed many of the implicated accounts and repositories, the problem persists. Users are urged to carefully evaluate project quality and exercise caution when downloading software from GitHub.

Read more
(www.bleepingcomputer.com)
Development Fake Stars

Migrating from CockroachDB to PostgreSQL: $110k in Yearly Savings

2025-05-14
Migrating from CockroachDB to PostgreSQL: $110k in Yearly Savings

A company struggled with high latency issues in CockroachDB, with complex SQL queries leading to performance degradation and difficult query cancellation. They ultimately migrated to PostgreSQL. The migration process spanned several weeks, involving building a custom ETL tool and data transformation, but ultimately completed the production database migration in 15 minutes, resulting in a 33% reduction in request latency and over $110,000 in annual savings.

Read more
(engineering.usemotion.com)
Development

Blend2D's Blazing-Fast PNG Codec: Outperforming C/C++

2025-03-26
Blend2D's Blazing-Fast PNG Codec: Outperforming C/C++

Blend2D library introduces a new high-performance PNG codec that significantly outpaces other C/C++ implementations. Optimized for the DEFLATE algorithm's inherent limitations, this decoder achieves speed improvements through fast decode table construction, optimized decoding loops, and clever use of literal pair techniques. Benchmarks demonstrate superior performance in PNG image decoding, even surpassing the speed of some QOI decoders in certain cases. The project is fully open-source and welcomes contributions.

Read more
(blend2d.com)
Development Codec

Technicolor Files for Bankruptcy Protection, Thousands of VFX Jobs at Risk

2025-02-27
Technicolor Files for Bankruptcy Protection, Thousands of VFX Jobs at Risk

Visual effects giant Technicolor, home to MPC, The Mill, and other prominent brands, has filed for bankruptcy protection in France after failing to secure new investors. The move threatens thousands of VFX jobs across the US, UK, Canada, and India. Some employees are already finding new opportunities; for example, a portion of The Mill's US team is launching a new venture, Arc Creative, with Dream Machine FX. Technicolor's struggles highlight challenges in the post-production industry, raising concerns within the VFX community and jeopardizing numerous ongoing film projects.

Read more
(variety.com)
Tech Visual Effects

Sleep: More Than Brain Rest, a Full-Body Reset

2025-04-10
Sleep: More Than Brain Rest, a Full-Body Reset

Harvard Medical School research challenges our understanding of sleep. While long considered mere brain rest, sleep deprivation experiments showed that fruit flies and mice died within ten days due to a buildup of reactive oxygen species in their guts. Sleep resets the brain, clearing neurotoxins, consolidating memories, and regulating genes, metabolism, and hormones for bodily homeostasis. Even organisms without central nervous systems, like hydra, exhibit sleep-like behavior, indicating that sleep's function transcends the brain and is crucial for whole-body health, reducing cardiovascular disease and cancer risk.

Read more
(www.nature.com)
Tech

Instacar Bypasses Apple's Payment System, Sparking Controversy

2025-05-15
Instacar Bypasses Apple's Payment System, Sparking Controversy

The app Instacar, unavailable in the US App Store, displays a warning in the EU App Store for using an external payment system, bypassing Apple's private and secure payment system. The article argues that Apple's understanding of online payments is naive, clinging to the outdated notion that its system's security is a unique advantage. This ignores the fact that mainstream online payment platforms like Stripe and Amazon already offer secure and private transactions.

Read more
(daringfireball.net)
Tech Payment System

Haskell Concurrency: Escape from Thread Hell

2025-04-17

This article recounts the author's journey from embedded systems development in C/C++/Rust to Haskell, highlighting Haskell's advantages in concurrent programming. Haskell uses green threads and event-driven IO, avoiding the complexities of traditional threading models. Through the `async` package and STM (Software Transactional Memory), Haskell offers a cleaner and safer approach to concurrent tasks. Functions like `concurrently`, `race`, and `mapConcurrently`, along with data structures such as `TVar` and `TBQueue`, simplify concurrent operations and prevent common concurrency issues like deadlocks and race conditions.

Read more
(bitbashing.io)
Development

Conquering JavaScript Fatigue: MESH, a Modular SSR Framework Built on HTMX

2025-09-23

Web development is facing "JavaScript fatigue" and "framework fatigue." This post explores using HTMX, a declarative approach to web development using HTML attributes, as a solution. However, HTMX's lack of structure led the author to create MESH, a modular server-side rendering (SSR) framework. MESH uses a "one component, one endpoint" model, leveraging Go and Web Components for SSR and hydration. Challenges with HTMX's inability to cross shadow DOM boundaries were overcome with clever workarounds. Real-time collaboration with Server-Sent Events (SSE) was also implemented. Ultimately, the author even removed HTMX entirely, using cleaner JS to achieve the same functionality, and reflects on the shortcomings and future directions of HTMX.

Read more
(ajmoon.com)
Development

Lightweight GRPO Training: No Transformers, No vLLM

2025-04-13
Lightweight GRPO Training: No Transformers, No vLLM

This project implements a lightweight GRPO (Group Relative Policy Optimization) training framework, built almost from scratch, relying only on tokenizers and PyTorch. It improves upon the original algorithm by removing KL divergence and incorporating overlong episode filtering, enhancing training stability and GPU memory usage. The project trains the Qwen2.5-3B-Instruct model on the CountDown task, which requires generating a mathematical expression to reach a target value given a set of numbers. The model solves this by learning to generate chain-of-thought reasoning before the final answer, guided by format and answer rewards. The entire process is straightforward and reproducible, running on a single A40 GPU with minimal commands.

Read more
(github.com)
Development

Zig's Comptime: Powerful Yet Restrained Metaprogramming

2025-04-20

Zig's comptime feature is renowned for its capabilities: generics, conditional compilation, and more. However, it's deliberately restrictive, disallowing dynamic code generation, custom syntax extensions, runtime type information (RTTI), and I/O. This article explores the reasoning behind these limitations, showcasing how Zig achieves efficient and understandable metaprogramming through partial evaluation and type specialization. A custom printing function example demonstrates how Zig performs type-safe runtime reflection without RTTI. The article concludes by praising Zig's unique elegance in metaprogramming; while less powerful than alternatives, it's remarkably efficient and easy to use in practice.

Read more
(matklad.github.io)
Development comptime metaprogramming

Max Planck Society: Elite Science, Toxic Culture?

2025-03-16
Max Planck Society: Elite Science, Toxic Culture?

The Max Planck Society, a renowned German research institution boasting 31 Nobel laureates among its 84 institutes, faces allegations of misconduct. A joint investigation by DW and Der Spiegel reveals accounts from over 30 young scientists detailing abusive behavior and toxic work environments within the prestigious institutes. Fear of reprisal silenced many, while others who reported misconduct claim they were discouraged. The investigation delves into why these issues persist despite opposition.

Read more
(www.dw.com)
Tech scientific misconduct toxic workplace Max Planck Society

IBM z17: A Deep Dive into the Next-Gen Mainframe

2025-04-24
IBM z17: A Deep Dive into the Next-Gen Mainframe

IBM's April 2025 announcement of the z17 mainframe unveiled a powerhouse featuring the new Telum II processor. This boasts a 4x AI acceleration boost over its predecessor and includes a low-latency DPU for enhanced I/O. Paired with the IBM Spyre accelerator, a cutting-edge ASIC designed for AI workloads with 32 cores and 25.6 billion transistors, the z17 offers up to 64TB of memory in a 4-frame configuration. This represents a significant leap forward in mainframe technology.

Read more
(community.ibm.com)
Hardware Telum II Processor Spyre Accelerator

An Eight-Year-Old Polyomino Tiling Algorithm: Backtracking Search with Heuristics

2025-03-15

This article details an algorithm for solving the polyomino tiling problem. The core idea is to transform the geometric problem into a graph theory problem and use a backtracking search algorithm with various heuristics. First, the algorithm preprocesses to calculate all possible placements, constructing a bipartite graph representing all possibilities. Then, a backtracking search algorithm finds a subset of placements satisfying the conditions, optimized by heuristics such as prioritizing constrained grid points and splitting the grid. The algorithm demonstrates good generality and robustness in handling arbitrary grid shapes and polyomino sets. The author also discusses limitations and future improvements, such as transforming the problem into a SAT problem for solution.

Read more
(gfredericks.com)
Development backtracking search

Waypoint: Automating Urban Planning with AI – Hiring First Engineer

2025-05-01
Waypoint: Automating Urban Planning with AI – Hiring First Engineer

Waypoint is revolutionizing urban planning through AI automation, tackling the inefficiencies and high costs associated with traditional consulting firms. They're seeking their first engineer to build their engineering systems from the ground up. Projects include fine-tuning YOLO models for sidewalk segmentation, developing a system for processing city planning documents, and automating the generation of intersection safety recommendations. The ideal candidate is a strong programmer, a quick learner, a problem-solver, and passionate about improving urban planning.

Read more
(www.workatastartup.com)
AI

From Arc to Dia: The Browser Company's AI Pivot

2025-05-27
From Arc to Dia: The Browser Company's AI Pivot

The Browser Company announces its pivot from its browser project, Arc, to a new AI-powered browser, Dia. The article explains that despite Arc's loyal user base, it fell short of expectations due to growth limitations, hesitation in embracing AI, and communication issues. The author admits Dia isn't a simple iteration of Arc but rather a response to the future of browsers—AI will fundamentally reshape how we interact with computers, and traditional browsers will be replaced by AI browsers. Dia aims to address Arc's shortcomings, offering a simpler, faster, and more secure experience. The article also touches on Arc's future, stating it won't be open-sourced immediately but will continue to be maintained.

Read more
(browsercompany.substack.com)
Development Pivot

Zig 0.14.0 Release Incoming: Improved x86 Backend and Incremental Compilation

2025-01-17

After a period of quiet development, the Zig Software Foundation is gearing up for the release of Zig 0.14.0. This release focuses on improvements to the x86 backend (potentially becoming the default for debug builds) and incremental compilation (disabled by default, but enabled via a compiler flag). Other improvements include labeled switch continue and upgraded support for nearly all target platforms. The team encourages users to upgrade to the latest master branch and will focus on ensuring a smooth upgrade during the release month. While a 1.0 release is still some time away, the team is committed to delivering stable, incremental releases.

Read more
(ziglang.org)
Development

Norwegian Startup's Airhull Tech Lets Electric Boats Glide on Air

2025-05-26
Norwegian Startup's Airhull Tech Lets Electric Boats Glide on Air

Pascal Technologies, a Norwegian electric boat startup, is equipping two of its boats, the Nabcrew Zero AirBlue 1240 and Hugin DC, with Airhull technology. This innovative technology creates an air cushion under the hull, reducing drag and significantly increasing efficiency, potentially saving up to 50% of energy consumption. Simpler to implement than hydrofoil technology, Airhull uses a comb-like structure on the hull's underside and a blower at the bow to lift the boat 15-20cm out of the water. Suitable for boats from 6m to 30m, the technology is showcased on a 12m workboat (Nabcrew Zero AirBlue 1240) and a 9.15m leisure boat (Hugin DC), both slated for launch later this year.

Read more
(www.heise.de)
Tech electric boats

OpenSearch 3.0: 9.5x Faster Search and Analytics for AI

2025-05-07
OpenSearch 3.0: 9.5x Faster Search and Analytics for AI

The OpenSearch Software Foundation announced the general availability of OpenSearch 3.0, boasting a 9.5x performance improvement over version 1.3. This release tackles the challenges of scaling vector databases for AI applications like generative AI and recommendation engines. Key features include GPU acceleration (reducing costs by up to 3.75x), enhanced data management (gRPC support, pull-based ingestion), and improved vector search capabilities. Core upgrades, such as Lucene 10 and Java 21 support, ensure future-proofing and enhanced performance. OpenSearch 3.0 empowers developers to build more efficient and scalable AI applications.

Read more
(opensearch.org)
Tech
1 2 … 33 34 35 37 39 40 41 … 596 597