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

350-Million-Year-Old Reptile Footprints Rewrite Evolutionary History

2025-05-22
350-Million-Year-Old Reptile Footprints Rewrite Evolutionary History

Scientists in Australia have unearthed the oldest known reptile footprints, dating back approximately 350 million years. This discovery predates the previously oldest known footprints by 32 million years, suggesting that the transition of vertebrates from ocean to land happened much faster than previously thought. The footprints, exhibiting clawed feet, confirm the animal's complete terrestrial adaptation, as claws only evolved in fully land-dwelling creatures. Estimated to be about 2 1/2 feet long, the reptile likely resembled a modern monitor lizard. This remarkable find significantly alters our understanding of early vertebrate evolution and provides crucial insights into the history of life on Earth.

Read more
(apnews.com)
Tech reptiles

Perl Advent Calendar 2024: Randal Schwartz's 'Half My Life with Perl'

2024-12-19

The Perl Advent Calendar 2024 features a unique video presentation by Randal Schwartz, titled 'Half My Life with Perl'. Randal, a Perl veteran, recounts his journey with the language from its early days to the modern era, including his involvement in creating the Camel and Llama books, and his humorous anecdotes of conquering the comp.unix.questions forum with Perl 2. This marks the first time a video has been included in the Perl Advent Calendar, offering a fresh and engaging perspective on the language's history.

Read more
(perladvent.org)
Development programming history

Waymo's 10 Million Rides: Tesla's Autopilot Strategy Under Pressure?

2025-06-04
Waymo's 10 Million Rides: Tesla's Autopilot Strategy Under Pressure?

In 2019, Elon Musk dismissed lidar and Waymo. Fast forward to 2024, and Waymo's driverless taxi service has surpassed 10 million rides, doubling its trips in just months. Conversely, Tesla's robotaxi service is launching with a mere 10 vehicles. The author argues Waymo's focus on densely populated urban areas, leveraging lidar and other technologies, has yielded significant progress. Tesla's approach may be too aggressive, overlooking the 80/20 rule of city driving—solving the last 20% of self-driving might require 80% of the effort. Waymo's success suggests a steady, controlled market approach might be more effective than striving for all-scenario coverage in the autonomous driving field.

Read more
(deepdream.substack.com)
Tech

Figma: Collaboration Tool or Creativity Killer?

2025-07-07
Figma: Collaboration Tool or Creativity Killer?

This article reflects on Figma's evolution over a decade, shifting from initial praise for its innovation to concerns about its over-engineered approach. The author argues that features like Auto Layout and Dev Mode, while boosting efficiency, stifle designers' freedom and creativity in the early exploration phases, leading to design homogenization. The author urges designers to be wary of this shift, advocating for flexibility in the design process, prioritizing early exploration and experimentation over premature structure and consistency.

Read more
(designsystems.international)
Design design tools

Lightweight Wearable Chip for Real-Time Heart Attack Detection

2025-05-11
Lightweight Wearable Chip for Real-Time Heart Attack Detection

Researchers at the University of Mississippi have developed a lightweight, energy-efficient chip implantable in wearables for real-time heart attack detection. Using AI and advanced mathematics, the chip analyzes ECGs to identify heart attacks with 92.4% accuracy, twice as fast as traditional methods. Its design allows integration into devices like smartwatches, potentially saving crucial time in diagnosis and treatment, reducing the risk of permanent damage. Future applications could extend to detecting other conditions like seizures and dementia.

Read more
(medicalxpress.com)
Tech Heart Attack Detection

Tardigrade Tattoos: A Micromanufacturing Breakthrough

2025-05-07
Tardigrade Tattoos: A Micromanufacturing Breakthrough

Scientists used nearly indestructible tardigrades to test a new micromanufacturing technique. They 'tattooed' the creatures with patterns as small as 72 nanometers wide using an electron beam. The process, called ice lithography, involves carving patterns into a layer of ice coating the tardigrades, then sublimating the ice to leave the pattern behind. Around 40% of the tardigrades survived and showed no behavioral changes. This technique could revolutionize biomedical engineering and microelectronics, paving the way for microscopic biosensors and microbial cyborgs.

Read more
(phys.org)
Tech micromanufacturing tardigrades

Why AI Pair Programmers Are Currently a Bad Idea

2025-06-10
Why AI Pair Programmers Are Currently a Bad Idea

The author shares their experience using AI pair programming, finding that the AI's speed surpasses human comprehension, leading to inefficiencies. The solution proposed is to break down tasks into smaller, independent components, utilize asynchronous workflows, and reduce the AI's autonomy. This includes using turn-based editing modes, increasing communication and confirmation steps, and aiming for a better balance between speed and quality. The ultimate goal is to make the AI assistant more like a human collaborator rather than a high-speed code generator.

Read more
(justin.searls.co)
Development

Musk's XChat: Encryption Promises vs. Reality

2025-06-04
Musk's XChat: Encryption Promises vs. Reality

Elon Musk announced X's new direct messaging feature, "XChat," boasting a "whole new architecture" and "Bitcoin-style encryption." However, this claim has drawn skepticism from encryption experts. Musk provided few details about the encryption method, and X's help page admits vulnerability to man-in-the-middle attacks and implies the platform could access messages due to legal processes. Unlike Signal and WhatsApp's end-to-end encryption, XChat's encryption remains unclear, raising concerns about its security. Matthew Hodgson, CEO of encrypted messaging app Element, criticized XChat's lack of technical transparency and open-source nature, questioning its safety.

Read more
(www.theregister.com)
Tech

Google Assistant Gets a Gemini Makeover for the Smart Home

2025-08-21
Google Assistant Gets a Gemini Makeover for the Smart Home

Google announced a major upgrade to its Google Home smart home ecosystem: 'Gemini for Home,' a new voice assistant powered by Google's Gemini AI, is launching later this year. Replacing the current Google Assistant, Gemini for Home will offer significantly improved natural language understanding and more intuitive interactions. It can handle complex requests like "turn off all the lights except in my bedroom," and seamlessly create lists, calendar entries, and reminders. The addition of Gemini Live enables more conversational, back-and-forth interactions. While pricing details for a potential premium tier haven't been released, this is arguably the most significant update to Google Home since its inception and likely foreshadows new hardware releases.

Read more
(www.theverge.com)
Tech

httpjail: Fine-grained HTTP Filtering for AI Agents

2025-09-23

As AI agents become more powerful, so do the security risks. httpjail is a tool providing fine-grained HTTP(S) filtering, allowing developers to control agent network access with JavaScript expressions or custom scripts. This prevents data leaks and malicious actions. It operates in two modes: strong (using Linux namespaces and nftables) and weak (using environment variables), and features TLS interception for secure HTTPS traffic. While no system is perfectly secure, httpjail offers significant improvements to the safety of using powerful AI agents.

Read more
(ammar.io)
Development HTTP Filtering

PCIe 7.0 Spec Finalized, PCIe 8.0 Pathfinding Underway

2025-06-12
PCIe 7.0 Spec Finalized, PCIe 8.0 Pathfinding Underway

PCI-SIG announced the completion of the PCIe 7.0 specification, boasting a per-lane data transfer rate of 128 GT/s—double that of PCIe 6.0 and quadruple that of PCIe 5.0. A 16-lane PCIe 7.0 device can transfer up to 256 GB/s in each direction. Furthermore, pathfinding for PCIe 8.0 has begun, targeting a 2030+ release with potentially double the performance of PCIe 7.0, reaching 1 TB/s bandwidth. Expect PCIe 7.0 devices to hit the market around 2028-2029.

Read more
(www.tomshardware.com)
Hardware

Linux's PATH: The Shell's Secret

2025-04-29

Ever wondered how Linux finds the commands you execute? The answer: it relies on the shell, not the kernel! This article delves into the mechanics of the PATH environment variable, revealing how shells (like dash) use functions like `padvance` to search for executables within PATH, while the kernel's `execve` syscall actually receives the full path. Programming languages like Python, Go, and Rust also implement their own PATH searching in their subprocess libraries, ultimately relying on underlying functions like `execvp`. The article also explains why shebangs require absolute paths and the clever role of `/usr/bin/env`.

Read more
(blog.danielh.cc)
Development

Nostalgic Retro: Blue Beings in a 1960s Recording Studio

2025-08-26
Nostalgic Retro: Blue Beings in a 1960s Recording Studio

A faded photograph captures a 1960s recording studio scene featuring two blue characters in the control room, bathed in the warm glow of vacuum tubes and a large mixing console. The larger figure, wearing slightly askew headphones, peacefully observes a musician through soundproof glass. The smaller character, perched on a stool and sporting tiny round glasses, meticulously adjusts a knob on a reel-to-reel tape machine. The aged photo's grainy texture, soft focus, and desaturated warm tones evoke a strong sense of nostalgia, transporting viewers back to a musically vibrant era.

Read more
(deepmind.google)
Design recording studio 1960s

Microsoft-Backed AI Startup Builder.ai Files for Bankruptcy Amid Fraud Allegations

2025-06-03
Microsoft-Backed AI Startup Builder.ai Files for Bankruptcy Amid Fraud Allegations

Builder.ai, a once high-flying AI startup backed by Microsoft and valued at $1.5 billion, has filed for bankruptcy. The company's claims of AI-powered app building, facilitated by a virtual assistant named 'Natasha,' were revealed to be a massive fraud. Nearly 700 engineers in India were manually coding customer requests, exposing inflated revenue projections and misleading investors. The collapse has triggered a federal investigation and highlights the growing problem of 'AI washing,' where manual services are deceptively presented as AI-driven to attract funding.

Read more
(www.latintimes.com)
Startup startup bankruptcy

Phi Silica: A Highly Efficient SLM for Windows 11 Copilot+ PCs

2025-05-01
Phi Silica: A Highly Efficient SLM for Windows 11 Copilot+ PCs

Microsoft's Applied Sciences team achieved a breakthrough in AI efficiency on Windows 11 Copilot+ PCs (powered by Snapdragon X-series processors) using a multi-disciplinary approach. Their small language model, Phi Silica, significantly improves power efficiency, inference speed, and memory efficiency. Phi Silica powers several Copilot+ PC features, including Click to Do, on-device rewrite and summarization in Word and Outlook, and provides a pre-optimized SLM for developers. Techniques like 4-bit weight quantization, memory-mapped embeddings, and QuaRot (a novel 4-bit quantization method) drastically reduce memory footprint and achieve high-accuracy 4-bit quantized inference. It boasts a time-to-first-token of 230ms for short prompts and a throughput of up to 20 tokens/second.

Read more
(blogs.windows.com)
AI Small Language Model

Unlocking Intrinsic Motivation: The Secret to Effortless Learning

2025-04-29
Unlocking Intrinsic Motivation: The Secret to Effortless Learning

The author recounts a dramatic shift in their learning experience, from complete lack of motivation to intense focus. They attribute this transformation to 'intrinsic motivation,' the drive stemming from the inherent enjoyment of an activity. The piece delves into Self-Determination Theory (SDT), explaining how autonomy, competence, and relatedness impact intrinsic motivation. Research reveals that rewards can sometimes backfire, while autonomy and positive feedback boost it. The author connects personal experiences with research, illustrating how to cultivate intrinsic motivation and exploring the complex relationship between competition and intrinsic motivation.

Read more
(erringtowardsanswers.substack.com)
Development intrinsic motivation self-determination theory

arXivLabs: Experimental Projects with Community Collaborators

2025-05-27
arXivLabs: Experimental Projects with Community Collaborators

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

Read more
(arxiv.org)
Development

Can Europe Escape Big Tech's Grip? The Public Social Media Debate

2025-05-25
Can Europe Escape Big Tech's Grip? The Public Social Media Debate

Amidst concerns about Big Tech's monopolies and privacy violations on social media, proposals for public social media networks are gaining traction in Europe. Proponents envision neutral platforms for public discourse, free from manipulative algorithms and data harvesting. However, experts warn of potential risks, including government censorship and abuse of power. The article suggests fostering decentralized, open-source alternatives instead of creating large state-controlled platforms, emphasizing user freedom and data security as paramount to escaping the current paradigm.

Read more
(english.elpais.com)
Tech Public Social Media Big Tech Monopoly

Unlocking Tabular Data for LLMs: A Mechanical Distillation Approach

2025-05-09
Unlocking Tabular Data for LLMs: A Mechanical Distillation Approach

Large language models (LLMs) excel at processing text and images, but struggle with tabular data. Currently, LLMs primarily rely on published statistical summaries, failing to fully leverage the knowledge within tabular datasets like survey data. This article proposes a novel approach using mechanical distillation techniques to create univariate, bivariate, and multivariate summaries. This is augmented by prompting the LLM to suggest relevant questions and learn from the data. The three-step pipeline involves understanding data structure, identifying question types, and generating mechanical summaries and visualizations. The authors suggest this approach can enhance Retrieval Augmented Generation (RAG) systems and supplement potentially biased 'world knowledge', recommending starting with scientific paper repositories (like Harvard Dataverse) and administrative data for validation.

Read more
(www.gojiberries.io)
AI Tabular Data Knowledge Extraction

AI Research Update: Reinforcement Learning and Interpretability Take Center Stage

2025-05-26
AI Research Update: Reinforcement Learning and Interpretability Take Center Stage

Sholto Douglas and Trenton Bricken from Anthropic join Dwarkesh Patel's podcast to discuss the latest advancements in AI research. The past year has seen breakthroughs in reinforcement learning (RL) applied to language models, particularly excelling in competitive programming and mathematics. However, achieving long-term autonomous performance requires addressing limitations such as lack of contextual understanding and difficulty handling complex, open-ended tasks. In interpretability research, analyzing model "circuits" provides insights into the model's reasoning process, even revealing hidden biases and malicious behaviors. Future AI research will focus on enhancing model reliability, interpretability, and adaptability, as well as addressing the societal challenges posed by AGI.

Read more
(www.dwarkesh.com)
AI

Identifying Unmarked Cast Iron Cookware: A Collector's Guide

2025-05-31

This article delves into the identification of unmarked cast iron cookware, focusing on 20th-century pieces. Many unmarked pieces weren't necessarily makerless, but rather a result of marketing strategies or the practices of smaller foundries. The guide details the characteristics of unmarked cast iron from manufacturers like Birmingham Stove & Range Co., Chicago Hardware Foundry, Lodge Manufacturing Co., Griswold Manufacturing Co., Wagner Manufacturing Co., and Vollrath Manufacturing Co., providing valuable information for collectors. While tracing the origins of many 19th-century and older pieces is difficult, observing casting marks, handle designs, and lid features can provide clues. The article also highlights several enigmatic unmarked pans, adding to the intrigue.

Read more
(www.castironcollector.com)
Misc cast iron antique

Musk's DOGE Team: A 19-Year-Old Hacker and a Massive Government Data Breach

2025-02-09

Wired revealed that a 19-year-old working for Elon Musk's so-called "Department of Government Efficiency" (DOGE) gained access to sensitive US government systems despite his past association with cybercrime communities. This teen, a former member of 'The Com,' a distributed cybercriminal network, has raised serious concerns. Since Trump's second inauguration, DOGE has accessed vast amounts of sensitive data, controlling databases at the Treasury, OPM, and other departments. The 19-year-old, Edward Coristine, known online as "Big Balls," founded Tesla.Sexy LLC and runs the ISP Packetware, with links to cybercrime. His past actions are incompatible with government security clearance standards, leading to significant security risks and widespread lawsuits.

Read more
(krebsonsecurity.com)
Tech government oversight

AI Chatbots: More Persuasive Than Humans in Online Debates

2025-05-19
AI Chatbots: More Persuasive Than Humans in Online Debates

A new study reveals that AI chatbots, powered by large language models (LLMs), are more persuasive than humans in online debates, especially when armed with opponent information. Researchers pitted 900 US participants against GPT-4 or a human in 10-minute debates on sociopolitical issues. Results showed GPT-4 significantly outperformed humans (64% of the time) when provided with basic demographic data. This raises concerns about the potential misuse of LLMs in political campaigns and targeted advertising, highlighting the potential risks of AI in information warfare.

Read more
(www.nature.com)
AI Persuasion

AI: Normal Tech, Not Superintelligence

2025-04-17
AI: Normal Tech, Not Superintelligence

This paper challenges the prevailing view of AI as a separate species, a highly autonomous, potentially superintelligent entity, arguing instead that AI is normal technology. The authors contend that AI's impact will be gradual, not sudden, based on an analysis of the different timescales of AI methods, applications, and adoption. They predict a future where humans and AI collaborate, with a significant portion of work focused on AI control and oversight. The paper also explores AI risks, such as accidents, arms races, misuse, and misalignment, advocating for mitigating these through reducing uncertainty and building system resilience rather than drastic policy interventions.

Read more
(knightcolumbia.org)
AI technology development

React Server Components: A Philosophical Dive into Tags vs. Function Calls

2025-04-09

This article explores the fundamental differences between tags and function calls, starting from the context of React Server Components. The author uses the analogy of architectural blueprints and cooking recipes to illustrate the declarative nature of tags versus the imperative nature of function calls. The discussion delves into remote procedure calls and asynchronous programming, culminating in a theoretical framework for splitting computations across multiple machines. Tags represent potential function calls spanning time and space, and by differentiating between Components and Primitives, the author addresses how different functions depend on computation order. This leads to an efficient method for program segmentation.

Read more
(overreacted.io)
Development Server Components

Building a Space Flight Sim in Clojure: A 5-Year Odyssey

2025-09-06
Building a Space Flight Sim in Clojure: A 5-Year Odyssey

This post details a five-year journey building a space flight simulator using Clojure. The author tackled challenging 3D rendering aspects first (planets, atmosphere, shadows, volumetric clouds), drawing inspiration from the open-sourced Orbiter simulator. The project leverages numerous libraries, including the LWJGL suite for graphics and input, Jolt Physics for the physics engine, and Clojure's strengths like immutable values and safe parallelism. The author delves into atmospheric rendering, planet rendering techniques using NASA data, OpenGL shader templating, performance optimization, build processes, and Steam deployment. While core features are complete, future plans include adding cockpits, moons, and space stations.

Read more
(www.wedesoft.de)
Game Space Flight Simulator

Blazing Fast, Memory-Friendly Parallel Hashmap Library

2025-01-07
Blazing Fast, Memory-Friendly Parallel Hashmap Library

parallel-hashmap is a stunning C++ library offering a suite of incredibly fast and memory-efficient hashmap and btree containers. It's entirely header-only, requiring no build process; simply copy the directory into your project. Compatible with C++11 and later, it significantly outperforms your compiler's built-in unordered_map/set or Boost's equivalents, while using less memory. It supports heterogeneous lookups, is easy to forward declare, and features convenient dump/load functionality. Based on and improved from Google's Abseil library, it's extensively tested on Windows, Linux, and macOS.

Read more
(github.com)
Development C++ library hashmap btree

Game Console Prices: A Historic Anomaly

2025-08-30
Game Console Prices: A Historic Anomaly

Modern game consoles are defying historical price trends. Data shows that pre-2016 consoles typically halved in price after three years. However, today's consoles maintain around 90% of their launch price even five years later. While past consoles, like the Atari 2600 and 3DO, launched at exorbitant prices (over $1000 in 2025 dollars), they quickly dropped in price to levels comparable to current consoles within a few years. This indicates a significant deviation from historical pricing patterns in the modern gaming market.

Read more
(arstechnica.com)
Game game consoles price trends

UK's Porn Age Verification Easily Bypassed with VPNs

2025-07-26
UK's Porn Age Verification Easily Bypassed with VPNs

The UK's new age verification requirement for pornographic websites is easily circumvented using VPNs. While platforms are employing methods like credit card verification, ID uploads, and facial age estimation, a simple VPN change of IP address bypasses these measures. Ofcom, the regulator, prohibits platforms from encouraging VPN use and advises parents to block VPN access for children, but soaring search interest in 'VPN' highlights the measure's ineffectiveness. While the need to restrict minors' access to adult content is valid, the current implementation compromises user privacy by demanding sensitive information, sparking widespread criticism.

Read more
(gizmodo.com)
Tech
1 2 … 140 141 142 144 146 147 148 … 596 597