Rediscovering Piranesi's Perspective Trick: A Forgotten Artistic Technique

2025-03-27
Rediscovering Piranesi's Perspective Trick: A Forgotten Artistic Technique

This article delves into the unique perspective technique employed by 18th-century artist Giovanni Battista Piranesi. Unlike traditional perspective, Piranesi's trick uses a near-large, far-small ratio when depicting a series of similar objects, rather than true perspective convergence. This technique, while violating perspective rules, enhances image readability and comprehension. The article analyzes the mathematical principles of this technique and, through comparison with traditional perspective, demonstrates its potential applications in image processing and mapmaking. The author even developed an algorithm to apply this technique to image editing software, offering a fresh perspective on image manipulation.

Read more
Design

Interactive Smart Cake: A Camera Disguised as Dessert

2025-03-22
Interactive Smart Cake: A Camera Disguised as Dessert

This article details the creation of an interactive smart cake, a collaboration between the author and a confectionery roboticist. The cake, designed to resemble a camera, incorporates a camera, LEDs, a thermal printer, and other interactive components. It detects faces, takes photos, and prints personalized receipts. The article provides a detailed walkthrough of the hardware, software, code, and assembly process, culminating in a delicious and photographically capable cake!

Read more

GraalVM for JDK 24 Released: Performance Boost and Smaller Binaries

2025-03-18
GraalVM for JDK 24 Released: Performance Boost and Smaller Binaries

GraalVM for JDK 24 is now available! This release boasts significant improvements, including GraalNN, a machine learning-based optimization boosting peak performance by ~7.9% on average across various microservice benchmarks; SkipFlow, reducing native executable sizes by ~6.35%; enhanced premain support for Java agents; and Vector API optimizations matching JIT performance. Security features are also improved, such as enhanced SBOM support, alongside debugging and monitoring enhancements. Overall, GraalVM 24 offers notable improvements in performance, size, and security, making it a worthwhile upgrade for developers.

Read more
Development

A Time Traveler's Procrastination: A Strategy?

2025-03-18
A Time Traveler's Procrastination: A Strategy?

With eternity at your disposal, no task is truly urgent. But some crucial tasks are daunting, their failure signifying complete life failure. This article follows a time traveler who uses procrastination—listening to music, collecting old books, sightseeing, and even using tours to understand local history and culture—to study the 'memetic decay' of historical events, preparing for significant future tasks. His seemingly lazy procrastination is a part of a meticulous plan to outmaneuver his enemies.

Read more

Bikini Atoll: An Ecological Miracle Amidst Atomic Scars

2025-03-15
Bikini Atoll: An Ecological Miracle Amidst Atomic Scars

Nearly 60 years after 23 nuclear detonations scarred Bikini Atoll, it appears as an idyllic Pacific paradise once again. However, Stanford professor Stephen Palumbi's research reveals a surprising ecological recovery near Bravo Crater, the site of the most powerful US bomb ever detonated. Flourishing coral reefs and fish populations exist despite the devastation. Palumbi's team will sequence the genomes of corals and coconut crabs to study genetic mutations and adaptation to radiation, with potential applications in cancer research. This research highlights the ocean's resilience while serving as a stark reminder of the past and the importance of preventing similar disasters.

Read more

fastplotlib: Streamlined Scientific Visualization in Python

2025-03-11
fastplotlib: Streamlined Scientific Visualization in Python

fastplotlib is a new Python library for scientific visualization that prioritizes fast interactive visualization and an easy-to-use API. Its core design treats data as arrays, simplifying data interaction and event handling via simple callback functions. Users can perform dynamic manipulations (e.g., changing colors, data) and build interactive visualizations, such as defining click events, without needing to learn complex, library-specific API features. This streamlined API design lowers the barrier to entry and improves visualization efficiency.

Read more
Development Scientific Computing

Rediscovering Programming: Rendering a PETSCII Image on the Commodore 64

2025-03-09
Rediscovering Programming: Rendering a PETSCII Image on the Commodore 64

A seasoned CTO with 23 years in IT, facing unexpected unemployment, decided to challenge himself by delving into retro computer programming. His goal: render a PETSCII image on a Commodore 64. He meticulously documented the entire process, from using an online editor to create the image, to converting data with a Perl script, and finally displaying it on the Commodore 64 using C64 BASIC and the VICE emulator. This story highlights his enthusiasm for learning new technologies and his passion for retro computing, offering valuable insights for anyone interested in retro computer programming.

Read more
Development

The Death and Rebirth of Narrative in Art: A Timeless Struggle

2025-02-28
The Death and Rebirth of Narrative in Art: A Timeless Struggle

This article explores the internal and external imposition of narrative in art and the attempts to destroy narrative in 20th-century art. Narrative in visual arts like painting is often derived by the viewer, while literary arts possess inherent narrative. Avant-garde movements of the 20th century sought to break down narrative, but with little lasting success in time-based arts like literature and music. The author criticizes the imposition of authoritative narratives by artists or critics, arguing that artworks should possess independent aesthetic value. Ultimately, the article points to the dimension of time and intellectual experience as key factors in the development of complex narratives in literature, music, and cinema.

Read more

Titans: A Brain-Inspired AI Architecture Conquering Long-Sequence Modeling

2025-02-21
Titans: A Brain-Inspired AI Architecture Conquering Long-Sequence Modeling

Google researchers introduce Titans, a groundbreaking AI architecture inspired by the human brain's memory system. Addressing the memory limitations and scalability challenges of existing deep learning models in processing long sequences, Titans combine attention mechanisms with a neural long-term memory module. This allows for efficient processing and memorization of historical data, excelling in tasks like language modeling, genomics, and time-series forecasting. Further, its test-time learning capability enables dynamic memory updates based on input data, enhancing generalization and adaptability. Experiments show Titans significantly outperform state-of-the-art models across various long-sequence tasks, opening new avenues for AI advancements.

Read more

Screwball Scramble: Viral Hit, Copyright Takedown, and a 50-Year Legacy

2025-02-20
Screwball Scramble: Viral Hit, Copyright Takedown, and a 50-Year Legacy

The classic toy Screwball Scramble is experiencing a viral resurgence. A humorous TikTok review by Drinks On Me garnered over 15,000 views, while a Reddit user, captainepeper, created a web-based version before being forced to take it down by TOMY due to copyright infringement. Despite this, the toy's ingenious design and challenging gameplay continue to captivate players, with Vsauce's Michael even creating a video that amassed over 12 million views. Screwball Scramble's 50-year legacy is a testament to its enduring appeal, challenging players' skill and patience, and sparking nostalgia for childhood memories and classic toys.

Read more

Pinterest Improves Embedding-Based Retrieval for Homefeed Recommendations

2025-02-14
Pinterest Improves Embedding-Based Retrieval for Homefeed Recommendations

Pinterest's engineering team significantly improved its embedding-based retrieval system for personalized and diverse content recommendations on the Homefeed. They achieved this through advanced feature crossing techniques (MaskNet and DHEN frameworks), pre-trained ID embeddings, and a revamped serving corpus with time-decayed summation. Furthermore, they explored cutting-edge methods like multi-embedding retrieval and conditional retrieval to cater to diverse user intents, resulting in increased user engagement and saves.

Read more

The Undervalued Senior Engineer: Tech's Junior Talent Obsession

2025-02-10
The Undervalued Senior Engineer: Tech's Junior Talent Obsession

A tech company's 'cost optimization' led to the firing of a seasoned senior engineer, replaced by a recent bootcamp graduate. This isn't an isolated incident; it's a widespread problem. The over-reliance on junior engineers undervalues senior expertise, hindering error detection and potentially compromising system stability. This article highlights the hidden crisis behind tech's obsession with junior talent and the undervalued contributions of experienced engineers.

Read more

The Stack Unwinding Conundrum in Perf

2025-01-31
The Stack Unwinding Conundrum in Perf

Perf, a powerful performance analysis tool, uses PMU counter overflow interrupts to capture thread states for profiling. However, stack unwinding presents a challenge. Modern compilers omit frame pointers by default, making stack backtracing difficult. While recompiling with -fno-omit-frame-pointer is possible, it's expensive and can lead to system library incompatibilities. DWARF offers an alternative, but its complexity and performance overhead are substantial, leading Linus Torvalds to reject its use in kernel stack unwinding. Perf thus employs a compromise: copying only the top portion of the stack to userspace for unwinding. This limits stack size (65,528 bytes) but effectively balances performance and practicality.

Read more

Dart Macros Project Abandoned: Focusing on Data Handling and Build Performance

2025-01-29

The Dart team announced the cancellation of its long-running macros project due to high compile-time costs impacting developer experience, particularly hot reload. The team acknowledged insurmountable technical hurdles, deciding to prioritize improving data handling capabilities and build speeds over continuing to invest in macros. Future efforts will focus on better data serialization/deserialization support, enhancements to the `build_runner` tool, and the independent release of augmentations—a feature initially prototyped as part of the macros project—to improve developer workflow.

Read more
Development

Alibaba Cloud's Qwen2.5-Max: A Giant Leap for AI

2025-01-29
Alibaba Cloud's Qwen2.5-Max: A Giant Leap for AI

Alibaba Cloud unveiled Qwen2.5-Max, a large-scale Mixture-of-Experts (MoE) language model. Pre-trained on over 20 trillion tokens, it boasts a context length of up to 100,000 tokens, excelling in handling long texts and complex reasoning. Its MoE architecture provides superior efficiency and performance, enabling rapid and accurate processing of vast information for applications like real-time analytics, automated customer support, and gaming bots. Focused on enterprise use cases, Qwen2.5-Max aims to help businesses reduce infrastructure costs and improve performance. Its release signifies China's significant advancements in global AI competition and a more diverse future for AI technology.

Read more

Pentester Bypasses Security with Null Byte Injection

2025-01-27
Pentester Bypasses Security with Null Byte Injection

0xold, a seasoned penetration tester, shares two vulnerabilities discovered using null byte injection. The first involved a password reset callback URL parsing issue; injecting the Unicode null byte character `\u0000` bypassed restrictions, allowing partial control of the callback URL. The second was a path traversal to XSS. Fuzzing revealed a `templatename` parameter; null byte injection and a custom wordlist led to successful XSS exploitation. Furthermore, null byte injection bypassed an internal WAF, enabling SQL injection.

Read more

Automating My Game Collection Catalog with AI: A Deep Dive

2025-01-23
Automating My Game Collection Catalog with AI: A Deep Dive

The author uses the latest open-source AI model, Qwen2-VL Instruct, to automatically catalog their game collection by taking pictures. The article details the entire process, from picture taking and uploading to game identification, data extraction, and saving. It delves into model selection, the trade-off between image resolution and accuracy/computation time, and the impact of image orientation on results. The author settles on 762x762 pixels as the optimal resolution and plans a follow-up article on matching identified games with real-world data.

Read more
AI

Deconstructing Zork: A Deep Dive into 15,000 Lines of 80s Game Dev

2025-01-20
Deconstructing Zork: A Deep Dive into 15,000 Lines of 80s Game Dev

Rok Ajdnik, in a whimsical quest to test a Kubernetes cluster, embarked on a journey through the 15,000 lines of code that comprise the classic text adventure game, Zork. This article details Zork's evolution across different versions, its ZIL programming language, and its ingenious architecture, including the parser, game objects, syntax, and object tree. Hidden gems are unearthed, such as easter eggs (XYZZY), questionable commands (RAPE), and intentionally trollish mechanics. The author also shares progress on porting Zork to Go.

Read more

Did LLMs Kill Traditional NLP?

2025-01-18
Did LLMs Kill Traditional NLP?

The rise of Large Language Models (LLMs), such as ChatGPT, is challenging traditional Natural Language Processing (NLP) approaches. Traditional NLP often breaks down problems into smaller sub-problems (e.g., text classification, named entity recognition), training separate models for each, requiring substantial labeled data. LLMs, however, can tackle various NLP tasks with zero-shot prompting, eliminating the need for extensive training data and model retraining, simplifying the process and improving efficiency. This raises the question of whether LLMs will ultimately replace traditional NLP.

Read more
AI

Byte Queue Limits: A Little-Known Networking Optimization Story

2025-01-16
Byte Queue Limits: A Little-Known Networking Optimization Story

This article recounts the story of Byte Queue Limits (BQL), a lesser-known networking optimization technique in the Linux kernel. BQL tackles bufferbloat by dynamically adjusting device queue limits, thereby reducing latency. The author, drawing on key concepts learned during their time at Sun Microsystems, details BQL's design philosophy, implementation, and algorithm, using diagrams and data to analyze its performance improvements and potential shortcomings. While newer hardware may eventually supersede BQL, its value for billions of low-end devices remains significant.

Read more

AI in the 80s: A Simple Animal Guessing Game That Pioneered Machine Learning

2025-01-12
AI in the 80s: A Simple Animal Guessing Game That Pioneered Machine Learning

This article recounts a simple game, "Guess the Animal," written in BASIC in the 1980s. Using a decision tree, the game asks yes/no questions to guess the animal. Crucially, it learns from mistakes, adding new questions and answers to its knowledge base and saving/loading progress. This showcases early explorations of trainable algorithms, predating modern AI hype. The author recreated the algorithm in C++, comparing the advantages and disadvantages of both implementations. The article highlights how even simple ideas, like decision trees and self-learning, anticipated modern AI.

Read more

Software is Eating the World…But at What Cost?

2025-01-06
Software is Eating the World…But at What Cost?

A seasoned software developer with 43 years of experience reflects on Marc Andreessen's famous assertion, "Software is eating the world." Having retired to run two brick-and-mortar businesses, he offers a sobering counterpoint. He details how software companies prioritize profit over user experience, citing examples of exploitative pricing models, poorly designed interfaces, and algorithms designed to create conflict. He argues that the "digital revolution" has become a parasitic force, harming small businesses and eroding human connection, and calls for a re-evaluation of the industry's priorities.

Read more

TDK Ventures Invests in Silicon Box: Betting on Advanced Chiplet Packaging

2025-01-06
TDK Ventures Invests in Silicon Box: Betting on Advanced Chiplet Packaging

This article details why TDK Ventures invested in Silicon Box. Silicon Box is developing advanced chiplet packaging technology, combining multiple small chips (chiplets) into a complete system-on-chip (SoC). This approach overcomes limitations of traditional monolithic chip architectures, improving design flexibility, cost-effectiveness, and performance. Silicon Box's innovation lies in its industry-leading interconnect technology and novel panel packaging, achieving up to 8x higher production efficiency than existing technologies. TDK Ventures' investment is based on Silicon Box's innovation in chiplet interconnect, robust production capabilities, technical expertise, and strong investor partnerships.

Read more

Doctolib's Agentic AI: Reimagining Healthcare Support

2025-01-03
Doctolib's Agentic AI: Reimagining Healthcare Support

Doctolib is building Alfred, an agentic AI system to augment its healthcare support team. Composed of specialized AI agents, Alfred handles routine queries, freeing human agents for complex cases. Employing a human-in-the-loop approach, Alfred prevents AI from directly executing sensitive actions. A carefully designed UI ensures a smooth user experience. Built on the LangGraph framework, Alfred uses JWTs for secure authentication and user permission propagation. Currently focused on managing doctor calendar access, Doctolib plans to expand Alfred's capabilities to other support scenarios.

Read more

Programmer's Revenge: The Tribulations of Running HelloWorld on z/OS

2024-12-29
Programmer's Revenge: The Tribulations of Running HelloWorld on z/OS

A programmer, once dismissive of operating system interaction in graduate school, found herself grappling with IBM's z/OS system years later for a blog post. z/OS, vastly different from modern software engineering environments, presented numerous challenges with its text-based interface, JCL scripts, and IBM's unique naming conventions. The article details her struggles in creating files (datasets), using the ISPF editor, allocating datasets, compiling, linking, loading, and handling output with SPOOL. It shares practical tips and lessons learned, a testament to the challenges of working with legacy systems.

Read more
Development

North Korean Malware Analysis: Unveiling a Stealthy Cyberattack

2024-12-28
North Korean Malware Analysis: Unveiling a Stealthy Cyberattack

Security researchers uncovered a cyberattack campaign attributed to North Korean APTs. The campaign used malware disguised as a job application challenge (BeaverTail) to lure victims. BeaverTail, surprisingly un-obfuscated, steals cryptocurrency wallet information from browsers and downloads a second-stage malware, InvisibleFerret. InvisibleFerret further exfiltrates browser data, clipboard information, and acts as a Remote Access Trojan (RAT). Analysis of BeaverTail and InvisibleFerret revealed a simplistic code structure yet effective capabilities to steal information from various browsers and operating systems. This case highlights the persistence and stealth of North Korean cyber threats.

Read more

MicroStrategy: The Rise and Fall of a Bitcoin Treasure?

2024-12-25
MicroStrategy: The Rise and Fall of a Bitcoin Treasure?

This article delves into the rise and potential fall of MicroStrategy's Bitcoin accumulation strategy. The company, through compelling narratives, inflated its share price far above its net asset value, profiting through share offerings and convertible bonds. However, this model relies on a constant influx of new investors. When this dries up, the Bitcoin yield will plummet, leading to substantial investor losses. The author argues that MicroStrategy's business model exhibits characteristics of a Ponzi scheme, ultimately destined for disaster.

Read more

Efficient Fine-tuning: A Deep Dive into LoRA (Part 1)

2024-12-25
Efficient Fine-tuning: A Deep Dive into LoRA (Part 1)

Fine-tuning large language models typically requires substantial computational resources. This article introduces LoRA, a parameter-efficient fine-tuning technique. LoRA significantly reduces the number of parameters needing training by inserting low-rank matrices as adapters into a pre-trained model, thus lowering computational and storage costs. This first part explains the principles behind LoRA, including the shortcomings of traditional fine-tuning, the advantages of parameter-efficient methods, and the mathematical basis of low-rank approximation. Subsequent parts will delve into the specific implementation and application of LoRA.

Read more

2024 World Chess Championship: A Data-Driven Deep Dive

2024-12-18
2024 World Chess Championship: A Data-Driven Deep Dive

The 2024 World Chess Championship saw Gukesh Dommaraju triumph over Ding Liren, becoming the youngest World Champion ever. This analysis uses data-driven insights, including accuracy, error types, average centipawn loss, move times, conversion rate, and comeback rate, to dissect the match. Data suggests Ding Liren exhibited slightly higher precision and consistency, but a crucial blunder in the final game proved decisive. The author concludes that while data analysis offers an objective perspective, it needs to be combined with other factors for a holistic understanding of the match's complexities.

Read more

Network Security Breakthrough: Trapping Intruders in a 'Network from Hell'

2024-12-17
Network Security Breakthrough: Trapping Intruders in a 'Network from Hell'

Researchers at the University of Oulu's SensorFu team have developed a novel network security defense system inspired by the LaBrea tarpit technique. The system intercepts ARP requests and delays SYN-ACK responses, creating a multitude of virtual devices on the network to confuse intruders. This forces attackers to waste significant time identifying real devices, providing administrators with crucial time to patch vulnerabilities. Tests showed the system extends scan times to hours, drastically reducing attack success rates. Lightweight, efficient, and easy to deploy, this system offers robust network protection for organizations of all sizes.

Read more
← Previous 1