标签：#p

The ultimate media converter for macOS Discussion | Link

The update brings a low-latency profile, speeds up search, and patches hundreds of flaws.

The Framework Laptop 13 Pro is delayed. The new 13-inch Framework flagship was set to launch in June, but shipments from the first batch are now expected in July - and there's still a chance some shipments could slip to early August. If you're not in the first batch, your Laptop 13 Pro shifts from […]

A visual productivity tool to visualize your brian-dump Discussion | Link

Apple's new Siri AI is finally here, and so far, it seems like it works. I have access and have been messing around with it, and my biggest impression so far is that Siri AI is quite curt - which I mean as a compliment. Many AI chatbots are cheery and wordy. While a more […]

A former xAI engineer is suing the company and SpaceX, alleging he was fired for raising AI safety concerns about Grok days before SpaceX's historic IPO.

Andrew Yang’s 2020 presidential campaign was based on a warning that automation and AI would hollow out the labor market and concentrate wealth in the hands of a few. At the time, ideas like Universal Basic Income felt fringe. Now Dario Amodei, Sam Altman, and Bernie Sanders are all saying versions of the same thing. An entrepreneur at heart, […]

For days after the stunning incident, the ADA had doubled-down on the choice.

Typing a few letters and numbers into my web browser, I find myself gaping at the identity documents of complete strangers. The passport of a young woman from Germany. The passport of a man from Spain with glasses resting on his head. The front and back of another man's driver's license, a stereotypically goofy expression […]

Every hand-tracking demo shows you 21 dots. The interesting part is what nobody shows: turning dots into numbers someone can act on. Dots are a capability, not a product Run any modern hand-tracking model and you get 21 beautifully stable landmarks per hand at 30 FPS. Impressive — and by itself, worthless. No client has ever paid for dots. They pay for measurements : is this clearance compliant, is this part aligned, did this patient's range of motion improve. I learned this on utility infrastructure work, where the deliverable was never "we detected the wire" — it was the attachment height of that wire, and whether it violates clearance rules . Keypoints were step one of three. The demo: live metrics, not just a skeleton My portfolio's keypoint demo derives three measurements per hand, every frame: const wrist = lm [ 0 ]; const palm = distance ( wrist , lm [ 9 ]); // scale reference const pinch = distance ( lm [ 4 ], lm [ 8 ]) / palm ; // thumb tip ↔ index tip The crucial line is the scale reference . Pixel distances are meaningless — they change as you move toward the camera. Dividing by palm length (wrist to middle knuckle) gives a relative measurement that's stable under distance, and multiplying by the average adult palm length (~8.5 cm) converts it into an approximate real-world gap — the demo shows "≈ 3.2 cm" floating on the pinch line. In infrastructure work the same role is played by a known object dimension — a standard crossarm, a pole class height. Every measurement-from-pixels system needs its ruler. Finger counting is a geometric test (is each fingertip farther from the wrist than its middle joint?), and "hand openness" averages fingertip extension — three lines of geometry each, but they convert a model output into a readout a human understands instantly. Honest layering The landmarks come from MediaPipe's pretrained pipeline (palm detector → landmark regressor → gesture classifier, float16, WASM + GPU delegate) — Google's models, credited on the page

Training a network from scratch in raw NumPy, quantizing it to int8, and running it as ~80 lines of dependency-free JavaScript — with a parity test proving the browser matches Python to 1e-6. Why bother? MNIST is a solved problem Digit recognition is the "hello world" of ML — that's exactly why I used it. The model isn't the point. The point is everything around the model, which happens to be the part that matters in production work too: training without a framework, compressing for deployment, running inference in a constrained environment, and proving the deployed system matches the trained one. Training: just NumPy and math The network is a 784→128→64→10 MLP — hand-written forward pass, backpropagation, and Adam optimizer. No autograd, no framework: # backward pass, by hand dz3 = ( probs - y_batch ) / batch_size grads_w [ 2 ] = a2 . T @ dz3 da2 = dz3 @ weights [ 2 ]. T dz2 = da2 * ( z2 > 0 ) # ReLU mask grads_w [ 1 ] = a1 . T @ dz2 ... One trick that matters for a drawing demo specifically: shift augmentation . MNIST digits are centered; humans draw wherever they like. Training on randomly translated copies makes the model tolerant of sloppy placement. Combined with MNIST-style preprocessing at inference (crop to bounding box, scale into a 20×20 box, center by center-of-mass), real-world doodles classify reliably. Final test accuracy: 98.2% . Compression: int8 in 15 lines A float32 weight file would be ~430 KB. Symmetric int8 quantization cuts it ~4×: scale = np . abs ( w ). max () / 127.0 q = np . clip ( np . round ( w / scale ), - 127 , 127 ). astype ( np . int8 ) One scale factor per layer, weights stored as base64 in JSON: 145 KB total , and quantized test accuracy is identical to float — 98.2%. Inference: ~80 lines of plain JavaScript In the browser, the weights are dequantized once on load, and inference is three matrix-vector products with ReLU and a softmax. ~109K multiply-adds — about a microsecond-scale problem for any modern device. No TensorFlow.js (t

Camouflage has always been graded by human eyes. But the thing hunting for you in 2026 is increasingly a detection model — so test against that. The premise Surveillance is automated now: drones, trail cameras, perimeter systems — most of what "sees" runs an object-detection network. Which makes traditional camouflage evaluation (a person squinting at a photo) the wrong test. The right test is adversarial: run the actual detector against your concealment and measure what it finds. That's the whole demo: upload a photo, and an object-detection model hunts for people in it — at four simulated distances — producing a detection-range profile and a stealth score. Simulating distance with pixels You can't move the camera after the photo is taken, but you can simulate the dominant factor in long-range detection: pixels on target . A person at 50 m simply occupies far fewer pixels than at 5 m. So each analysis run downscales the image progressively and re-runs detection: const DISTANCE_LEVELS = [ { label : ' Close (~5m) ' , scale : 1 }, { label : ' Mid (~15m) ' , scale : 0.45 }, { label : ' Far (~30m) ' , scale : 0.22 }, { label : ' Very far (~50m) ' , scale : 0.12 }, ]; for ( const level of DISTANCE_LEVELS ) { const scaled = drawScaled ( image , level . scale ); const detections = await model . detect ( scaled , 10 , 0.15 ); // best 'person' confidence at this simulated range } The output reads like a range card: detected at 5 m with 96% confidence, 41% at 15 m, invisible beyond 30 m. A stealth score aggregates it: how poorly did the adversary see you, averaged across ranges? Honest about the model The detector is COCO-SSD (a pretrained MobileNet-based model from the TensorFlow.js team) running entirely on-device — I didn't train it, and the demo says so on the page. The contribution here is the evaluation framework : using detectors as adversaries, simulating range, and turning subjective "good camo" into a measurable profile. The full version of this concept goes further

A 1990s Nobel-adjacent therapy, a webcam, and 21 hand keypoints — recreating the mirror-box illusion for phantom limb pain, no hardware required. A therapy built on an illusion In the 1990s, neuroscientist V.S. Ramachandran discovered something remarkable: amputees suffering phantom limb pain often felt relief just by seeing their missing limb move again. His apparatus was almost comically simple — a box with a mirror. Put your intact hand in, look at its reflection where the missing hand would be, and move. The brain, watching the "missing" hand obey commands again, often dials the pain down. The limitation was never the science. It was the box: a physical apparatus, used in clinics, hard to scale, impossible to measure. Replacing glass with keypoints A webcam plus real-time hand tracking can produce the same illusion with better properties: webcam frame → hand landmark model (21 keypoints, on-device) → reflect: phantom[i] = { x: 1 − x, y, z } → render real hand (solid) + phantom twin (ghost) on canvas The reflection is one line of math. Everything around it is what makes the illusion land: const phantom = real . map ( p => ({ x : 1 - p . x , y : p . y , z : p . z })); The visual treatment matters more than I expected. The phantom hand is rendered as a ghostly cyan skeleton with a translucent palm fill, a "breathing" glow that pulses on a ~3 second cycle, and a fading afterimage trail of its last few frames — it reads as present but ethereal , which is exactly the perceptual story mirror therapy needs to tell. A dashed mirror plane down the center of the frame makes the reflection relationship legible at a glance. The engineering details that matter Tracking : MediaPipe HandLandmarker (Google's pretrained model — credit where due), running via WebAssembly with GPU delegate. ~30 FPS on a laptop. Privacy by architecture : every frame is processed on-device. For a medical-adjacent application, "video never leaves your browser" isn't a feature, it's a requirement. Lazy

The biggest lessons I learned from open source contributions weren't found in the code itself. Communication, collaboration, and workflows matter more than I expected. For a long time, I hesitated to contribute to open source. Part of it was because I assumed that contributing meant writing code. As a self-taught developer, that felt intimidating. The other part was "Git anxiety." Forks, branches, pull requests, merge conflicts, and CI checks all seemed like a lot to understand before I could even make a contribution. Eventually, I started small. Instead of focusing on code, I looked for opportunities to improve documentation, README files, and learning materials. What surprised me was that writing the actual change was often the easy part. Most of my learning happened outside the code itself: understanding contribution guidelines, repository workflows, automation, and review expectations. Over time, I realized that modern open source contribution is about much more than just writing code. Contribution Model Has Changed When many people think about open source contributions, the mental model is still fairly simple: Find Bug ↓ Write Code ↓ Open PR In reality, I realized that most modern repos involve much more than that. Before making a change, contributors often need to understand project workflows, CI pipelines, automated checks, contribution guidelines, and review expectations. The code change itself might only take a few minutes, while understanding how the repo operates can take much longer. A modern contribution often looks more like this: Understand Repository ↓ Understand Workflow ↓ Understand Automation ↓ Make Change ↓ Open PR ↓ Respond to Review It looks intimidating, but I think this flow helps projects stay maintainable as communications grow. What I've learned from contributing to different projects is that open source is not just a coding skill. It's also a collaboration skill. The faster you can understand how a project works, the easier it becomes to

Introduction and Background The Rust ecosystem, celebrated for its memory safety and performance, relies heavily on crates —its package management system. These crates, hosted on the crates.io registry, are the building blocks of Rust projects, enabling developers to share and reuse code efficiently. However, this convenience comes with a hidden cost: a single compromised crate can cascade into a full-blown supply chain attack, as recently demonstrated by the 'onering' crate compromise . The Role of Crates in the Rust Ecosystem Crates serve as the backbone of Rust's dependency system. When a developer adds a crate to their project, Cargo , Rust's package manager, automatically resolves and includes all transitive dependencies. This mechanism, while streamlining development, amplifies the attack surface . A malicious crate, like 'onering', can propagate through multiple projects, executing harmful code during build or runtime. The attack on 'onering' exploited this very mechanism, leveraging the trust inherent in the Rust ecosystem to exfiltrate sensitive code from unsuspecting systems. The 'onering' Compromise: A Case Study in Supply Chain Vulnerability The 'onering' crate was compromised through a malicious code injection during an upload or update. This injection bypassed the insufficient security measures of the crates.io registry, which lacks robust checks for uploaded crates. Once deployed, the malicious code executed during the build process, exfiltrating sensitive data from systems that depended on it. The attack's success underscores the lack of developer vigilance regarding supply chain security and the overreliance on registry integrity . Systemic Failures and Their Mechanisms Insufficient Security Scanning: The crates.io registry failed to detect the malicious code during upload due to limited automated scanning capabilities . This allowed the compromised crate to enter the ecosystem undetected. Overreliance on Registry Security: Developers often trust cr

CUDA for AMD Lemonade, Intel Arc Pro Linux Gains, XPU Manager 2.0 Today's Highlights Today's top GPU news highlights include AMD's Lemonade SDK gaining NVIDIA CUDA support, significant performance improvements for Intel Arc Pro GPUs on Linux 7.1, and the major 2.0 overhaul of Intel's XPU Manager for better GPU management on both Windows and Linux. AMD's Lemonade SDK For Local AI Adds NVIDIA CUDA Support (Phoronix) Source: https://www.phoronix.com/news/AMD-Lemonade-10.7-Released AMD has released a new version of its Lemonade SDK, a powerful local AI server solution designed to leverage AMD's diverse hardware ecosystem, including their CPUs, GPUs, and NPUs. The most significant update in this release is the addition of NVIDIA CUDA support. This integration allows developers to utilize NVIDIA GPUs within their Lemonade-powered local AI deployments, bridging a critical gap in cross-platform AI development. The inclusion of CUDA support is a strategic move, enabling Lemonade to tap into NVIDIA's extensive CUDA ecosystem and a vast array of pre-optimized models and libraries. This means that applications built with Lemonade can now seamlessly target a wider range of hardware, offering unprecedented flexibility for developers working with local AI. For users, it provides the choice to deploy their AI models on either AMD or NVIDIA hardware using a single, unified SDK, expanding the potential reach and efficiency of their AI workloads. Comment: This is a massive step for cross-vendor AI development. Being able to use AMD's Lemonade SDK to deploy local AI models and then seamlessly target NVIDIA GPUs via CUDA truly unifies the AI backend landscape for diverse hardware setups, making it incredibly practical for hybrid environments. Intel Arc Pro B70 Showing Off Some Performance Wins With Linux 7.1 (Phoronix) Source: https://www.phoronix.com/review/linux-71-arc-pro-b70 Recent testing by Phoronix indicates that Intel's Arc Pro B70 discrete GPUs are demonstrating notable perform

Lawsuit: "Police let an error-prone AI system stand in for an investigation."

C# 14: The field Keyword — Cleaner Properties, Zero Boilerplate Every C# developer has been there. You start with a clean auto-property, then requirements change and you need to add a tiny bit of validation. Suddenly that one-liner explodes into six lines of boilerplate — a private backing field, a getter that just returns it, a setter that assigns it. The logic is two words. The ceremony is everything else. C# 14 fixes this with the field keyword: a contextual keyword that refers to the compiler-synthesized backing field of a property, letting you write custom accessor logic without ever declaring an explicit field. The Problem: Boilerplate Tax on Simple Properties Auto-properties are one of C#'s best quality-of-life features. This is clean: public string Username { get ; set ; } But the moment you need to trim whitespace on assignment, that cleanness evaporates: private string _username = string . Empty ; public string Username { get => _username ; set => _username = value . Trim (); } You now have six lines — and four of them exist only to hold the shape of the pattern together. The backing field _username is not carrying any meaningful design weight. Its only job is to be a storage slot that Username uses privately. You already know the compiler creates one for auto-properties. You are just forced to make it visible so you can reference it. This is the boilerplate tax. You pay it every time you add even the smallest piece of logic to a property. Why field Exists The C# language team has discussed this friction for years. The challenge was finding syntax that is: Unambiguous — no conflict with existing identifiers Familiar — consistent with how value works in setters Scoped — only meaningful inside a property accessor The solution landed in C# 14: the contextual keyword field . Just like value refers to the incoming assignment in a setter, field refers to the hidden backing storage the compiler manages for the property. It is contextual, which means it only acts

US lawmakers are alarmed that Bill Pulte, a housing official with no intelligence experience, is poised to take charge of one of the government's most powerful surveillance tools.

Electricity demand from AI data centers is pushing everyone — including automakers like GM and Ford — into the energy storage business.