The Human-Centered Computing Foundation aims to create an alternative web architecture that prioritizes user data and attention. They are launching a campaign to secure a new Top-Level Domain for ethical, human-centered technology.
User is impressed with Qwen 3.6, a local model that performs well as general intelligence. It recommends Qwen 3.6 27B for its balance of speed and power.
Apple's MacOS 27 Golden Gate has improved app icons, restoring distinct shapes and sharper designs. However, MacOS 26 Tahoe's forced squircle shape for third-party icons is still a usability issue that Apple should address.
Fil-C supports memory-safe context switches using longjmp/setjmp and ucontext APIs. It enforces strict rules to prevent stack corruption and capability model violations.
We’re checking your connection to prevent automated abuse
WORK IN PROGRESS, THIS IS AMENDED AS I LEARN PDP-1 LISP MYSELF.But hopefully, it will get you started. Feedback is very much solicited in the Google Group :-) Lisp was developed in 1958 by John McCarthy at MIT. It pioneered a high-level, symbolic programming language designed for AI research. Early Lisp introduced key concepts such as recursion, symbolic expressions, and automatic storage ...
Ornith-1.0 is a state-of-the-art coding agent with four models (9B-Dense, 31B-Dense, 35B-MoE, 397B-MoE) that achieve top performance on coding benchmarks. It uses self-improving training and can generate reasoning blocks and solution rollouts.
We are introducing and open sourcing LongCat-2.0, a large-scale MoE language model with 1.6 trillion total parameters and ~48 billion activated per token — a substantial step up from previous LongCat models, accompanied by several architectural improvements. Both the full training run and the large-scale deployment are built entirely on AI ASIC superpods. Pretraining spans millions of ...
A graphical shell called Outer Shell allows remote access to server apps via SSH, providing a home screen of apps with APIs for integration. This approach enables native, platform-tailored apps for work and HTML for casual use, raising the bar for web development.
The US Supreme Court ruled that law enforcement's use of geofence warrants requires privacy protections under the fourth amendment. The court held that individuals have a reasonable expectation of privacy in their cell phone location data, even if they are in public areas.
South Korea's government and top tech companies are investing $1 trillion in megaprojects to bolster memory chip supply, build AI data centers, and deploy humanoid robots by 2028. The initiatives aim to secure South Korea's position in the AI industry but raise concerns about labor displacement and wealth distribution.
Trealla is a local-first knowledge base with content-addressable storage and hyper-relational graph. It uses Prolog as the main runtime and supports features like time-travel, deduplication, and full-text search.
The Apple Neural Engine is a fixed-function matrix accelerator in Apple chips since A11-class, exposed through Core ML. A reverse-engineered guide documents its datapath, dispatch route, and compiler format for A11-A18 and M1-M5 chips.
The Trump administration is cracking down on dissent, sentencing eight defendants to 450 years in prison for possessing or distributing information deemed anti-government. This move erodes the First Amendment and could lead to the criminalization of holding certain viewpoints.
A Game Boy emulator, WATaBoy, was created to test the feasibility of using JIT-to-Wasm for console emulation on iOS. The emulator's JIT compiler running in Wasm was found to be ~1.2x faster than the interpreter running natively.
Every time you switch tools, your AI forgets you. You repeat yourself. Context is lost. The AI that was finally starting to know you resets to a stranger. This is the AI memory silo problem. And it has the same solution as every silo problem before it: an open protocol. Open Memory Protocol is a vendor-neutral specification for how AI tools store, retrieve, and share memory about users and ...
To optimize insert speed with unordered data, batching and pre-sorting can help. A fast comparison method is needed, such as converting the first 8 bytes of a 20-byte random ID to an unsigned long.
The author shares their experience with astronomy and outdoor activities, presenting their telescope Wallace and its technical details. They also discuss the challenge of capturing the visual experience of hiking and trail running.
A journalist discovered nearly one million exposed passports and IDs on public internet due to a company's failure in data security practices. The documents were accessible through direct URLs with no authentication or encryption, creating a potential window for criminal exploitation.
Carson Gross discusses the benefits and drawbacks of using AI in coding, specifically with the Sorcerer's Apprentice problem where developers rely too heavily on AI and lose understanding of their code. He shares a personal experience with AI assistance in fixing a bug in his hyperscript project, highlighting the importance of human oversight and expertise in controlling complexity.
A CUDA program is compiled into machine code that runs on a GPU, involving multiple compilers and transformations from PTX to SASS, and is then launched on the GPU through a complex process involving the CUDA runtime, driver, and kernel-mode driver. The GPU executes the kernel in parallel, using warps and scheduling control bits to hide latency and optimize performance, and eventually copies ...
Dark-sky lighting reduces light pollution by directing light downwards and using amber colors, which helps protect wildlife and our circadian rhythm. It also reduces energy consumption and makes our stars visible again, promoting a safer and healthier environment.
As a part of the journey to transition Netflix’s compute infrastructure to be more Kubernetes-native, we have leaned into incorporating components from the Kubernetes ecosystem into our container platform Titus. One example of this is our use of Kueue, a cloud-native job queueing system for batch workloads, which has largely replaced the custom queuing and scheduling logic in our homegrown ...
Routers are becoming the control plane for AI inference, enabling collaboration between multiple models to improve accuracy and efficiency. The vLLM Semantic Router is a programmable, observable, and open serving layer that selects the right collaboration recipe for each request.
The website is using Cloudflare security and blocked my action, possibly due to a trigger like a certain word or SQL command. I was trying to perform an action when this page came up, and the Cloudflare Ray ID is needed to report the issue to the site owner.
Sandia National Laboratory built capacity to design and test ICs for radiation-hardened devices in the 1970s and 1980s for use in space missions and nuclear warheads. They successfully created the SA3000 processor, which could handle 1×106 rads of radiation with minimal performance reduction, and was used in the W88 nuclear warhead and other space applications.
JumpServer is an open-source PAM platform providing secure access to various endpoints through a web browser. It's licensed under GPLv3 with no warranties, distributed on an 'AS IS' basis.
Ornith-1.0 is a self-improving open-source model for agentic coding tasks, achieving state-of-the-art performance on various benchmarks. It learns to generate solutions and task-specific harnesses, outperforming leading models of similar size.
EuroISPA warns the European Commission that piracy blocking measures are becoming more extreme and causing overblocking incidents. The ISP organization argues that the Commission should prioritize implementing current law rather than introducing new enforcement obligations.