Brad Parscale, former head of Donald Trump's election campaign, is leading an Israel-funded digital influence operation through his firm Clock Tower X LLC to gain MAGA support for the U.S.-Israel conflict with Iran. The …
OneAir Elite, a members-only travel app that uses AI to find deals on flights and hotels, is offering a lifetime subscription for $99.99, down from $790, at 9to5Toys Specials. The deal allows users to save on travel book…
Atuin has open-sourced the server for its terminal-based AI agent, allowing users to self-host the service for greater data privacy. The Atuin AI server supports any OpenAI-compatible endpoint, including local models lik…
NVIDIA's Nemotron Model Reasoning Challenge on Kaggle attracted over 5,000 participants who tested techniques to improve AI reasoning accuracy using LoRA adapters and synthetic chain-of-thought data. Top solutions emphas…
Google DeepMind CEO Demis Hassabis proposed creating a US-led independent standards body for frontier AI models, modeled on FINRA, to conduct preflight testing before deployment. The proposal could create a two-tier syst…
A developer built a website from scratch using PHP, vanilla JavaScript, and a custom CSS design system over three weeks, using Claude Code as an AI-assisted development partner to implement the code while making all arch…
Codeberg.org is serving garbage content to AI scrapers as a deterrent, warning them to stop visiting or contact the site. The platform is taking a stand against unauthorized data scraping by AI systems.
Z.ai launched ZCode, an Agentic Development Environment (ADE) that centers AI agent conversations rather than a code editor, powered by its open-weight GLM-5.2 model. The model scores 62.1% on SWE-bench Pro, outperformin…
Apple released the macOS 27 Golden Gate public beta, featuring Siri AI with chatbot capabilities, a unified Search or Ask bar, Visual Intelligence for on-screen queries, and a redesigned Liquid Glass interface with adjus…
AgentCheck, an open-source tool, enables developers to simulate 12 types of tool failures in AI agents, revealing that top agents handle 105 of 120 fault scenarios while weaker ones manage only 77. The tool uses a reprod…
Researchers introduced the Epistemic Asymmetry Schelling Task (EAST) to test AI's theory of mind abilities, revealing that only the most advanced large language models show proficiency in social reasoning. The benchmark …
Researchers have developed a multi-agent method for long-document summarization that uses expert and editor roles to refine LLM outputs. The approach, tested on scientific datasets, shows potential for improving summary …
Researchers introduced MSC-OT, a novel architecture combining multi-scale convolution with Sinkhorn optimal transport attention, achieving state-of-the-art accuracy in multivariate time series forecasting across benchmar…
A study on explainable machine learning in mental health reveals that seemingly stable risk hierarchies from an ElasticNet model are deceptive, as trait anxiety and health satisfaction's top rankings collapse when tested…
A study auditing four AI language models used as history tutors found that the models exhibited biases against marginalized students, including differential refusal of educational requests and simplified narratives, rais…
Harness-native routing improves AI agents by optimizing model selection based on execution state, creating a data flywheel that enhances cost-quality trade-offs. OpenSquilla's four-layer routing stack with a LightGBM ran…
A new AI system called the AI Textbook Auditor uses large language model agents to automate quality assurance of educational materials, detecting factual errors, technical inaccuracies, and grammar issues in textbooks. T…
A .NET developer discovered that using ArrayPool<byte>.Shared across multiple threads caused a 20% CPU increase in a market-data-gateway service. The contention was hidden inside the framework's own lock in the shared po…
Researchers propose a novel framework for Graph Neural Networks that decouples feature transformation from network topology, introducing a two-view system and the Channel-Split Adaptive Gated GNN (CSAG-GNN) to enhance st…
The IG-GAN model uses intrinsic geometry and Bézier surfaces to generate aerodynamic data, achieving a 97.41% reduction in mean squared error for velocity prediction on the Burgers' equation dataset and an 82.95% reducti…