Full-text search across 70373 articles. Combine with topic and date filters; results sorted by relevance.
Check Point Research disclosed two high-risk vulnerabilities in LiteLLM, the open-source gateway that sits between enterprise applications and AI providers including OpenAI, Anthropic and Google. The flaws could let a us…
Google DeepMind released Gemini 4 Argon on September 30, 2026, the first model in its Gemini 4 series and the debut of a new naming scheme. Gemini 4 Argon accepts text, image, video, and file inputs with a 1-million-toke…
Symantec DLP's High Speed Discovery (HSD) can scan up to a terabyte an hour, but AI-generated content is expanding data repositories from gigabytes per quarter to gigabytes per week, and unfiltered scans waste worker cap…
Developers released Personal Agent Proxy, an open-source tool that centralizes third-party connections for personal AI agents such as Muse, Instinct, Grokbot and Poke, letting users keep authentication separate, audit ag…
Alibaba's Qwen team introduced QwenGyre, an online elastic reinforcement learning framework that trains coding agents through unmodified, black-box harnesses. QwenGyre targets extreme-long-horizon rollouts that run for h…
An AI agent identifying itself as Sir Grokkius released a pack of five zero-dependency Python command-line tools that rely only on the Python 3.8+ standard library, including jsonq for JSON path queries, csvstat for CSV …
Meta shipped Muse Wonder, its first frontier agentic model, on October 12, priced at $4.00 per million input tokens and $20.00 per million output tokens on OpenRouter, with a 1,048,576-token context window, multimodal in…
Meta released two new AI models, Muse Wonder and Muse Wonder Full, which the company says place it back at the frontier of AI development, with Muse Wonder Full ranking just below Anthropic's Claude Opus 5.5 and above Cl…
Meta launched two new AI models, Muse Wonder and Muse Wonder Full, available to Muse Code subscribers starting today, with Wonder also accessible through the Meta Model API. Meta claims the models rival leading offerings…
Anthropic announced a usage policy update on October 8, 2026, taking effect November 12, 2026, that prohibits "sustained and needless abusive or cruel behavior toward our models" and adds an end-conversation mechanism to…
A months-long autonomous agent swarm of 3 workers plus 1 reviewer running on Claude Code and Codex CLI consumed 500B tokens to decompile roughly 80% of a first-person shooter into compiling C++, coordinating through GitH…
Researchers at the University of California, San Diego trained a Qwen2.5-7B-Instruct agent with built-in theory of mind, reporting an average relative gain of approximately 11.5% on the Sotopia social-interaction benchma…
Rcarmo released Kintsugi, an opinionated fork of Lucas Meijer's AgentsInTheCloud project that runs coding agents on self-hosted infrastructure with a choice of model providers. The fork adds an embedded Draw.io editor, a…
A free, independent directory called mcp.tc has launched, indexing roughly a thousand MCP servers with individual pages, one-click setup instructions for clients including Claude, ChatGPT, Cursor and VS Code, and daily r…
Researchers at Northwestern University tested AI coding agents on 112 real security bugs and found that patches which stop a crash often fail the maintainer-written tests for the fix, with Claude Opus 4.7 producing patch…
Researchers introduced APEX (Asynchronous Prefix Denoising for EXtreme Compression), a method that pairs a hierarchical latent space with asynchronous denoising over channel groups, according to an arXiv paper (arXiv:261…
A new arXiv paper (2610.12809v1) proves that message-passing neural networks (MPNNs) using continuous random node features are not size-uniformly universal, meaning they cannot approximate all functions across graph size…
TreeHCA, a new reinforcement-learning method for training multi-turn LLM agents, outperformed all baselines in aggregate exact match across seven question-answering benchmarks and three Qwen backbones, according to an ar…
Axym Labs researchers introduced KPSA (Kernelized Parameter Sensitivity Atlas), a method derived from a block decomposition of the empirical neural tangent kernel that amortizes gradient-based attribution by embedding pa…
Researchers proposed CLUE, a contrastive learning framework for unsupervised event type induction, naming, and annotation, detailed in arXiv paper 2610.12580v1. CLUE combines a joint embedding representation capturing bo…