Full-text search across 70304 articles. Combine with topic and date filters; results sorted by relevance.
OpenAI said on November 9th that a single dot can now connect to multiple local computers and choose which machine handles a given task, per a thread on X from its developer account. OpenAI framed the change around a pra…
A developer built agentcheck, an open-source CLI and TypeScript library whose mcp-scan command starts a stdio MCP server, enumerates its tools and lints every name, description and schema string for tool-poisoning attack…
Anthropic released the Claude Fable 5.2 and Claude Mythos 5.2 models on Monday, saying both generate outputs 30% faster than their predecessors with improvements in coding, reasoning and vision. Fable 5.2 is available ac…
Jane Street released a collection of agent skills, distributed as directories each containing a SKILL.md file with instructions for AI agents and chatbots. The first listed skill, socratic-tutor, turns an LLM into a stri…
Meta Superintelligence Lab released Muse Ember on Monday, its lightest and most cost-efficient model yet, priced at 10 cents per million input tokens and built for agent execution tasks such as tool calls, coding subtask…
AWS released Nitro Sandboxes, lightweight VMs managed by the Nitro Hypervisor that give each Amazon EKS pod, Amazon ECS task, or standalone VM — including AI agents — its own hardware-isolated virtual machine on a shared…
Baidu's ERNIE Evolve automated research system, which edits code, runs experiments and retains only verified gains, beat public best records on the NanoChat and NanoGPT Speedrun benchmarks and sped up inference for ERNIE…
Snowflake's Internal Audit team cut fieldwork time for evidence collection, testing and audit report production by 50% using Snowflake CoCo, the AI coding agent in Snowflake, according to the company's own account of int…
Cloudsmith announced an AI Agent Skills registry at KubeCon + CloudNativeCon North America 2026, adding a skills package type to its cloud native artifact repository so platform teams can host, publish, distribute and go…
A developer released Kiosk, an open interface layer that lets AI assistants such as the Hermes assistant discover, register with and transact with online services, after failing to order groceries through an assistant at…
German web hosting provider Mittwald CM Service disclosed that hackers compromised 56 customer projects after going undetected on its systems for eight days, exploiting a flaw in the company's Model Context Protocol (MCP…
CogitatorTech released Ordered, a Zig library of sets and maps that keep elements sorted by key at all times, targeting Zig 0.16 and 0.17 with older releases expected to work on Zig 0.15. The library is built around a si…
Meituan released LongCat-2.5, a natively multimodal model built on LongCat-2.0 that the company says strengthens native visual understanding and agentic coding for long-horizon agentic tasks. The model is listed with pro…
0sec Labs has open-sourced 0, an AI security harness for code review, pentesting, reverse engineering and fixes that runs locally through a browser console, a headless CLI and an MCP server using the user's own model pro…
A developer built Hearthmear, a Roblox RPG with 205 scripts, 10 dungeons, 7 bosses, 17 enemy types and three classes, entirely from a phone by editing .rbxl place files through an AI assistant and publishing via the Robl…
A paper posted to arXiv (2611.08704v1) by researchers studying English Wikipedia editors and students finds that real-world identifications of AI-generated writing rest less on specific textual features than on "pragmati…
Researchers posted arXiv:2611.09372v1, describing OpenMeter, a schema-agnostic multi-agent system that translates natural language queries over industrial smart meter telemetry into executable code and human-readable ins…
A new arXiv paper (2611.08711v1) introduces RL-ATTACK, a trainable reinforcement-learning framework that generates tailored adversarial transformations against black-box language models by learning from the target model …
A new arXiv paper (2611.08998v1) separates visual perception from computation in multimodal LLMs by pairing counting questions about the same diagram that fall in different counting-complexity classes, such as spanning t…
A community forum reply recommends running multiple trading agents against a single vLLM server, which provides continuous batching, chunked prefill, prefix caching and an OpenAI-compatible API so each agent acts as a cl…