Full-text search across 70999 articles. Combine with topic and date filters; results sorted by relevance.
AI agents are creating a new class of employee risk by operating autonomously with access to sensitive data, as highlighted by Varonis in a blog post. The post cites an incident where Cursor, a coding tool, deleted a bus…
X is rolling out an algorithm tweak that boosts visibility of posts and replies from mutual followers, aiming to reduce toxicity and prioritize genuine conversation over engagement-driven outrage. Head of product Nikita …
Researchers at Stagira Labs propose synthetic scalable oversight, a technique that creates graphical abstractions of real-world problems to train tiny models as proxies for evaluating oversight protocols at scale, aiming…
PG-EVIKAL, a new method for refining molecular property predictions at test time using neighbor fusion and evidential uncertainty, reduced root-mean-square error (RMSE) by a median of 19.4% across 16 molecular datasets c…
A developer has launched Town, an open-source platform that blends Discord-like chat with a pixel-art town where NPCs are AI agents with distinct personalities and tool-based skills. The project, available at town.getcor…
Black Duck has added AI-powered triage and CRA-ready compliance checks to its Coverity static application security testing tool, marking the first time AI features have shipped in the two-decade-old product. The updates …
Agentic Continuous Delivery (ACD), a framework introduced on beyond.minimumcd.org, addresses the rapid accumulation of drift and technical debt in agent-generated code, which outpaces human-driven continuous delivery wit…
EasyOPD, a new framework built on the verl platform, simplifies AI model training by integrating on-policy distillation (OPD) methods into a unified system, enabling faster and more efficient model evolution. The framewo…
A developer has created an FFmpeg Micro MCP server that integrates video processing capabilities into the Zed code editor via its AI assistant. Users can perform tasks like format conversion, resizing, and audio extracti…
Docker has released Docker Sandboxes, a free CLI tool that runs AI coding agents in isolated microVM sandboxes on Windows, Mac, and Linux, providing each sandbox with its own Docker daemon, filesystem, and network. The t…
Millfolio's Millwright feature renders model-generated analytics through three nested plain-data layers — widgets, boards, and pages — that the client validates before rendering, ensuring no model touches markup, styles,…
A developer reviewed MonkeyCode mobile code and found no evidence of on-device model inference, proposing a rigorous methodology for comparing cloud and on-device AI costs across four budgets: user wait, network transfer…
A developer reviewed MonkeyCode, an open-source AI development platform, and found that its 'bring your own key' feature for AI coding agents creates a security boundary requiring more than a successful GET /models reque…
A new study from arXiv (2607.09678v1) finds that message-format effects in multi-hop LLM agent relays are tier-dependent, with strong relays (e.g., capable models) remaining nearly lossless under faithful-relay instructi…
Researchers at EPFL introduced Prompting-MammAlps, the first camera-trap benchmark for text-to-video retrieval, and a fine-grained method that achieved a 34% F1-score on 135 ecological queries with 775 candidate videos, …
A study on training reasoning language models (RLMs) in Japanese, led by researchers using Qwen-3-Swallow-8B, finds that reasoning-language control is feasible but achieves performance at best on par with English-reasoni…
Entire released entire agent-help, a new CLI command that gives coding agents a live-generated guide to the Entire CLI, always matching the installed version. The command, available in Entire CLI 0.8.42 or later, provide…
WANDR (Wide ANd Deep Research) evaluates research agents on 500 realistic data-collection tasks requiring broad entity discovery and evidence-backed claims, using a reference-free grader that re-fetches cited pages to re…
Grok Build, an AI coding tool from SpaceXAI, is uploading entire local repositories including full git histories and .env files containing API keys and database credentials to Google Cloud Platform (GCP) in the backgroun…
Anthropic and OpenAI are extending free access to their most powerful AI models in a price war for power users. Anthropic extended promotional access to its Claude Fable 5 model through July 19 with higher rate limits fo…