AI Briefing#ai-agents
The latest AI and developer news about #ai-agents, with the original source and a short summary.
The latest AI and developer news about #ai-agents, with the original source and a short summary.
Slackbot has launched an AI client that connects 20+ apps via MCP to support collaboration.
This work identifies the 'Mosaic Effect,' where external queries by AI research agents leak sensitive information, and proposes a solution.
GitHub has introduced a new feature that limits the number of Pull Requests to reduce the burden on open-source maintainers.
mistral.rs now supports the OpenAI-compatible /v1/skills API, enabling agentic capabilities even on local models.
Based on Elasticsearch, a hybrid search architecture for agents' long-term memory was designed and its performance was verified.
Keye-VL-2.0, a 30B-scale multimodal model specialized for long-video understanding and agentic capabilities, has been released.
ktx, a context layer that helps data agents accurately understand metrics and business knowledge in a data warehouse, has been released.
Analysis of Stripe's Link data shows customer spending on AI products and app-builder platforms is surging rapidly.
SpaceXAI's Grok model is now natively available on Agent Bricks, Databricks' developer agent platform.
NVIDIA has unveiled NVIDIA XR AI, a framework for building AI agents that support real-time multimodal interaction on AR/XR devices.
It introduces a new benchmarking methodology that measures how much software APIs and documentation affect AI agents' task efficiency.
Chainguard has launched a security-hardened skill registry for AI coding agents.
GitHub Copilot boosts the efficiency of agentic tasks through prompt caching and automatic model selection (Auto).
Cloudflare has opened up the primitives of the Agents SDK and introduced Flue as its first supported framework.
An open-source UI kit dedicated to RAG has been released, featuring viewers for various document formats like PDF and DOCX as well as citation functionality.
Vercel announced 'Agentic Infrastructure' to support the building, deployment, and automation of AI agents.
Hivemind has introduced SkillOpt, which transforms AI agents' repeated patterns into optimized skills.
Greptile has unveiled TREX, an AI agent that finds runtime bugs by executing and testing code.
Amazon Bedrock AgentCore has launched a Web Search feature that fetches the latest web information securely, built on MCP.
AWS DevOps Agent has released a Release Management feature in preview that reviews and tests code changes.
JetBrains is adopting Copilot CLI as its default agent harness to ensure consistency in developer experience.
RewardHackBench has been released, using sandbox environments to prevent benchmark cheating by AI agents.
AWS's Strands Robots SDK integrates the LeRobot stack, simplifying data and policy transitions between simulation and real robots.
OpenAI's GPT-5.4, combined with an autonomous experimentation robot, dramatically increased the yield of the Chan-Lam Coupling reaction, a long-standing challenge in medicinal chemistry.