LangChain Improves Deep Agents with Harness Engineering
LangChain has demonstrated a significant leap in coding agent performance, moving from Top 30 to Top 5 on Terminal Bench 2.0 through "harness engineer...
LangChain
Alibaba releases Qwen3.5: Towards Native Multimodal Agents
Alibaba's Qwen has launched the first two models in its Qwen 3.5 series, featuring native multimodal capabilities with vision input. Among them is the...
Simon Willison
OpenAI Introduces GABRIEL Toolkit for Social Science Research
OpenAI has released GABRIEL, an open-source toolkit designed to assist social scientists. This tool leverages GPT models to convert qualitative data f...
OpenAI
GitHub Introduces Agentic Workflows for Repository Automation
GitHub has launched Agentic Workflows in technical preview, enabling developers to automate various repository tasks using coding agents within GitHub...
GitHub Blog
Hugging Face Explores Custom Kernels for AI Agent Skills
Hugging Face is exploring the creation of custom CUDA kernels, leveraging the capabilities of models like Codex and Claude. This initiative aims to en...
Hugging Face Blog
GPT-5.2 Achieves Breakthrough in Theoretical Physics
OpenAI's GPT-5.2 has reportedly derived a novel formula for a gluon amplitude, a significant development in theoretical physics. This new result was s...
OpenAI
OpenEnv in Practice: Evaluating Tool-Using Agents in Real-World Environments
Details in article.
Hugging Face
Z.ai Unveils Massive GLM-5 Model
Z.ai has released GLM-5, a substantial new MIT-licensed model featuring 754 billion parameters and 1.51TB of data on Hugging Face, making it twice the...
Simon Willison
LangChain Details Agent Sandbox Connection Patterns
LangChain has published an analysis of two key patterns for how AI agents connect with sandboxed environments. This exploration is vital for agents th...
LangChain
New Paper Explores Structured Context Engineering for File-Native Agentic Systems
A new paper by Damon McMillan delves into 'Structured Context Engineering for File-Native Agentic Systems,' addressing challenging LLM context tasks. ...
Simon Willison
Hugging Face Releases Transformers.js v4 Preview on NPM
Hugging Face has made a preview of Transformers.js v4 available on NPM, signaling an update to their popular library for running transformer models in...
Hugging Face
Anthropic Unveils Claude Opus 4.6 with Industry-Leading Performance
Anthropic has released Claude Opus 4.6, an upgrade to its most advanced AI model, showcasing industry-leading performance across various domains. This...
Anthropic