Notebook Intelligence: A JupyterLab Extension for AI Code Assistants
Notebook Intelligence is a new JupyterLab extension designed to integrate various AI coding assistants, including Claude Code, Copilot, Ollama, and OpenAI-compatible LLMs. It aims to enhance the notebook development experience by providing features like MCP, skills, plugins, and notebook agents.
AgentFM: Decentralized AI Supercomputer via Peer-to-Peer Network
AgentFM introduces a peer-to-peer network designed to transform everyday computers into a decentralized AI supercomputer. This system enables the execution of large-scale AI workloads by leveraging a global mesh of idle CPUs and GPUs.
Mindburn-Labs Releases helm-ai-kernel for Secure AI Agent Execution
Mindburn-Labs has released helm-ai-kernel, an open-source execution firewall designed to enhance the security of AI agents. This tool acts as a fail-closed system, quarantining potentially risky tools and proxying requests to ensure secure operations.
AISix: Open-Source AI Gateway for LLMs and AI Agents
AISix is a new open-source AI gateway built in Rust, designed to provide a unified OpenAI-compatible API for various large language models and AI agents. It offers features like routing, guardrails, caching, rate limiting, and observability, aiming to simplify the management of AI infrastructure.
Benchmark Assesses Durability and Cross-Language Transfer of Teaching Feedback Classification Protocol
A new benchmark study evaluates the durability and cross-language transfer capabilities of a validated protocol for classifying teaching evaluation feedback. The research re-runs the protocol using updated representation methods, including prompted large language models, and tests its sentiment task transfer to English.
Open-KNEAD: Framework for Agentic Nutrition Estimation from Meal Images
Researchers have introduced Open-KNEAD, a novel framework for estimating meal nutrition from images using an agentic decomposition approach. This system aims to provide accurate portion estimates and traceable records while maintaining user privacy and minimal burden, even for non-US cuisines.
Managing AI Investments in the Agentic Era
OpenAI's latest guidance focuses on how enterprises can effectively manage their AI investments as agentic systems become more prevalent. The core recommendation is to measure 'useful work per dollar' to drive efficiency and scale high-value workflows.
TerraZero: Procedural Driving Simulator for Scalable Zero-Demonstration Self-Play
Researchers have introduced TerraZero, a procedural driving simulator and self-play training stack designed for developing robust autonomous driving agents. It achieves high simulation speeds and generates diverse, safety-critical scenarios by procedurally populating real-world map geometries with randomized elements.
Model Routing Is Simple. Until It Isn’t.
This article explores the complexities of model routing in AI systems, moving beyond simple solutions to address more intricate scenarios. It highlights the challenges and considerations involved in effectively directing AI workloads to appropriate models.
SpectraReward: MLLMs as Zero-Shot Reward Models for Text-to-Image Generation
Researchers have introduced SpectraReward, a novel training-free method that repurposes pretrained Multimodal Large Language Models (MLLMs) as reward models for text-to-image generation reinforcement learning. This approach measures how well an image reconstructs its original text prompt, bypassing the need for preference labels or fine-tuning.
AI Agent Skills for Chinese Knowledge Workers
A GitHub repository, 'skills', offers AI agent skills tailored for Chinese knowledge workers. It includes workflows for iMandalArt, FIRE, planning, and publishing, designed to work with Claude Code, Codex, and general LLM agents.
Multi-Expert Routing for Low-Resource Manchu OCR
Researchers have developed a multi-expert routing system designed to improve Optical Character Recognition (OCR) for historical Manchu documents. This system leverages a lightweight image classifier to route pages to specialized OCR models, addressing the challenge of diverse writing styles and limited labeled data.
PAT: A RAG-Based System for Whole-Document Translation
This research paper introduces PAT (Pragmatic Auto-Translator), a RAG-based system designed to move large language models (LLMs) beyond sentence-level translation. PAT utilizes a comparable corpus of authentic longform texts to inform whole-document translation, aiming to produce draft translations that are contextually appropriate for the target language and culture.