We explore why AI agents for data analysis often provide incorrect answers and how a special context layer helps fix this.
AI: Events
How Robots Learn Precise Movements: Online Reinforcement Learning from Physical Intelligence
Research
Physical Intelligence has introduced an approach for teaching robots precise manipulations using online reinforcement learning directly during interaction with the environment.
AI: Events
Databricks Launches Cloud Access to NVIDIA GPUs – No Server Setup or Infrastructure Management Required
Infrastructure
Databricks has introduced AI Runtime – an environment for training and fine-tuning models on NVIDIA GPUs without the need to deploy your own infrastructure.
JetBrains has introduced Tracy, an open-source library for Kotlin developers that helps monitor the behavior of AI applications under real-world operating conditions, offering insights into their performance and issues.
MiroMind has released MiroThinker 1.7 H1, a research AI agent capable of solving long, multi-step tasks with verification at each step.
AI: Events
How AI Agents Work Directly Inside a Repository: Breaking Down the Squad Approach
Development
GitHub explained how the Squad system works – an AI agent coordinator that operates directly within a repository and makes multi-step tasks transparent.
AI: Events
Gemini in Google Sheets: AI Assistant Now Works with Data at a State-of-the-Art Level
Products
Google has announced new beta features for Gemini in Sheets: the AI assistant can now create, edit, and analyze entire tables based on a text prompt.
GitHub has shared its approach to securing agentic workflows, relying on isolation, limited permissions, and detailed logs to ensure AI agents operate safely.
AI: Events
On-Device Voice AI Agents: How PyTorch is Building a Unified Platform for Voice Tasks
Infrastructure
PyTorch has introduced an approach for running voice AI agents locally on devices – without the cloud, with support for multiple platforms and real-time tasks.