Field Notes

Agents for Google Home, Claude unified, NVIDIA leads MLPerf, DeepMind launches DMI

Today with concrete agent interfaces, product bundling, hardware signals, and policy pressure.

Top stories

Google opens up Google Home to agents via an MCP server. In early access, Claude, ChatGPT, and others can control devices, fetch camera summaries, and read out activity by voice. If you build agents, you can connect them now. Only grant the permissions they truly need.1

Anthropic merges Claude Chat and Cowork into one product. At the same time, Claude Docs and Slides arrive for documents and decks straight from chat. You find conversations and work in one place and export results as usual.23

NVIDIA reports top scores in the MLPerf Inference v6.1 debut with the Vera-Rubin NVL72. The focus is on system performance, efficient scaling, and software optimization to lower inference costs. The figures come from the company.4

Emerald AI, Google, and NVIDIA found the AI Energy Management Alliance. The coalition wants to couple AI data centers more flexibly to the power grid. The founders say it is the first alliance of its kind. Concrete programs or tools are still pending.5

DeepMind launches the DeepMind Institute for interdisciplinary questions around AGI. It bundles research and perspectives across disciplines.6

In Washington, political opposites align to rein in AI. From Bernie Sanders to Steve Bannon, there are voices for it. If you work in the US, expect tighter rules.7

Tools and releases

OpenAI lays out a framework for reporting model misbehavior and publishes six cases, including unintentional file uploads.89

Reports say Apple is working on an enterprise AI server with its own M8 Ultra chips and is considering a partnership with Nvidia. This is not confirmed.1011

Mozilla and Mistral announce a browser assistant. Smart Window is based on Mistral models and brings open, private, multilingual features into the browser.1213

Microsoft introduces 365 G7 for government, with new AI features for teams in administration and field work.14

According to NVIDIA, TensorRT Edge-LLM runs 6.4 times faster in the MLPerf Edge Agentic Benchmark on Jetson AGX Thor.15

Research

TypeSafe AI presents a model that evaluates options instead of generating text. The goal is more stable decisions.16

For health and life sciences, developers release 38 open agent skills across 11 domains to reduce policy misapplication.17

Sources

  1. Your AI agents can now control your Google Home devices (techcrunch.com)
  2. Anthropic merges Claude Chat, Cowork, and more into a single product (the-decoder.com)
  3. Claude comes for Gemini with its own take on Docs and Slides (theverge.com)
  4. NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference v6.1 Debut (blogs.nvidia.com)
  5. Emerald AI, Google and NVIDIA Launch Alliance to Advance Flexible AI Data Centers (blogs.nvidia.com)
  6. Google Deepmind launches interdisciplinary institute to tackle the big questions around AGI (the-decoder.com)
  7. Political opposites unite in Washington to rein in AI (the-decoder.com)
  8. OpenAI Creates a New Framework to Disclose Bad AI Behavior (wired.com)
  9. Our framework for reporting model misalignment (openai.com)
  10. Apple is reportedly building an enterprise AI server with its own M8 Ultra chips (the-decoder.com)
  11. Apple might make servers again to cash in on the AI rush (theverge.com)
  12. Mozilla's new Smart Window assistant runs on Mistral's models (the-decoder.com)
  13. Mistral and Mozilla are bringing open, private and multilingual AI to your web browser (mistral.ai)
  14. Introducing Microsoft 365 G7: AI tools for government teams (microsoft.com)
  15. TensorRT Edge-LLM Completes the MLPerf Edge Agentic Benchmark 6.4x Faster on Jetson AGX Thor (developer.nvidia.com)
  16. Former OpenAI researcher builds an AI model that judges options instead of writing text (the-decoder.com)
  17. Improving HCLS AI reasoning with open-source agent skills (aws.amazon.com)