Daviluri focused on the technical details of agents, including Microsoft Execution Containers, which is moving to general availability. These are sandboxes that are integrated with the OS to decide what data is shared with agents.
Windows will also require agents to attribute their actions to themselves. There will be IT actions for enterprises that want to set agentic rules once and use them across their fleet of devices. Microsoft's Divya Venkataramu demonstrated this system, showing how she had access to documents even when agents didn't.
Davuluri also described hybrid intelligence, a mix of agents running locally and in the cloud to reduce token budgets. Microsoft's HydraFusion, which balances local and cloud agents, is coming from GitHub to Windows on October 15 for GitHub Copilot, the CLI, and VS Code, and will use new models like MAI Code 1.1 Flash and Nvidia's Nemotron. Using advanced quantization, DeepSeek V4 Flash can be used at 1.6 bits to run locally at 60GB of memory, he said.
Windows ML will serve as the runtime to deploy models across GPU, NPU, and CPU. Llama.cpp will come to Windows ML to give developers chances to experiment with new models at launch.
Microsoft's Kayla Cinnamon demonstrated Hybrid intelligence on a Surface RTX Dev Box by prompting GitHub Copilot to detect unused codes, packages, and feature flags to clean up code. With an Efficiency mode, Copilot selects a model automatically based on how complex a task is. A more complex prompt used GPT-6 Luna, while separate sub-agents ran locally.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
Andrew E. Freedman Social Links Navigation Senior Editor Andrew E. Freedman is a senior editor at Tom's Hardware focusing on laptops, desktops and gaming. He also keeps up with the latest news. A lover of all things gaming and tech, his previous work has shown up in Tom's Guide, Laptop Mag, Kotaku, PCMag and Complex, among others. Follow him on Threads @FreedmanAE and BlueSky @andrewfreedman.net . You can send him tips on Signal: andrewfreedman.01
kealii123 "We've reinvented the personal computer completely" translation: Nothings changed. "The system used a mix of local and cloud models" translation: it's almost certainly 99% cloud models Reply
Bigshrimp I have a feeling that these won't sell well. I mean just look at those prices and there is no real innovation either, just more AI slop on top. Reply
bit_user Really curious whether they're going to use SOCAMM2, or just soldered RAM. Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/laptops/SPONSORED_LINK_URL
- https://www.tomshardware.com/laptops/microsoft-and-nvidia-launch-surface-laptop-ultra-with-rtx-spark-rtx-spark-preorders-live-now-coinciding-with-major-windows-11-changes-for-agentic-ai#main
- https://www.tomshardware.com/membership
- Our favorite budget gaming chair just hit an all-time low of $189
- Grab a SteelSeries Apex Pro keyboard at up to 50% off — save up to $120 on one of the best gaming keyboards you can buy today
- Researcher develops method for fingerprinting cheaters using Counter-Strike mouse and keyboard input patterns
- From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI
- New Linux tech compresses memory in RAM, as RAM, for 452x speedup
Informational only. No financial advice. Do your own research.