continuing to love using @herdrdev . You can simply use as a tmux for dummys. But i've just had a galaxy brain moment after realising there is a herdr skill that lets your agent control herdr so you can delegate jobs to another agent in seperate tab ๐ง
THIS CHANGES EVERYTHING
Its weird how I know when a post is 100% AI written too. Subtle tells like the forced enthusiasm.
"this is exactly what they needed"
Maybe its cos I tried to curate a my newletter for a while with AI for a while, it just churns out the same style of copy over and over again...
Microsoft is so back! This is the Surface Dev Box powered by NVIDIA RTX Spark.
Microsoft just turned Surface into a serious local AI workstation brand.
The new Surface RTX Spark Dev Box is not a regular mini PC. Itโs a Windows 11 Pro developer machine built around NVIDIAโs RTX Spark superchip, with a Blackwell RTX GPU, Grace CPU, up to 128GB unified memory, and up to 1 petaflop of AI compute.
That means you can run 120B+ parameter models locally with a 1 million token context, fine-tune models that usually need cloud GPU instances, and build agentic workflows without paying for every single cloud call.
The stack is also very real: WSL 2 with GPU passthrough, CUDA support, VS Code, GitHub Copilot, Git, Python, Node.js, AI Toolkit for VS Code, Windows ML with TensorRT, Copilot Runtime, and Microsoft Foundry.
This is exactly what Windows needed.
Not another thin AI PC with an NPU sticker, but a proper desk-side developer box for local-first AI work.
Surface finally has a machine for people building the future, not just consuming it.
Microsoft is so back! This is the Surface Dev Box powered by NVIDIA RTX Spark.
Microsoft just turned Surface into a serious local AI workstation brand.
The new Surface RTX Spark Dev Box is not a regular mini PC. Itโs a Windows 11 Pro developer machine built around NVIDIAโs RTX Spark superchip, with a Blackwell RTX GPU, Grace CPU, up to 128GB unified memory, and up to 1 petaflop of AI compute.
That means you can run 120B+ parameter models locally with a 1 million token context, fine-tune models that usually need cloud GPU instances, and build agentic workflows without paying for every single cloud call.
The stack is also very real: WSL 2 with GPU passthrough, CUDA support, VS Code, GitHub Copilot, Git, Python, Node.js, AI Toolkit for VS Code, Windows ML with TensorRT, Copilot Runtime, and Microsoft Foundry.
This is exactly what Windows needed.
Not another thin AI PC with an NPU sticker, but a proper desk-side developer box for local-first AI work.
Surface finally has a machine for people building the future, not just consuming it.