The AI features are actually pretty cool!

Using a local AI model (running on your own GPU), you can:

• Drag files on to the "AI bar" to bring them into the context of a prompt.
• Tell it to perform actions like, "merge these two PDF files."
• Use it to search your filesystem, "find pictures of spiders."
• Ask it to find system settings or tell it to change them, "turn on dark mode."

It supports voice and probably visual (e.g. with a webcam) I think.

Best of all: It doesn't send any of that to some data center in the cloud! I mean, you can configure it to do that but you can just as easily use say, qwen3.5.

Note: It's not realistic to expect to be able to use local models if you have less than 16GB of VRAM (in your GPU). I mean, some 8-billion parameter model will work with say, 8GB but you're not going to be satisfied with the results most of the time 🤷