To maximise the VRAM available for an LLM on the same machine. That's why I asked myself the same question, anyway.

An MCP server lets you avoid giving the agent your API key so it can't leak it. At least in theory.

You could do the same with a CLI tool but it's more of a hassle to set up.

I find it really weird that it's not until you get to the $200 range that anything has USB-C DP Alt mode support. You'd have though that anyone trying to squeeze space would want to drop full-size HDMI or DisplayPort ports at the first opportunity.

I really want a small SBC with USB-C DP Alt mode that I can stuff into a ~60%ish mechanical keyboard-sized case to make into a headless laptop thing so I can justify buying some display glasses like the XReal One or similar. Seems like it would be the ultimate travel computing solution.

That's precisely what I want. I've got all the parts other than the SBC. Even designed my own keyboard with this in mind. You can do it with a pi zero 2 but going from mini-HDMI to full-HDMI through an adapter to USB-C to only then pop out too the XReals is a complete mess of cables and little boxes, not least of which is caused by the HDMI->USB-C adapter needing to be independently powered. I'm tempted to print up a case to hide the mess but it's still, you know, a mess.

Also the zero 2 is a little under-powered. It's fine as far as it goes, it just feels like there's probably performance on the table with a more modern chipset.

I switched from SBC to thin clients. You get the Dell wyse 5070 for 50 euros in my country and it has USB c dp alt mode and much more.

I will note that in theory there's the Radxa Cubie A7S, but that's unobtainium. Or I'd have bought three by now.

I definitely get the impression there's something not quite right with qwen3.5 in llama.cpp. It's impressive but just a bit off. A patch landed yesterday which helped though.

Which patch are you referring to?

In some ways the handholding is the point. The way I used qwen2.5-coder in the past was as a rubber duck that happens to be able to type. You have to be in the loop with it, it's just a different style of agent use to what you might do with copilot or Claude.

It failed to detect me at all.

Same here.

If you haven't already, give the models that Handy supports a try. They're not Whisper-large quality, but some of them are very fast.

Charging in and DisplayPort out on the same socket would mean an additional dongle or hub or something, so there's at least that reason for having both.

Couldn't one charge from the display connected to?

Both ports support power delivery, so you can still be plugged in and use DisplayPort out.

Not necessarily - you just need a display that has USB-C input and supports USB-PD.

The camera on the Surface is nowhere near as good as on my M1 Macbook Air, either. That seems to be a weird blind spot on laptops in general, it's very obviously an afterthought on my personal Dell XPS as well.

I'm still working on an 8GB M1 Pro. It's just about ok. VS Code plus podman plus Teams plus Slack plus Firefox and it hits the limits; usually Slack is the thing to get killed.

I’m on a 16GB M1 Pro. Slack, Zoom, then browser tabs for Jira, PagerDuty, GMail, Confluence, and a Google doc and it’s swapping like mad. It’s that fast SSD and the integrated memory, so it swaps quickly. Swapping is still swapping though.