I don't know where you've got those numbers, but they're wrong.

https://www.reddit.com/r/LocalLLaMA/comments/1n79udw/inferen... seems comparable to the Framework Desktop and reputable - they didn't just quote a number, they showed benchmark output.

I get far more than 3 t/s for a 70B model on normal non-unified RAM, so that's completely unfeasible performance for a unified memory architecture like Halo.

Did you casually glance at how the hardware in the Framework Desktop (Strix Halo) works before commenting?

A Framework Desktop exposes 96GB of RAM for inference and costs a few thou USD.

There is an open-source app ("Casio G-Shock Smart Sync") for syncing time with g-shock BLE watches.

Some of them are also compatible with Gadgetbridge (also OSS, but much richer in functionality).

The term to search for is "UKI".

A UKI is a kernel+initramfs+boot-arguments bundle all as a single WinPE/UEFI executable using the "EFI Stub Loader".

You configure your system firmware to execute it, passing no arguments. It boots using the command line you set earlier. It's signed, and verified by the platform secure boot.

Hibernation works fine with this approach.