It's not Broadcom so it is supportable by OpenWrt. It runs Qualcomm IPQ5332 SoC which recently got supported in upstream linux ath12k drivers... There are a few patches to make it work and the only blocking issue is with the Realtek switch connected to the LAN ports.
Mainly because Microsoft wants to have "connected standby": the CPU is running in a low power mode (not powered off like "old" S3 sleep), can be turned on periodically and can turn on other devices even when the computer is "sleeping".
My Zen2 based Lenovo laptop has 6-7 hours of battery when doing basic tasks in both Windows and Linux, but sleep on Linux lasts a week while on Windows it's empty in 24 hours.
And it can fail in great ways. Last example: I asked claude for a non-trivial backup and recovery script using restic. I gave it the whole restic repo and it still made up parameters that don't exist in the code (but exist in a pull request that's been sitting not merged for 10+ months).
SNI routing is such a bad way to do what should be L3 problem that people implemented PROXY protocol to send information about user's endpoint address without doing MITM.
All desktop/mobile OSes today use "Stable privacy addresses" for inbound traffic (only if you are hosting something long-term) and "Temporary addresses" for outbound traffic and P2P (video/voice calls, muliplayer games...) that change quickly (old ones are still assigned to not break long-lived connections but are not used for new ones).
And the worst part about CGNAT is that you have two bad solutions:
Either EIM/EIF (preferably with hairpinning) where you can practically do direct connections but you have to limit users to a really low number of "connections" breaking power users.
Or EDM/EDF where users have a higher number of "connections" but it's completely impossible to do direct connections (at least not in any video/voice calling system).
.docx and .xlsx are also just zip files with XML and attachments. The bad thing is that the XML is Word's internal document structure serialized and behavior for some values is only defined in Microsoft's code.
This only works for one SSID. Even then, one thing that can mitigate this is using Private-PSK/Dynamic-PSK on WPA2, or using EAP/Radius VLAN property.
On WPA3/SAE this is more complicated: the standard supports password identifiers but no device I know of supports selecting an alternate password aside from wpa_supplicant on linux.
You can't think all the way about refining your prompt for LLMs as they are probabilistic. Your re-prompts are just retrying until you hit a jackpot - refining only works to increase the chance to get what you want.
When making them deterministic (setting the temperature to 0), LLMs (even new ones) get stuck in loops for longer streams of output tokens. The only way to make sure you get the same output twice is to use the same temperature and the same seed for the RNG used, and most frontier models don't have a way for you to set the RNG seed.
https://storage.googleapis.com/play_public/supported_devices...