nyarlathotep_
Joined 4 years ago
Joined 4 years ago
This has always been my core Linux desktop woe--there's always, no matter what CPU GPU (including (and most often), no dGPU at all) combo/distro I've used, been sleep wake issues of some variety.
I've had AMD CPUs with both fedora and Ubuntu that would never sleep if they weren't woken to desktop (i.e if not logged in, then allowed to automatically turn off the display) they'd never sleep. On those and any of the machines since, I never get more than a month of uptime without it being unresponsive when trying to wake it or similar, regardless of if Kernel updates have been installed etc.
I watched some of his streams on the development of the JIT compiler (what this change set seems to replace) on one of his streams, and was blown away by the quality of Copilot generated suggestions/completions on an (under-represented) domain (x86 opcodes, etc). This was before Claude Code et al, and pretty early-on for LLM tooling.
In the decade+ since, there's no way I'd do so.
I know three personal friends that are long time (allegedly not addicted) heavy marijuana users that all suffer from Cannabinoid hyperemesis syndrome to the extent that it's effected their daily lives substantially.
Two of the aforementioned are roughly my age, and are half as bright as they were as teens. Neither of them can follow a train of thought particularly well and are difficult to hold conversations with.
How much of this is attributable to how much more powerful "modern" strains (or whatever the geo-engineered differences are) is unknown, but I can't imagine it's not a factor. This is not the dopey "get stoned and play XBox and eat a whole pizza" stuff we had in the 00s.
I'm sure there's plenty of counterexamples or something, but my perspective on this has completely changed, influenced by examples like this.
Worth noting this absolutely impacted usability and stability to a massive degree. The machine ran far hotter to the touch than my personal (equivalent model) MBP, and would make it maybe a month of uptime before it failed to wake from sleep/kernel panic'd/locked up the desktop.
Most other typical desktop software was "vended" via internal software "store" thing (managed browsers, etc), but I could, and did, install various extensions on Firefox (internal Wikis even encouraged using Tampermoney (or whatever the successor is called now) like UBO/Sideberry etc.
Current employer issued machine is a Windows laptop with no admin and basically locked-down.
Even getting something like Docker installed/WSL configured is a whole episode in frustration.
The huge positive is this Enterprise-whatever version of Windows has minimal slop--no CoPilot things or ads in the start menu/lockscreen, but I can't even change the desktop wallpaper. Also, the CPU idles at basically 40% utilization with the various agent things/endpoint security running. For any sort of local development, I largely "sidestep" things by running whatever I need in containers/WSL, so it's really not a huge problem. There's minimal Windows-specific use outside of Teams/Outlook whatever.
Any SOTA model can one-shot something that looks pretty similar to something from Three's examples, but things go south quickly when attempting to increase the complexity, even with pretty unambiguous instructions.
This is frequent, if not constant, on iOS for me. I never witnessed it before the 26 update.
How can it take an entire second or more to display an icon in list in the settings application? It was literally a solved problem for every iOS version I've ever used.
MicroCenter has(had? OOS near me) M4 Minis for $400!
A remarkable bargain, even more so considering the recent hardware price hikes.
IDS is probably overkill for a home network anyway.
I recently replaced said router with a Dream Router 7.
This was all pretty simple to setup given hardware that supports VPN, DDNS, etc.
Worth noting too that just blocking hostnames is not enough; Netflix, as an example, uses Google's DNS servers (plain UDP) on some/most clients.
If the network the client operates on can't prevent DNS to other servers, that's a simple "bypass" vector.
AFAIK, the best you can get, given sufficient time, patience, and hardware is:
• something like the above
• blocking outbound DOT (853), DOQ (784, IIRC) excepting, perhaps, upstreams you trust
• blocking HTTPS to known DOH endpoints (Cloudflare, Google, etc)
• DNAT for plain DNS cases like the Netflix example above. (to your DNS server(s))
Even that there's plenty of hypothetical opportunities for clients to just use another DOH resolver outside of your domain/IP block lists.