> Many many idle GPUs.
Yes, but what use will they have? Rail lines and dark fibre could be used post-bust (and don't depreciate all that quickly).
26,498 karma · joined January 1, 2017
Yes, but what use will they have? Rail lines and dark fibre could be used post-bust (and don't depreciate all that quickly).
What are the civil works costs for a small(er) reactor versus a large(r) reactor?
I started using Phoenix at version 0.3 (compiling the port on my FreeBSD then-daily-driver) and still use Firefox (and Tor Browser) today. I'm not sure what problem(s) people have with it, but it works for me.
(I also use Safari on my Apple systems, to further diversify web browser usage.)
How many organizations and individuals use, and depend on, those services for their livelihood?
People seem to have forgotten the bust in railroads and Internet (etc), even though those technologies were transformative:
* https://en.wikipedia.org/wiki/Technological_Revolutions_and_...
One thing I'm wonder about: previous busts left useful infrastructure behind (rail, fibre) that could still be used. If/When the AI/LLM companies go bust, what will be left over afterwards and how useful will it be?
Why is the "scrap of paper" a problem now after they've been around for several decades (as has Photoshop and colour printers). It's a legal document: if the cop thinks it's fake they can call the insurance company and confirm.
In Georgia, where one of the incidents took place, the cops can lookup insurance status in real-time:
> The GEICS system works through required insurer reporting. Every insurance company doing business in Georgia must notify the state when a policy is issued, renewed, canceled, or lapses. Companies report these changes within 30 days. The Department of Driver Services maintains this data and makes it available to law enforcement statewide.
* https://www.bubbahead.com/blog/can-police-pull-you-over-for-...
The book discusses expectations and 'standards of performance' and what is reasonable.
The system / cards is not about keeping score: each card is a task, and whomever holds the card is responsible for it from start to finish. E.g., three of the cards are "Meals (Weekday Breakfast)", "Meals (Kids; School Lunch)", and "Meals (Weekday Dinner)"; this may fall to one person, but there is also the "Groceries" card:
> Your family needs to eat. Can we all agree on that? Good. So how do the pantry and refrigerator get stocked on an ongoing basis? With CPE. This includes noticing what’s about to be used up or expire, keeping a running grocery list, and getting to the market . . . ideally before you realize there’s not a drop of milk in the fridge. Even if your family orders groceries online and your go-to dinner is frozen pizza, this card requires CPE nearly 365 days a year. Consult the various “Meals” cardholders regularly so you can be sure you have taco shells in the cupboard for taco night.
* https://fairplaypolicy.org/the-cards
So the Meals person(s) have to plan the meal(s), and make an ingredient list for those specifically, and add to to a grocery list, but grocery buyer needs to take stock of what is already present in the fridge/freezer/pantry and buy anything missing generally, and specifically for the meals planned.
You can find quite a lot of data showing that the burden of managing domestic matters falls primarily on women. I would recommend people check out the book Fair Play which examines the problem and puts forward a system where people are assigned various personal tasks that need to be done (there are 100 in total, not all of which may apply to your life/household):
> The result is Fair Play: a time- and anxiety-saving system that offers couples a completely new way to divvy up domestic responsibilities. Rodsky interviewed more than five hundred men and women from all walks of life to figure out what the invisible work in a family actually entails and how to get it all done efficiently. With four easy-to-follow rules, 100 household tasks, and a figurative card game you play with your partner, Fair Play helps you prioritize what's important to your family and who should take the lead on every chore from laundry to homework to dinner.
* https://www.goodreads.com/en/book/show/44071899-fair-play
* Interview: https://www.youtube.com/watch?v=u6FfxfRMQkw
A large golden parachute for firing a board member could be a disincentive mechanism to do it: a 'poison pill' of a kind.
* https://en.wikipedia.org/wiki/Shareholder_rights_plan
They could have felt they were doing the right thing, and making a public statement of the situation, but knew it was a bit of a kamikaze tactic and so made it more painful for when the eventual backlash occurs.
In Canada, Ontario and Quebec have their own province-level police forces, but the rest of the provinces 'contract' the RCMP to do street-level policing in many areas (with larger municipalities having their own forces). Part of the reason for this is that with any organization there's overhead (training, HR, payroll/pensions, etc), and for smaller communities a larger portion of a fixed budget would go to those things, so by leveraging the RCMP those fixed costs are amortized wider.
In Ontario/Quebec, many smaller communities use the OPP/QPP for local policing (larger cities have their own forces) for the same reason.
For the community in the article, is there a reason why the same 5 cops couldn't become employees of the county's force? For really small counties, it may be necessary / useful to 'contract' the state police to run a local detachment.
See also Ben Felix:
* https://www.youtube.com/@BenFelixCSI/videos
Puts (DOI) links to the peer-reviewed papers he cites. And he often gets the authors of those papers on the podcast his co-hosts:
From his uni bio:
> Patrick Boyle is a Visiting Professor of Finance at Queen Mary University of London in the School of Economics and Finance. He has taught at Queen Mary University since 2009. Patrick teaches the Financial Derivatives and Portfolio Management Classes to Master’s in Finance Students.
> Patrick Boyle is a Founding Partner at Palomar Capital Management, a quantitative Hedge Fund. Patrick began his finance career in 1997 and has worked with Victor Niederhoffer, a former partner of George Soros' and has been a Portfolio Manager at RBS, Millennium and Nomura.
[…]
* https://www.qmul.ac.uk/sef/staff/patrickboyle.html
So the university itself vouches for the academic side of things, and the UK business registry has a "BOYLE, Patrick" for Palomar Capital Management (resigned as of 2018):
* https://find-and-update.company-information.service.gov.uk/c...
The users can develop on >100 compute nodes, but choose not to bother doing any kind of forwarding/proxying/jumping to them and just do stuff on the login nodes.
If they can't be bothered to do a "ssh -J …" then it's on them. The resources are there.
We put the MOTD up ≥7 days in advance and put it in relevant Slack channels.
We still get "Is Foo down?" the day of.
Probably why both soft and hard limits were developed.
There will be a system outage 2026-11-11 (08:00 ET) to 2026-11-13, Wed to Fri.
And yet we still regularly loads of >100 on our 64 core HPC login codes, and swap is regularly used even with 96G of system memory (we have per UID memory limits too).
What's hostile is the VSCode (and Codex and Claude) makers developing tools that basically DoS a system because they assume it will operate only on single-user machines.
(And WTF are you doing that you're forking 256 processes? We have quite a few expensive HPC nodes: use those to build, not the damn login nodes.)
It also (AIUI) tries to walk the entire file tree, so have fun with NFS (auto)mounts.
It also amounts to letting off fork bombs: we set up limits for a maximum of 256 process per UID, and regularly get folks asking "what does this 'cannot fork' message mean?": it mean you're trying to DoS the system.
No? When I'm asked about Apple Pay on an iPhone, it says "Maybe later". In the Settings it has reminder, and if I go in and say "No" it never pops up ever again (maybe on a iOS major version upgrade?).
Generally with Apple it's one "Maybe later" with a reminder flag set someplace, and then that's it.
macOS has a 'service (priority) order':
* https://sixcolors.com/post/2025/05/service-please-manage-mul...
It's no different that having multiple default gateways in Linux (with different metric/priorities):
* https://superuser.com/questions/1600190/change-order-of-rout...
And when your default route changes so will the TCP connection end-points. This is mostly an issue when you have a stateful session, like SSH, and less so with 'bursty' traffic like web surfing or e-mail.
By default wired connections are higher priority. I've been on Wifi with a MacBook and patched in with the Network Settings panel open and see the change, which is reflected in the output of "netstat -rn". Underneath the hood most of the macOS kernel is (Free)BSD/Unix.
Until we get SCTP multihoming popularized this will generally be an issue on any operating system. Though it would perhaps be nice to perhaps have a socket (TCP) API flag to tell the kernel "keep me on the initial interface as long as it exists" so sessions don't flip over.
> and have enough Linux know how to fix the occasional hick up
A Linux sysadmin teammate of mine has >10 years of running Linux servers, and chose to install Ubuntu on his work Dell laptop, and in the ~2 years he's been running that thing he's had regular (2-3/month) issues with audio during Zoom, Teams, and Google Meetings sessions. Both Bluetooth and wired headset connections.
At my last job (pre-COVID) I had a Linux workstation at my desk and a Windows laptop. At where I'm currently at we're hybrid and so we get laptops and I chose a MacBook because I knew I could both do my daily sysadmin chores through Terminal and also have access to 'business apps'.
I've been using Linux since the a.out days and MacOS X since the PowerPC days, and I find MacOS less of an ordeal for day-to-day tasks as my terminal/workstation. I sysadmin Linux (and previously also Solaris and IRIX) servers for a living. It's been several years since I had to daily drive a Windows machine (though I've heard PowerShell and WSL2 are pretty good).
I'm not sure what "ordeal" you are referring to. I'm mostly in Terminals SSHing places, so perhaps my work use cases are not 'advanced' enough to bump into them? (I also use macOS for my personal systems.)
"Consumers" can be running it just as much as "consumers" could be running Raspberry Pis, which could also be considered 'embedded' by some. SPARC is no less (or more) proprietary than ARM: you can license the ISA, fab a CPU, and sell some kind of hardware stack.
Do you think it would be more or less difficult to get a license from Intel/AMD for x86(_X64) than ARM and/or SPARC so you could fab your own chips?
Yes:
* https://en.wikipedia.org/wiki/SPARC
Fujitsu was selling Solaris-compatible systems (as were many others). Linux and BSDs also ran on them.
There were a number of x86 makers up until the 1990s or so, but who is left of them? Can you still license x86(_x64) to fab your own CPUs? You can with SPARC:
[…]
> Citizen Lab at the University of Toronto called the bill’s data-retention requirements, which would mandate electronic service providers keep detailed data of nearly every Canadian for a year, revealing their movements and who they contacted, “almost certainly unconstitutional.”
> Major tech companies such as Signal, which provides secure messaging, and DuckDuckGo, a privacy-focused search engine, have said they would withdraw from Canada if they were forced to reveal or retain user data that compromises users’ privacy.
When one of your Proxmox (or XCP-ng, or VMware/Nutanix/OpenStack) servers won't boot and you need go into single-user/rescue mode, then talk to me about the convenience or not of a remote KVM (versus having to drive to the DC and hook up the crash cart).
It's all very well if all you manage is VMs/instances, but some of us deal with the hardware that those VMs run on, and things break.
That 0.01% of servers is probably the layer that the other 99.99% run on (and yes, a lot of that 0.01% can also be automated, but if PXE isn't working (or you're getting strange CPU3VOLT or SLOT17ECC or BUS1RESET errors) then neither is re-imaging).
I currently admin several racks of Dells (and Lenovo, and SuperMicro, and previously HP). Just make sure you buy the license for KVM in case you ever need to go into single-user/rescue mode (otherwise better not mind having to drive to the DC and hooking up a crash cart).
Ask @bcantrill (the author of this article/story) about how well integrators integrate (as a hint to the answer, he co-founded Oxide Computers because he ran a cloud service with Dells/HPs/etc).