And besides Brussels (EU government) and Brussels (national government), both prone to internal disagreements, there’s also the prone-to-disagreement Brussels (provincial government).

https://www.brusselstimes.com/1862716/542-days-brussels-breaks-belgian-record-for-longest-period-without-a-government

‘Tumbling into the abyss’ – Brussels breaks record for longest period without government

Monday, 1 December 2025

As of today (Tuesday), the Brussels-Capital Region has been without an elected government for 542 days – breaking the world record of 541 days, held by Belgium’s Federal Government led by Elio Di Rupo (PS) in 2011.

EDIT: Here we have Brussels (provincial government) being threatened by Brussels (national government) over Brussels (provincial government) going to Brussels (EU government) for funds:

https://www.courthousenews.com/a-city-without-government-how-brussels-keeps-running-despite-its-leaders/

Most damning: Brussels’ auditors have been unable to certify annual accounts since 2017, with the federal Court of Auditors in recent years going from “refusing to certify” to outright “abstaining.” It’s a sign the data is no longer reliable enough to evaluate.

The crisis turned public in Schuman Square, at the geographic heart of EU power flanked by the EU Council and the European Commission headquarters. Unable to cover a 12-million-euro cost overrun on renovations, the Mobility Minister wrote to five European institutions begging for financial help — and was publicly rebuked by Prime Minister Bart De Wever.

“It is truly a disgrace, a total humiliation, the Brussels Region begging for money to build a square,” the N-VA leader said. De Wever, the first Flemish nationalist politician to hold the office of prime minister of Belgium, has escalated pressure with unprecedented threats.

In May, he declared bluntly: “If Brussels comes asking the federal government for money, I will put it under guardianship.” He promised to “impose strict austerity conditions to put a stop to the ‘malgoverno’ in Brussels.”

I use Google’s Noto Sans as my default browser sans-serif font. It does a better job having a different capital-I and lower-case-l than does Calibri:

http://www.identifont.com/differences?first=Noto+Sans&second=Calibri

I once wrote code for an elderly researcher who would only review code as a hard copy. I’d bring him stacks of paper and he’d get going with his pen and highlighter. And I’ll grant that the resolution is normally higher on paper than on most displays. I’m viewing this on a laptop screen that’s about 200 ppi. A laser printer is probably printing at a minimum of 300 dpi, maybe 600 or 1200 dpi.

I still think that the few people reading things in print are the exception that proves the rule, though.

Times New Roman was designed for the print era, and Calibri for onscreen viewing. Onscreen viewing is a lot more common today. Based on that technical characteristic, I’d be kind of inclined to favor Calibri or at least some screen-oriented font.

That being said, screens are also higher-resolution than they were in the past, so the rationale might be less-significant than it once was.

https://en.wikipedia.org/wiki/Calibri

Calibri (/kəˈliːbri/) is a digital sans-serif typeface family in the humanist or modern style. It was designed by Lucas de Groot in 2002–2004 and released to the general public in 2006, with Windows Vista.[3] In Microsoft Office 2007, it replaced Times New Roman as the default font in Word and replaced Arial as the default font in PowerPoint, Excel, and Outlook. In Windows 7, it replaced Arial as the default font in WordPad. De Groot described its subtly rounded design as having “a warm and soft character”.[3] In January 2024, the font was replaced by Microsoft’s new bespoke font, Aptos, as the new default Microsoft Office font, after 17 years.[4][5]

I suspect that the Office shift is probably a large factor in moving to Calibri.

That being said, there are many Times New Roman implementations, but it sounds like Calibri is owned by Microsoft, so I’d be kind of inclined to favor something open.

Brussels had just about had it with Brussels.

goes looking

Of the the last few presidents to die:

https://www.findagrave.com/memorial/4244/ronald-reagan

Ronald Reagan is buried at his presidential library.

https://www.findagrave.com/memorial/4311/george_h_w-bush

George H.W. Bush is buried at his presidential library.

https://www.findagrave.com/memorial/6734/jimmy-carter

Jimmy Carter is buried at his national historical park:

https://www.findagrave.com/memorial/4243/gerald_rudolph-ford

Gerald R. Ford is buried at his museum:

https://thehill.com/homenews/administration/5275798-trump-says-children-could-have-two-dolls-instead-of-30-with-his-tariff-plan/

Trump says children could have ‘2 dolls instead of 30’ with his tariff plan

He added, “You know, someone said, ‘Oh, the shelves, they’re going to be open.’ Well, maybe the children will have two dolls instead of 30 dolls, and maybe the two dolls will cost a couple of bucks more than they would normally.

I don’t really think that talking about Average Joe’s standard of living is really Trump’s strong point.

I’m not that concerned. There’s a famous Reagan joke that he did during a sound test:

https://www.youtube.com/watch?v=kifJ_mQdpZA

monster trucks

“Monster truck”

https://en.wikipedia.org/wiki/Monster_truck

“Road train”

https://en.wikipedia.org/wiki/Road_train

If you mean distributing inference across many machines, each of which could not individually deal with a large model, using today’s models, not viable with reasonable performance. The problem is that you require a lot of bandwidth between layers; a lot of data moves. When you cluster current systems, you tend to use specialized, high-bandwidth links.

It might theoretically be possible to build models that are more-amenable to this sort of thing, that have small parts of a model run on nodes that have little data interchange between them. But until they’re built, hard to say.

I’d also be a little leery of how energy-efficient such a thing is, especially if you want to use CPUs — which are probably more-amenable to be run in a shared fashion than GPUs. Just using CPU time “in the background” also probably won’t work as well as with a system running other tasks, because the limiting factor isn’t heavy crunching on a small amount of data — where a processor can make use of idle cores without much impact to other tasks — but bandwidth to the memory, which is gonna be a bottleneck for the whole system. Also, some fairly substantial memory demands, unless you can also get model size way down.

I’d also note that OP only created their account a day ago.

I wonder how much exact duplication each process has?

https://www.kernel.org/doc/html/latest/admin-guide/mm/ksm.html

Kernel Samepage Merging

KSM is a memory-saving de-duplication feature, enabled by CONFIG_KSM=y, added to the Linux kernel in 2.6.32. See mm/ksm.c for its implementation, and http://lwn.net/Articles/306704/ and https://lwn.net/Articles/330589/

KSM was originally developed for use with KVM (where it was known as Kernel Shared Memory), to fit more virtual machines into physical memory, by sharing the data common between them. But it can be useful to any application which generates many instances of the same data.

The KSM daemon ksmd periodically scans those areas of user memory which have been registered with it, looking for pages of identical content which can be replaced by a single write-protected page (which is automatically copied if a process later wants to update its content). The amount of pages that KSM daemon scans in a single pass and the time between the passes are configured using sysfs interface

KSM only operates on those areas of address space which an application has advised to be likely candidates for merging, by using the madvise(2) system call:

int madvise(addr, length, MADV_MERGEABLE)

One imagines that one could maybe make a library interposer to induce use of that.

I mostly use terminal-based software on Linux.

I think that the only programs I use much that embed a web browser are:

• Firefox

• Steam

• Some games that are Web-based and which I only run one of at once (Neo Scavenger, some RPGMaker-based games, probably some others).

In most respects, I’d rank a five-year-old human mentally above a raven or octopus.

https://en.wikipedia.org/wiki/Cannibalism_in_Oceania

The Korowai tribe of south-eastern Papua could be one of the last surviving tribes in the world engaging in cannibalism.

Sometimes we do.

In the US, looks lile it’s illegal to trade in dolphin meat due to the Marine Mammal Protection Act.

It’s apparently legal in Japan.

I’ve also noticed that is you want a chest smaller than DDD, it’s almost impossible with some models — unless you specify that they are a gymnast.

That’s also another point of present generative AI image weakness — humans have an intuitive understanding of relative terms and can iterate on them.

So, it’s pretty easy for me to point at an image and ask a human artist to “make the character’s breasts larger” or “make the character’s breasts smaller”. A human artist can look at an image, form a mental model of the image, and produce a new image in their head relative to the existing one by using my relative terms “larger” and “smaller”. They can then go create that new image. Humans, with their sophisticated mental model of the world, are good at that.

But we haven’t trained an understanding of relative relationships into diffusion models today, and doing so would probably require a more sophisticated — maybe vastly more sophisticated — type of AI. “Larger” and “smaller” aren’t really usable as things stand today. Because breast size is something that people often want to muck with, people have trained models on a static list of danbooru tags for breast sizes, and models trained on those can use them as inputs, but even then, it’s a relatively-limited capability. And for most other properties of a character or thing, even that’s not available.

For models which support it, prompt term weighting can sometimes provide a very limited analog to this. Instead of saying “make the image less scary”, maybe I “decrease the weight of the token ‘scary’ by 0.1”. But that doesn’t work with all relationships, and the outcome isn’t always fantastic even then.

There are also things that present-day generative AI is not very good at in existing fields, and I’m not sure how easy it will be to address some of those. So, take the furry artist. It looks like she made a single digitally-painted portrait of a tiger in a suit, a character that she invented. That’s something that probably isn’t all that hard to do with present-day generative AI. But try using existing generative AI to create several different views of the same invented character, presented consistently, and that’s a weak point. That may require very deep and difficult changes on the technology front to try to address.

I don’t feel that a lot of this has been hashed out, partly because a lot of people, even in the fields, don’t have a great handle on what the weaknesses are and what might be viably remedied and how on the AI front. Would be interesting to try to do some competitions in various areas, see what a competent person in the field and someone competent in using generative AI could do. It’ll probably change over time, and techniques will evolve.

There are areas where generative AI for images has both surpassed what I expected and underperformed. I was pretty impressed with its ability to capture the elements of what creates a “mood”, say, and make an image sad or cheerful. I was very surprised at how effective current image generation models were, given their limited understanding of the world, at creating things “made out of ice”. But I was surprised at how hard it was to get any generative AI model I’ve tried to generate drawings containing crosshatching, which is something that plenty of human artists do just fine. Is it easy to address that? Maybe. I think I could give some pretty reasonable explanations as to why consistent characters are hard, but I don’t really feel like I could offer a convincing argument about why crosshatching is, don’t really understand why models do poorly with it, and thus, I’ve no idea how hard it might be to remedy that.

Some fantastic images are really easy to create with generative image AI. Some are surprisingly difficult. To name two things that I recall !imageai@sh.itjust.works regulars have run into over the past couple years, trying to create colored car treads (it looks like black treads are closely associated with the “tire” token) and trying to create centaurs (generative AI models want to do horses or people, not hybrids). The weaknesses may be easy to remedy or hard, but they won’t be the same weaknesses that humans have; these are things that are easy for a human. Ditto for strengths — it’s relatively-easy for generative AI to create extremely-detailed images (“maximalist” was a popular token that I recall seeing in many early prompts) or to replicate images of natural media that are very difficult or time-consuming to work in in the real world, and those are areas that aren’t easy for human artists.

You’d think that large drones would be a lot more useful for smuggling via air than balloons.

I assume that the balloons can broadcast their position so that their cargo can be picked up wherever they land, but they aren’t controllable. Drones are.

I don’t believe — without looking at the protocol — that there’s a way to do so, unless you’re an admin on that user’s home instance.

It’s probably possible to find whether a given home instance has any users at all subscribed to your instance, since if that isn’t the case, it won’t be receiving posts to your community. But I don’t believe that home instances expose subscriptions.