I think that in such case there are in fact multiple OPs. But I don’t have strong feelings about how to name them and if differentiate them at all

Looks like growth on a Petri dish

You think he might have more tricks up his… sleeve?

That was quick. I guess he won’t get charges for obstructing the law, then. AFAIK he’s the first mammal to lay a golden egg

I don’t know much about AP’s insides but couldn’t it be handled in frontend?

Posts would need to have an additional, optional field “Original OP” (or make “author” a list?) and during the action of cross posting it would be filled from the original (or that’s what you mean - the link is not enough, it would require a db lookup?) That would also have an added benefit of being able to choose if dragging OP along is a good idea or not

Pixelfed and Friendica treat language box more rigorously than Mastodon and Lemmy?

I think when it was GPUs it was also crypto. Also, it’s harder for two things to be a bottleneck at the same time

Maybe the current squeeze is because some players decided to set up their farms at the same time. About that I don’t know much

As for the purely technical “why RAM” - when they’re running a model it has to fit in the memory. A model which weights 8GB on the drive requires at least 8GB of RAM (rest for the os etc). AFAIK there have been some papers how to get more out of smaller weight but in general we are at “more is better” part of development of that technology. And probably, because of the hype, having service throttled outside of any plan might be worse RN than getting worse results. So if I were setting up an AI farm and had the money, “the tendency is to push it as far as we can”
And free market is figuring out who is the highest bidder

That’s maintenance and you wouldn’t get all the info in the more partitioned servers

My point is that, unless they are doing that already, companies will probably soon be able to sign a deal with them to get access to models trained on the messages on Discord side. Without requiring the bots to log in or be present in the servers

The problem is threatening the whole Fediverse
(…)
This is a grave abuse of the protocol: messages are silently dropped. It stands against everything the Fediverse is trying to do: allow users to communicate. My experience with open protocols allows me to say that it is a critical problem and that it cannot be tolerated. Would you settle for a mail provider which silently drop all emails you receive if they contain the letter “P”?

Oh, the drama. I think this is a wrong take. The point of ActivityPub is not that we all see everything. Is so the servers can exchange activities. But whether it makes sense to put it in our inboxes or not, depends on what particular application is supposed to do. If using AP would require the application to show everything, what would be the differences between the apps?

I use Threadiverse because I prefer a bit more structured content than what’s happening on Mastodon. While I would love to be able to follow the few accounts and tags on Mastodon that interest me, the fact that I don’t see all that traffic is not an abuse of the protocol

Also, correct me if I’m wrong, since

Pixelfed is designed to mimic Instagram

Then what Pixelfed does is also correct, Instagram also does not show the whole wall from Facebook. Only the pictures, right?

Maybe try something a bit further? Spain or Germany? AFAIK the latter has bigger fedi crowd. Former is just closer to you

We have peertube. AFAIK mostly done by French framasoft and it’s federated

It even has option to mirror yt account but I guess doing so without consent might be a netiquette problem

I suspect they’ve been cooking up their own models. They’ll be selling the data outside of logins

Rolling release means that you won’t be forced to reinstall the whole system when the number after the name flips. And you won’t be locked out of some newer version of a package because the distro you use decided they’re cutting off the updates to “old” versions

If you want something immediate to daily drive and want more of a custom system as opposed to Arch then maybe give NixOS a shot

IMO the main customization part of Gentoo is that you can compile the world without the libs you don’t want to have. With NixOS (AFAIK) being also package-based, how can it offer more custom system than Arch?

Yeah, with time the compile times only get longer and longer

I loved how tailored to me was Gentoo. But as time passes and your hardware gets older, the compilation times get longer and longer. That’s what made me to do the hop

I’ve heard some time ago that now Gentoo is offering more pre-compiled packages. But I don’t know the extent. libstd, gcc and libreoffice were the worst offenders in my time

If you’re going to be compiling your own kernel (or now Gentoo ships with pre-compiled ones too?) my word of advice would be “don’t forget to compile in the filesystem support”

I haven’t done that myself but from https://pimylifeup.com/pangolin-linux/ I understand that will only be subdomain to access pangolin dashboard

how do I keep the vps.domain.tld while also allowing for pangolin.domain.tld to be valid at the same IP?

Domains are just translation from name to IP. What gets served on which subdomain is then handled by nginx or traefik. AFAIK you can have all 3 (VPS, pangolin and root) to point at the same IP

I’ll probably get eaten here but here goes: I do use LLMs when coding. But those should NEVER be used when on unknown waters. To quickly get the 50 lines boilerplate and fill out the important 12 - sure. See how a nested something can be written in a syntax I’ve forgotten - yes. Get some example to know where to start searching the documentation from - ok. But “I asked it to do X, don’t understand what it spewed out, let’s roll”? Hell no, it’s a ticking bomb with a very short fuse. Unfortunately the marketing has pushed LLMs as things one can trust. I feel I’m already being treated like a zealot dev, afraid for his job, when I’m warning people around me to not trust the LLMs’ output below search engine query

To me the idea of temporality of communities and no instances is interesting. It’s definitely something new

If I use something like Context Free to create something, would that count? If you paint or take a picture, but apply some filters / effects?

No, because what we see is not a randomized output based on statistics of previous content