Every Go application since 1.24 will activate MPTCP on their sockets by default, so it's in fact quite widely used.

The fact that simply opening the store gives you a faceful of AI slop (12 out of 15 apps on the front page), and the editor picks are almost exclusively slop or partially slopped (one application without explicit AI code in it), it does paint a pretty clear picture of them allowing more of it compared to flathub.

So; flathub, but with less polished packaging, and more AI slop.

So, this is basically like flathub except filled with more AI slop?

Router-wise, I can throw in a recommendation for the Turris devices by CZ.NIC. (MOX, Omnia, Omni-NG)

They're using their own Turris OS, which is built on a clean OpenWrt base, they don't upload telemetry or display ads of any kind - but they do have a multi-step opt-in if you wish to run honeypots and/or push data to the Sentinel distributed firewall research project.

Managed to get into the queue for a 1TB here in EU, first bit of proper luck I've had with the Valve hardware so far

I dunno, that looks exactly like something you'd expect to see an Analyst/Therapist - an analrapist if you will - use.

Been using Yggdrasil for quite a while at this point, both for encrypted service communication and also as unbreakable VPN links to various places where I don't want to have to mess about with NAT traversal.

It's been nice to be able to just drop it onto my router, add an ALFIS DNS entry, and have all my devices just work with Yggdrasil with no additional configuration.

No, the Rust standard library only contains functions useful for standard Rust, not POSIX/Linux user-space -specific functionality from libc.

Because it's a rust implementation of coreutils, and not a rust implementation of coreutils and libc?

Well, Valve did a randomization of the order queue. So your account value really shouldn't matter

I'm holding off on Fluxer until they decide how they're going to implement federation, since the designs they've communicated publicly so far have all seemed like they prioritize siloing and putting excessive load on self-hosted nodes.
Their first proposed solution would've required each self-hosted server to be able to handle every user on every other server in the network - a proposal which they've since scrubbed from their page.
The latest proposal I can find at least speaks about aggregating connections through the users server, so it's not as insane (Only requiring each self-hosted server to be able to handle requests from every other server on the network). But it still forbids intelligent caching, and instead seems to consider recommending the use of cloudflare to reduce the load from their design to be a good solution.

I made an NFC/QR code card for my wifi details at home, and that also includes the details written in clear text on the card, since some devices simply don't do QR (or NFC in my case).

Someone pulled some DRAM cost graphs, and with the >500% increase that's happened, the Deck did have to rise at least $200 to remain viable.

We really can't have nice things, first crypto and now AI insanity is really doing its best to remove computing as an affordable hobby.

It doesn't really solve my use-case, since we do a combination of direct stream when possible, as well as local sync/cache but only for transcoding purposes. As well as local users only, with the federated sharing taking care of permissions, so that we don't transfer user information between eachother.

Jellyswarrm is certainly a lot closer than other solutions people have suggested though.

The lack of library sharing is what's keeping me on Plex.

The ability to watch media across servers, and even transparently pull it for local transcode when necessary is just an amazing feature.

And it's not even just the people traveling here from the future trying to do it, it's also travelers from the past.

Det är värt att notera att upphovsrättsfri kod inte alls behöver vara open-source, för det är ett begrepp med djupare betydelse än bara att den underliggande koden går att få tag i.

Sen så är ju ett av de största problemen med AI-generering specifikt runt datakvalitén, något som man vanligtvis inte hade något problem med i traditionell maskinlärning, eftersom dataset:en som användes för det var väl kuraterade. Om det fanns minsta chans till dåliga värden i traditionella dataset:et så kastades det och byggdes om, eftersom man ville få kvalitativa resultat.
Inte som dagens AI som istället desperat lärts med allt som kunnat skrapas från internet.

Moderna varianterna här verkar mer fokuserade runt idén att dränka skräpet i mängden data. Men det finns inte något sätt att i efterhand ta bort input-data ifrån en modell när det väl lärts in i den, utan att behöva rulla tillbaka hela modellen till innan det lärdes in. Så allt vi kan veta idag är att alla publika AI modeller garanterat har skräp i sin lärodata, och därför också garanterat kommer generera skräp. Det enda vi inte är lika säkra på är procenten av det de genererar som kommer vara skräp - eller baserat på skräp.

Jag har sett folk kunna använda AI för att generera data som inte är felaktig, men det har också krävt att personen i fråga har nog med kunskap i ämnet för att veta vad som är skräp och vad som är kvalitativ data, något som verkligen inte är en garanti.
Speciellt på projekt som försöker förklara sig som FOSS när de är till huvuddelen AI-genererade så lämnar det här en extremt dålig smak i min mun, för det lutar extremt åt hållet att personen som rattat AI:n inte vet vad de håller på med.

Jag tar mycket hellre och kör genererad kod om personen som tillhandahåller den också noterar att den är genererad och därför saknar garantier, samt inte bör försöka läsas av människor eller användas för FOSS verksamhet, för det antyder iallafall att de vet hur verktyget de använt fungerar.

More expensive than the base-level Xbox/PS controller, but cheaper than their pro-version - as well as the Joy-Con 2s.

I'm down for it, nobody else is doing trackpad controllers after all.