pict-rs caches images locally.

I’d think most trolls would pass up on doing something that can easily get them imprisoned if ever found out …

It’s not so much their death, but that they won’t create more art*. Just a short while ago I was watching the invincible special, it had Lance Reddick in it. I don’t care about him, because I don’t know him, but I know that he played many roles I loved, and now that he’s dead, that will not happen again, and it makes my life just a little bit emptier.

Edit: * actually, not just art. Politicians like Nelson Mandela could, just by talking, improve the world a little bit. Genius scientists who do science. All that stops. And it’s sad.

Huh? We do, though I think it defaults to undetermined, so communities with a different main language don’t get filtered out.

But I mod one with 2 languages, if you post with “German” only people who have German active in their profile will see it.

1. Civ V: 2,444h (mostly with Vox Populi and its predecessors)
2. Pathfinder: Wrath of the Righteous 1,972h (this includes 800h Alpha and Beta)
3. Pathfinder: Kingmaker 1,383h
4. Stellaris 1,048h
5. Slay the Spire 1,042h

All single player ;)

edit: Typo.

Nope, the J4105 (which is not a server CPU, server ones are powerful, not power saving) has an iGPU that’s again way more powerful than the PI’s.

I think ARM is still more efficient. It’s specifically the PIs (or it’s chips) that’s not that efficient by my understanding. AFAIK Intel also is not as close for power draw under load.

The S740 is actually passively cooled, and it sits in my cupboard together with a pi4 backup server, router and modem ;) The S740 became very popular with German selfhosters (used prices actually went from 40€ to 80€ for just the base model because of high demand :D), so there’s a page with power measurements, it’s in German, but pretty self explanatory for the most parts.

Modern Intel CPUs are actually very efficient. And PI’s are not very efficient, just low-powered ;)

Another option: Used ThinClients.

I run Proxmox with HomeAssistant, Jellyfin, and some other services on a refurbished Futro-S740, it has a J4105 CPU and 8GB RAM (not officially supported, but 16GB is reported to work) and I use 2 m2 SSDs (required an adapter from AliExpress) for storage.
It could also support 2 proper SATA drives with adapters (power issues might start if you use 3 or more HDDs, plus connection issues), but that always depends on the ThinClient in question.

A good bit more powerful than a PI4, but can be found cheaper with roughly the same idle power draw.

Source: CPU-Monkey

Relevant, wie bei HN geschrieben wurde ist es nur günstiger, wenn man dadurch den prompt um 90% reduzieren kann, also muss man sehr genau prüfen, ob sich das hier lohnt.

Nach dem, was ich gehört habe (habe selber nur Llama1 64B im Schneckentempo via CPU probiert, war viel schlechter als GPT3.5), soll LLama2 70B wohl mit GPT 3.5 mithalten können. Aber du brauchst halt ne high-end GPU, damit das auch nur ansatzweise nutzbar ist.

Ehrlich gesagt brauche ich keine richtige AI für meine Sprachsteuerung, selbst definierbare Befehle reichen mir vollkommen aus. Leider ist was in HA ist noch nicht auf dem gleichen Stand wie Rhasspi bevor es da aufgehört hat mit der aktiven Entwicklung.

Paraphrased from what my dentist told me: You eat, there will probably be sweet and acidic stuff that attacks your teeth. Now you brush your weakened teeth and damage them more.

I only have coffee in the morning, so I simply wait an hour after my coffee to brush my teeth ;)

Always depends, I grabbed a .eu for 10 years on sale, I think 15€ in total.

Personally the load on the major servers by having one more instance that subscribes to everything is why I think people should back off from creating more than the 1500 instances Lemmy network already has.

Unless you use those anti-social seeding apps, you probably won’t subscribe to everything. My instance is subscribed to exactly the communities I want.

And Lemmy is not mature/resilient enough for me to want to invest my time into some random instance.

So is everyone hosting individual instances dealing with them being public?

Yeah, I’d love for it not to be public, but that’s currently not a feature. Still, there’s only one user, so it’s not too bad, you just have to be more diligent to not subscribe to anything that would be illegal for you to host in your and your server’s country.

Nice, works great!

FYI, I needed to add @sveltejs/adapter-cloudflare to the devDependencies to get it to run locally.

Yeah, here the use cases differ ;)

I only care about instances that won’t send me updates, I don’t care much about the reason they don’t, so I’d be perfectly fine with seeing 70 instances in the defederated block. But I guess this one is pretty simple to solve by making a new block “whitelisted, and you are not in it” ;)

baraza.africa is not checked despite being in the awesome lemmy list.

It’s my defed-checker testcase, because they recently went whitelist only and defederated me that way. So the way I understand it, your tool should work there, it’s just not checking this instance for some other reason

20€ to 35€, might change, I’ll see (I use nothing but Domains). Because I locked in the old price and am paid till 2026 ;)

QR code habe ich erwähnt.

Was A…-Pay angeht, ich gebe lieber Rewe meine Daten als Google/Apple.