That is something that some tech savy Lemmy users could already easily do. I repost stuff from all over the web. But some systematic preservation of good old subreddits aught to be automated.

If the current dystopian US government abolishes OSHA, the LiveLeak logo might appear over American workers and not only Chinese workers like in the 2000s.

Isn’t this powered by American groq AI inference accelerator cards. At least I hope the groqracks are physically in Europe.

People with a single lever mixer tap:

Look what they need to Imitate a fraction of our power

Hüte dich vor den Ide(e)n des Merz!

It’s disgusting. Is it? Honey comes from a bee’s behind. Milk comes from a COW’S behind. And have you ever used toothpaste?

Temporarily for a month for Canada and Mexico. China has new tariffs

Try being 37. what was a simple half marathon in your 20s becomes a painful reminder of your physical decline.

Completely agree. Especially less good architects couldn’t deal with it’s limitations. Concrete isn’t some indestructible wonder material—exposure to harsh weather and exhaust fumes leads to cracking, corrosion, and a dreary look if it isn’t properly engineered and maintained. So many old concrete buildings look shitty

Basically appointing an investigator for fentanyl smuggling for Canada and sending troops to the northern border for Mexico.

No just those Kroepoek chips in communion wafer form

What the us army chants before annihilating another third world country

My American friends made me drink Keystone Ice and this was one of the most disgusting things I did ever drink. The rest was also not great. A few American beers were passable but I had better ones in most European countries.

Die Gewichte sind offen. Daten und Training wie so häufig nicht. Aber bei den meisten Modellen in dieser Leistungsklasse bekommt man nicht Mal die Gewichte.

Man trainiert ein Modell mit weniger Parametern um ein Modell mit vielen Parametern zu imitieren. Führt zu Leistungsverlusten bei der Ausgabequalität, aber braucht dann weniger Ressourcen beim Ausführen und funktioniert einigermaßen so wie das Ursprungsmodell in schlechter. Modelle mit 1 -15 Mrd. Parametern laufen auf einem gängigen modernen Laptop. Aber einige Modelle aus denen solche kleinen Modelle distilled wird sind viel größer (zum Beispiel , Deepseek R1 mit 640 Mrd. Parametern) und brauchen hunderte Gigabyte (Video-)RAM um laufen zu können.

When the sea turns so dark, and a jaw makes its mark,

That’s a moray!

When it hides in a reef, flashing sharp, jagged teeth,

That’s a moray!

Correction I copypasted the wrong new testament verse reference from a large collection on things the bible had to say on strangers. Both are old testament, but the list is long and very clear in the new testament too.

Oh the old testament is quite clear: Thus says the Lord: Do justice and righteousness, and deliver from the hand of the oppressor him who has been robbed. And do no wrong or violence to the resident alien, the fatherless, and the widow, nor shed innocent blood in this place.

Jeremiah 22:30

And the new is no different:

You shall treat the stranger who sojourns with you as the native among you, and you shall love him as yourself, for you were strangers in the land of Egypt: I am the Lord your God.

Matthew 25:40

Das alles betrifft natürlich nicht die opensource Version von deepseek R1. Man kann sich mit Ollama die kleineren destillierten Versionen des modells laden und auf einem mittelstarken Laptop lokal ausführen. Hatte auf einem Rechner mit mehr Power sogar schon die 640B Parameter Version auf die Art installiert und benutzt.

Plato didn’t skip pre-workout, just pre-Socratics