Armin Weigl @arminweigl@librem.one

@raineer @drahardja @skyfire747 @nomenloony @blogdiva

Domain expert here: they aren't extremely effective for me in my field (programming) because to tell an LLM precisely enough what to do, I'd have to use a verbose English description, and I've already constructed the solution in my head in a more structural form that isn't words. Turning it back into words to tell an LLM what code to generate is a step backwards.

Es ist nicht so, dass alle Milliardäre in den Epstein Files vorkommen. Es gibt solche und solche Milliardäre. Einige kommen mehrmals vor während andere richtig oft vorkommen.

Dobrindt zahlt jetzt keine Integrationskurse (auch nicht für Ausländer, die schon eine Bewilligung haben, ihren Kurs aber noch nicht begannen). Dafür schimpft er in Zukunft auf die Ausländer, die sich nicht an unsere Geoflogenheiten anpassen. Was soll ich noch sagen: die CDU/CSU im Vollrausch.

@isotopp

“America is now what anthropologists call a Kardashian Type Three civilisation: more than fifty percent of GDP is in the attention economy.”

https://www.gregegan.net/MISC/RUINS/Ruins.html

RE: https://mastodon.online/@mastodonmigration/116013834072607458

Who owns Bluesky?

Let's ask Mike Masnick @mmasnick who is on the board of Bluesky PBC.

Mike is a famous long time journalism critic and founder of Techdirt. He regularly writes about how Bluesky is decentralized and how you can 'move your digital identity' around the Bluesky network.

Mike is a big advocate of open systems, so let's ask him to open up about who owns the company that he promotes so enthusiastically.

Mike, who really owns Bluesky?

#Bluesky #WhoOwnsBluesky

Thought-provoking ideas in chapter 2 of @emilymbender and @alex's The AI Con:

- Weizenbaum's ELIZA was meant to show that producing language does NOT imply thought.

- Babies don't learn language from passive exposure, so whatever LLMs are doing is fundamentally different.

- The notion of AGI is tied to the old racist notions associated with IQ tests. (I've thought of the G in AGI as meaning "not limited to a particular task", but rating relative intelligence has this issue.)

#ai

the financiers of the #pedophile Epstein Class want you to believe if their fanstasy valuations disappear, so will the global #economy.

girfters only know one thing: GRIFT. the crypto/AI house of cards is tumbling & they want to scare us into bailing them out.

AGAIN.

but economies are human activity, not financial instrumsnts.

BITCOIN’s FAILURE IS A “MARKET CORRECTION”

bitcoin billionaires are economics failures.

deal with it.

https://futurism.com/future-society/bitcoin-tanking-death-spiral

This NewPipe issue I hit and have been following is somewhat insightful on how the awfulness of AI codeslop plays out in practice. https://github.com/TeamNewPipe/NewPipe/issues/13139

No, this isn't NewPipe bad, they have a hard AI ban. It's seeing would-be contributors that was insightful.

If we hold all that against a H100 GPU as is being used for AI, we can see that this is a lost cause without launching any satellite.

The H100s GH100 die is about 814 mm2 on TSMC 4N with about 80 billion transistors.

This kind of die does not fly into space and survives longer than an afternoon.

https://indico.esa.int/event/165/contributions/1218/attachments/1205/1425/05b_-_KIPSAT_-_Presentation.pdf

ESA talks about 65nm processes, 16 times larger structures, 250 times larger area structures, for their space hardened compute.

The same amount of transistors for a H100 would be a 0.2 m2 slab, which also means that energy goes up and clock goes down, down, down.

Some compute ins pace runs on 28 nm structures, so a 50x times area increase compared to a H100. That's 40.000 mm2 for the same amount of transistors.

Or, in other words, a GPU does not work in space.

At all.

If you try, 80 bn transistors become around 25 bn transistors for overhead (n = 3) and redundant reserve.

So even if you sent a 4 nm node chip into space, it's no longer an 80 bn monster, but only a relatively modest 25 bn transistor GPU fragment.

Running hotter cools better, to the fourth power. So even relatively modest temperature increases will pay of big time in terms of radiative cooling (100ºC -> 120ºC), but that means more leakage, more power, less clock. So your 25 bn transistors net capacity will calculate slower than on earth.

A GPU does not work in space.

Did it ever occur to you why your computer has a fan, and why that fan usually stays quiet until the machine actually starts doing something?

When your laptop sits idle, very little is happening electrically. Modern processors are extremely aggressive about not working unless they have to. Large parts of the chip are clock-gated or power-gated entirely. No clock edges means no switching. No switching means almost no dynamic power use. At idle, a modern CPU is mostly just maintaining state, sipping energy to keep memory alive and respond to interrupts.

The moment real work starts, that changes.

Every clock tick forces millions or billions of transistors to switch, charge and discharge tiny capacitors, and move electrons through resistive paths. That switching energy turns directly into heat. More clock cycles per second means more switching. More switching means more heat. Clock equals work, and work equals heat.

This is why performance and temperature rise together. When you compile code, render video, or train a model, the clock ramps up, voltage often increases, and the chip suddenly dissipates tens or hundreds of watts instead of one or two watts.

The fan turns on not because the computer is panicking, but because physics is being obeyed.

Even when transistors are not switching, hot silicon still consumes power. As temperature increases, leakage currents increase exponentially. Electrons start slipping through transistors that are supposed to be off. This leakage does no useful work. It simply generates more heat, which increases temperature further, which increases leakage again. This feedback loop is one of the reasons temperature limits exist at all, and ultimately why we have fans – to keep the system under load below this critical temperature.

Above roughly 100 C, this leakage becomes a serious design concern for modern chips. Not because silicon melts (that's above 1400ºC) or stops working, but because efficiency collapses.

You spend more and more energy just keeping the circuit alive, not computing. To compensate, designers must lower clock speeds, increase timing margins, or raise voltage, all of which reduce performance per watt.

Reliability also suffers. High temperature accelerates wear mechanisms inside the chip. Metal atoms in interconnects slowly migrate. Insulating layers degrade. Transistors age faster. A chip running hot all the time will not live as long as one kept cooler, even if it technically functions.

This is why cooling exists, and why it scales with workload: It exists to keep the chip in a temperature range where switching dominates over leakage, where clocks can run fast without excessive voltage, and where the hardware will still be alive years from now.

In space, where you cannot rely on air or liquid to carry heat away, this tradeoff becomes unavoidable and very visible.

Run hotter, and you can radiate heat more easily.
Run hotter, and your electronics become slower, leakier, and shorter-lived.

Nikolai Kardaschow war ein Sowjetischer Radioastronom und Astrophysiker, der wissen wollte, wie man außeriridische Zivilisationen überhaupt erkennen kann, das "SETI-Problem".

Er dachte sich, "am Energieverbrauch". Die Kardaschow-Skala ist also keine ISO-Norm (ISO = Interstellare Normenorganisation?), sondern ein Denkwerkzeug für eine Person.

Die Skala sagt:

Typ 1: Nutzt die ganze Energie, die dem Planeten zur Verfügung steht. Also zum Beispiel eine Albedo-Null-Welt, die komplett mit Solarzellen vollgestellt ist.
Typ 2: Nutzt die ganze Energie, die dem Sonnensystem zur Verfügung steht. Also nicht etwa eine Million Satelliten in der Umlaufbahn, sondern eine Dyson-Sphere, die alles Licht, alle Strahlung eines Sterns aufnimmt.
Typ 3: Nutzt die Energie aller Sterne einer Galaxis.

Wir müssen die Idee, Rechenzentren ins All, genauer in LEO, zu schießen, also auf einer anderen Skala bewerten: Die Kardashian-Skala.

Eine Typ 1-Zivilisation auf der Kardashian-Skala hat einen gesellschaftlichen Diskurs, der nicht mehr durch Fakten, sondern durch Emotion und Aufmerksamkeit gesteuert wird.

Eine Typ 2-Zivilisation hat diesen Diskurs nicht mehr unter Kontrolle, und er kontrolliert Politik, Forschung und Wirtschaft, in denen jetzt unwissenschaftliche und quatschdumme Narrative die Steuerung übernehmen.

Da sind wird gerade.

Eine Typ 3-Zivilisation hat vollständige epistemische Entkopplung, also Sichtbarkeit Wahrheit vollkommen ersetzt, Reichweite Evidenz schlägt und Konsistenz irrelevant ist, solange das Narrativ nur spannend ist, physikalische Grenzen als "schlechte Vibes" gelten und Kritik als moralischer Makel und Zersetzung gelesen wird.

Wir messen also nicht Energie, sondern Realitätsverlust. Viele Tech-Visionen der letzten Jahre (alles ab und eingeschlossen Bitcoin) lassen sich direkt darauf abbilden.

Wir sind inzwischen schon eine Kardashian Typ 2-Zivilisation.

As someone to whom "incurious" is one of the sickest burns imaginable, a tool that hands me code saying "it passes tests - don't worry about how it works" is about the least useful thing i can imagine.

Saudi-Arabien-Besuch: Merz wohnt landestypischer Enthauptung bei https://www.der-postillon.com/2026/02/merz-saudi-arabien.html

inspired by CLAUDE.md, I’ve started putting markdown files named after coworkers into work code repos so I can remind them to stop doing shit to the codebase that annoys me

for some reason they’re all mad at me now, which means ill be adding commands to JEREMY.md for an attitude adjustment

OLG Dresden: Metas Datensammlung illegal, keine Revision zum BGH

Metas Datensammlung auf Drittwebseiten ist illegal - in Deutschland erstmals rechtskräftig entschieden. Das OLG Dresden lässt Meta nicht zum Bundesgerichtshof.

https://www.heise.de/news/OLG-Dresden-Metas-Datensammlung-illegal-keine-Revision-zum-BGH-11165785.html?wt_mc=sm.red.ho.mastodon.mastodon.md_beitraege.md_beitraege&utm_source=mastodon

#Datenschutz #Entertainment #MetaPlatforms #Recht #SocialMedia #Verbraucherschutz #Wirtschaft #news

Reactionary forces doing reactionary things

@VeroniqueB99 text:

Ashley Olivier ashleyaerials

My MAGA family keeps adding me to this group chat where they share MAGA propaganda. I've left three different times and asked them to stop re-adding me. But here we are again. So I've started responding with different screenshots of pages from the Epstein Files.

It's going great.

Who could have forseen this?

Oh, right, everyone who's been reading #solarpunk or thinking about how you can't shoot a virus with a gun or how survival skills in the Anthropocene involve gardening not weaponry.

(no diss on vortexegg, they get it!)

@T3Z Ich will fahrscheinfreien ÖPNV. Schluß mit der Kontrolliererei, mit Strafen oder gar Gefängnis für Menschen ohne Fahrschein, finanziert durch Steuern.

Ich will das 9€-Ticket zurück. Genauso, wie es war. Kein Abo. Einfach mit Kleingeld aus dem Automaten ziehen. Simpel, bezahlbar, barrierearm.