Nostr notes by Martin Seeger

Security Bug du jour: CVE-2026-98432 We head towards non-stop ...

2026-10-30T09:47:43Z

Security Bug du jour: CVE-2026-98432

We head towards non-stop patching.

Zitat des Tages: *Unsere ganze IT-Sicherheit basiert auf einer ...

2026-10-27T13:47:38Z

Zitat des Tages: *Unsere ganze IT-Sicherheit basiert auf einer unbewiesenen mathematischen Behauptung, abgebildet in einem ungeeigneten Trustmodell und umgesetzt unter Zeit- und Gelddruck von Leuten ohne Ahnung aber mit Interessenskonflikten.*

Ich hätte ich einen NETGEAR 5G Router mit SIM-Karte & WiFi 6 ...

2026-10-26T14:32:52Z

Ich hätte ich einen NETGEAR 5G Router mit SIM-Karte & WiFi 6 (Modell:MR5200-100EUS) abzugeben.

Ich denke, das so 150€ ein fairer Preis wären.

#flohmarkt

You're faster than most commercial services on the issues ...

2026-10-21T09:01:24Z

In reply to nevent1q…d5dn
_________________________

You're faster than most commercial services on the issues that matter 😉.

RE: https://chaos.social/@evawolfangel/117449081037348304 Wo ...

2026-10-16T06:25:40Z

RE: https://chaos.social/@evawolfangel/117449081037348304

Wo Eva Wolfangel (npub1fxr…dgxg) Recht hat, hat Sie Recht.

Ich habe jetzt Dutzende Reports von KI-Angriffen gelesen. Bei jedem davon hätte ein Schutz nach Stand der Technik den Angriff verhindert.

Wir haben da zwei Ebenen:<li>Aktuelle Software schleppt oft Jahrzenhnte an Alt-Lasten mit sich rum. Diese wurden, da sie laufen, vor 15 Jahren das letzte Mal angefasst. Hier werden wir einen Patch-Tornado erleben und 1.000 CVEs pro Monat und Betriebssystem sind nicht unrealistisch.</li><li>Der Stand bei den operativen Fähigkeiten hat sich über die letzten Jahre weiter verschlechtert. Da laufen Setups draußen, die wären schon 2015 nicht OK gewesen. Heute sind sie eine Einladung. Hier sehe ich noch keinen echten Push, das anzugehen und wir werden da noch durch ein sehr tiefes Tal der Tränen gehen.</li>

quoting
note1227…d3s5

Mehr ist eigentlich nicht dazu zu sagen. Ok vielleicht ein bisschen mehr :) Freebie zu meinem Kommentar: https://www.zeit.de/digital/datenschutz/2026-10/cyberangriffe-kuenstliche-intelligenz-frankreich-supermaerkte?freebie=db9d1c48
#cybersecurity #danfoss

RE: https://mastodon.ie/@ciaranmak/117438698166365017 If you play ...

2026-10-15T18:32:12Z

RE: https://mastodon.ie/@ciaranmak/117438698166365017

If you play a game showing ads, you’re the one being gamed.
nostr:note12jnmul6w93cqzh2l0wr66nsngrnhxyvkwcq723gm2w2hxmpecrvsgxjd25

Wie man in Deutschland KI einführt 😁 Meine Beobachtung aus ...

2026-10-14T17:55:41Z

Wie man in Deutschland KI einführt 😁

Meine Beobachtung aus einem KI-Workshop.

#AIgenerated

Kann jemand aus meinen Kreisen einen Stiftehalter für 100 ...

2026-10-14T15:33:50Z

Kann jemand aus meinen Kreisen einen Stiftehalter für 100 schmale Stifte/Bleistifte (max. 9mm Durchmesser) gebrauchen?

Ich habe das Teil gedruckt ohne vorher meine Stifte (Acrylstifte) zu messen und das Teil ist zu eng.

Es wäre Schade um das Filament, daher verschenke ich es gerne, falls es jemand brauchen kann.

#flohmarkt #fedigive

I am impressed about the reusable desiccant for my #3dprinting ...

2026-10-12T10:48:02Z

I am impressed about the reusable desiccant for my #3dprinting filament.

You put it into the microwave oven at 600W for 5min and it is good to go again.

But there are some cave-ats:<li>Don't do it in a porcelain container as it will get incredibly hot and take a long time to cool down.</li><li>Don't get hasty because it takes so long to cool down. Don't put it too early in your 3D-printed container as it will melt through it.</li><li>When it melts through, don't try to catch it with the (still incredibly hot) porcelain container which you grab with your bare hands.</li>

So many valuable lessons this hobby has 😄.

TIL: Identifying and Backing Up Game Saves I want to retire ...

2026-10-11T17:25:03Z

***TIL: Identifying and Backing Up Game Saves***

I want to retire my old gaming PC and hand it over to a friend.

But before doing so, I wanted to make sure I had a backup of every saved game on that machine.

I wasn't exactly looking forward to this task, expecting to spend hours hunting down save files scattered across countless directories.

Today, I discovered [Ludusavi](https://ludusavi.com/ ) ([on GitHub](https://github.com/mtkennerly/ludusavi )).

This tool does the entire job with a single command:ludusavi backup --path "E:\SteamSavegames" --force

In just seven minutes, it checked its database of over 50,000 games, identified 80 of them on my PC, and backed up all their save files to a network drive.

Some of those games I'd completely forgotten I'd ever played. 😄

Excellent tool!

Tips für Profis: **Wenn man in Deutschland eine ...

2026-10-09T12:00:47Z

Tips für Profis: **Wenn man in Deutschland eine Gesundheitsanwendung für die Behandlung von Depressionen auf den Markt bringt, dann sollten die Daten nicht in einem ungeschützten AWS S3 Bucket landen.**

Just saying ....

Immer wieder schön: Jemanden in einer anderen Branchen seine ...

2026-10-08T09:40:58Z

Immer wieder schön: Jemanden in einer anderen Branchen seine rechtlichen Grundlagen erklären.

In diesem Fall durfte ich meinem Hotel gestern Nacht das Bundesmeldegesetz erklären.

Der Checkin fragt inzwischen oft Daten ab, auf die das Hotel gar keinen Anspruch hat und verweist auf "gesetzlliche Grundlagen für den Meldeschein."

In einigen Hotels wird einem zudem zur Anmeldung einfach nur ein Tablet gegeben, dass die Registrierung ohne bestimmte Daten einfach nicht durchführt.

My ex-AV regularly was like that: Hey this 20-year-old mbox file ...

2026-10-07T18:13:11Z

In reply to nevent1q…wu25
_________________________

My ex-AV regularly was like that: Hey this 20-year-old mbox file with 30.000 emails contains some virus. But we won’t tell you which email and remove the complete mbox file. So it’s ex-AV now 😁

As a side note: during the invasion of Iraq, more than twenty ...

2026-10-03T11:22:03Z

In reply to nevent1q…995h
_________________________

As a side note: during the invasion of Iraq, more than twenty years ago, I used to summarize another problem with counterterrorism like this:

> **Killing one terrorist with a bomb produces ten new terrorists.**

The number ten was obviously rhetorical, not a statistic. But the underlying mechanism is remarkably similar.

Again, you have a selection problem. You are trying to kill a very small number of terrorists embedded in a much larger population of people who are not terrorists.

Your weapon and your intelligence are imperfect, so you inevitably produce false positives. Except that in this case, a false positive is not someone who gets investigated. It may be someone who gets killed.

And people tend not to react favorably to their children, parents, spouses, friends or neighbors being killed.

So the false positives feed back into the system. Every innocent person killed creates a circle of people with a very personal reason to hate whoever dropped the bomb.

Most of them will never become terrorists, of course. But you don't need many of them to do so before your counterterrorism operation starts producing some of the very thing it is supposed to eliminate.

In surveillance, false positives consume attention.

In kinetic counterterrorism, false positives can create enemies.

**Different consequences, same fundamental problem: when the target is extremely rare and your selection mechanism is imperfect, being wrong vastly more often than being right can dominate the entire system.**

In this case, there is also a simple mathematical reason why mass ...

2026-10-03T11:06:37Z

In reply to nevent1q…f8g5
_________________________

In this case, there is also a simple mathematical reason why mass surveillance works like shit.

You have three ingredients:<li>a very large population,</li><li>a tiny number of actual terrorists,</li><li>and a detection system with a measurable false-positive rate.</li>

That combination is mathematically nasty.

Assume you monitor one million people and ten of them are actually terrorists. Now assume you have an extraordinarily good detection system: it identifies 90% of the terrorists and has a false-positive rate of only 0.1%.

It will correctly identify nine terrorists.

But among the 999,990 innocent people, a 0.1% false-positive rate produces roughly **1,000 false alarms**.

So your supposedly excellent surveillance system gives you about 1,009 alerts, of which only nine are real. More than **99% of your alerts are false positives**.

And this example is deliberately generous. A false-positive rate of just **0.1% is unrealistically low** for a real-world system trying to infer something as vague and complicated as terrorist intent from communications, behavior, associations, travel patterns, social media activity and other surveillance data. At 1%, the same example would produce roughly **10,000 false positives for nine true positives**. At 5%, it would produce nearly **50,000**.

This is the base-rate problem. When the thing you are looking for is extremely rare, even a remarkably accurate detection system can produce vastly more false positives than true positives. In the real world, where detection is messy and the criteria themselves are ambiguous, the problem gets much worse.

And now you have created a second problem: alarm fatigue.

Someone has to investigate all those innocent people. Analysts learn that almost every alarm they see is bullshit. Attention becomes the scarce resource. The system designed to find the needle in the haystack responds to the difficulty of finding the needle by continuously adding more hay.

Mass surveillance therefore has a fundamental scaling problem: **collecting more data is easy. Producing more alerts is easy. Producing more human attention is not.**

# Surveillance Considered HarmfulThere is an interesting detail ...

2026-10-03T09:02:27Z

#

Surveillance Considered HarmfulThere is an interesting detail emerging from the attempted terrorist attack on Flydubai Flight 1073.

According to current reports, the co-pilot had already been barred from flying by Oman Air because of concerns about extremist views. Yet somehow he ended up flying for Flydubai. And somehow an Omani citizen, from a country whose pilots are not allowed to fly into Israel, ended up in the cockpit of a flight to Tel Aviv (Source: https://apnews.com/article/cf64a228799e6e788e834cac2a489ab2).

This happened in a region containing some of the most heavily surveilled societies on the planet. Saudi Arabia, where the aircraft eventually landed, is a particularly extreme example: internet activity, social media and private communications are subject to extensive state surveillance (Source: https://freedomhouse.org/country/saudi-arabia/freedom-net/2025). Other Gulf states are not far behind in the scale and reach of their surveillance.

And yet here we are. This points to a more fundamental problem with surveillance as a weapon against terrorism.

What does a terrorist actually want?

*Physical destruction is only the means. The real target is psychological: fear, distrust, disruption and ultimately the fragmentation of society. A terrorist cannot destroy a society through physical violence alone. But it may be sufficient to make that society afraid enough to start damaging itself.*

And this is where mass surveillance becomes dangerously counterproductive. A functioning society depends on trust: trust between citizens, trust in institutions, and the assumption that most people around us are not our enemies. Terrorism attacks exactly that trust.

Mass surveillance does, too.

It tells everyone that everyone is potentially suspicious. It normalizes observation, suspicion and self-censorship. In trying to make society more resilient against individual terrorists, we risk weakening the very social cohesion that makes a society resilient in the first place.

The irony of Flight 1073 is therefore hard to miss.

**We have built extraordinarily powerful machinery for watching millions of ordinary people, while apparently failing at the much more mundane task of asking why a pilot previously considered an extremism risk was sitting in the cockpit of an airliner bound for Israel.**

Surveillance considered harmful. Not merely because it invades privacy. But because it can fail even at the trivial things while quietly weakening the society it was supposed to protect.

Neulich stieß ich auf ein Video, bei dem mich die Frage deutlich ...

2026-10-02T13:11:06Z

Neulich stieß ich auf ein Video, bei dem mich die Frage deutlich mehr interessierte als die Antwort, die es gab:

**Merkten die (West-)Römer, dass ihr Imperium fällt?**

Wenn man der Frage nachgeht, stößt man zunächst auf das übliche Quellenproblem. Unsere schriftlichen Zeugnisse stammen fast ausschließlich aus der gebildeten Oberschicht. Über das Empfinden der Bevölkerung insgesamt können wir daher nur sehr eingeschränkt etwas sagen.

Nehmen wir einen gallorömischen Grundbesitzer um 470:<li>Seine Villa steht noch.</li><li>Er spricht Latein.</li><li>Seine Kinder bekommen eine klassische Bildung.</li><li>Der örtliche Bischof ist Gallorömer.</li><li>Steuern werden weiterhin eingezogen.</li><li>Die Städte existieren weiterhin.</li><li>Römisches Recht wird weiterhin benutzt.</li>

Nur die oberste politische Herrschaft hat gewechselt. Statt des Kaisers in Ravenna herrscht in seiner Region nun vielleicht ein westgotischer König.

Für einen solchen Menschen musste das nicht aussehen wie der Untergang einer Zivilisation. Vieles funktionierte weiter.

Aber es gibt Hinweise darauf, dass zumindest in vielen Regionen ein sehr reales Gefühl von Unsicherheit und staatlichem Rückzug vorhanden gewesen sein muss:<li>wiederholte Plünderungen und lokale Kriege,</li><li>steigende oder als ungerecht empfundene Steuerlast,</li><li>Landflucht und Aufgabe von Besitz,</li><li>zunehmende Abhängigkeit von Großgrundbesitzern,</li><li>regionale Aufstände wie die der Bagauden,</li><li>Städte und Gemeinden, die ihre Verteidigung zunehmend selbst organisieren mussten,</li><li>und Menschen, die in Gebiete unter germanischer Herrschaft wechselten.</li>

Salvian beschreibt ausdrücklich Römer, die zu den Goten oder zu den Bagauden flohen, weil sie ihre Lage dort für besser hielten.

Das ist vielleicht so nah an einem „Volksgefühl“, wie wir mit unseren Quellen kommen.

Und in den erhaltenen Quellen verschiebt sich über mehr als ein Jahrhundert langsam der Ton: von **"Das Reich steckt in einer schweren Krise"** über **"Unsere Welt erlebt eine Katastrophe"** bis hin zu **"Die alte politische Ordnung verschwindet."**

Daher würde ich es am Ende eher so formulieren:

**Die Römer erlebten nicht den "Fall Roms". Sie erlebten eine jahrzehntelange Serie von Dingen, die zu funktionieren aufhörten.**

Das ist ein Post über die Gegenwart.

RE: https://ohai.social/@GreenSkyOverMe/117360188583379509 Am 23. ...

2026-09-30T13:57:30Z

RE: https://ohai.social/@GreenSkyOverMe/117360188583379509

Am 23. Juni 2026 hat VW insgesamt 2.619.107.694,30 € Dividende ausgeschüttet.

Das ist keine 100 Tage her.

Da ging es dem Unternehmen anscheinend noch so blendend, dass man Milliarden an die Inhaber ausschütten konnte.
nostr:note1cdmfrr4wgc55cskc0xp00h6ea4kw865hhhsp7rtwsglepeq4yzssrqchva

Visiting the past I left my “home village” 34 years ago. ...

2026-09-29T09:02:51Z

**Visiting the past**

I left my “home village” 34 years ago. I arrived there at an age of 6 and left with 26.

One of my quirks is that I have problems with dentists. So if I ever trust one, I stick with him. Consequently for dental reasons, I occasionally still visit the village. With public transport it is a 90-min drive.

Large parts of the bus drive are identical to the route I took first to school and then to university. Going that route at least twice a day for 15 years left it deeply engraved in my brain.

Going back to the old places feels strange. When I arrived in 1972, fishing and agriculture were still important. You could still see the traces of WW2 (bomb craters, disposed ammo, bunkers, etc). Already during my time there, the village had changed: agriculture and fishing disappeared and tourism started to rule supreme.

Now, three and a half decades later, all traces of village identity have been filed off. It feels as if someone had cast a gigantic mold over the village, pressing all its individuality into a standardized form. The building style, the stores, the services: everything has been unified.

It has become totally alien to me. And the same thing has happened to every village along the way. The memories of the old places are still there, but they have no longer anything they can anchor to.

My apologies, I read your C# as C++. Sorry!

2026-09-28T07:48:57Z

In reply to nevent1q…g554
_________________________

My apologies, I read your C# as C++. Sorry!

It seems Inkscape is hiring C++ devs: ...

2026-09-27T07:02:39Z

In reply to nevent1q…z0ek
_________________________

It seems Inkscape is hiring C++ devs: https://social.treehouse.systems/@dysfun/116182391411729473

I am a bit unsure as my French is not what it used to be. Unsure if they offer remote work.

This [video](https://youtu.be/T-oXyXwD6sE ) makes an interesting ...

2026-09-27T06:50:57Z

This [video](https://youtu.be/T-oXyXwD6sE ) makes an interesting point about IPOs: **an IPO finally makes it possible to bet against a company.** Which, depending on the company, may be the most exciting feature of going public.

The video also references the chart below. No new technology has ever become cheaper this quickly than AI. The chart comes from this [article](https://epoch.ai/publications/the-plunging-price-of-thought ).

Discovered at IKEA... cc @npub16kg…afn8 ...

2026-09-23T14:41:28Z

Discovered at IKEA...

cc Cat 🐈🥗 (D.Burch) (npub16kg…afn8)

I love that channel ❤️

2026-09-21T06:30:05Z

In reply to nevent1q…dnnr
_________________________

I love that channel ❤️

IMHO the reference to AI is an attempt to diffuse responsibility. ...

2026-09-18T18:59:24Z

In reply to nevent1q…qywh
_________________________

IMHO the reference to AI is an attempt to diffuse responsibility. This is 100% on the human operators.

Freedom of choice must be part of AI policy. AI as a service and ...

2026-09-18T11:21:44Z

In reply to nevent1q…yz23
_________________________

Freedom of choice must be part of AI policy.

AI as a service and open-weight models both have a legitimate role. Regulation and public procurement should preserve the ability to choose between them: portability of data and evaluations, interoperable interfaces, the right to run models locally, and credible exit paths from providers.

Freedom of choice also includes the option not to use AI where it adds no value, and access to a human route where an essential service is involved.

Convenience today must not quietly become dependency tomorrow. Exit doors are best installed before anybody smells smoke.

25/34

Policy and procurement should prefer the smallest model that can ...

2026-09-18T11:21:29Z

In reply to nevent1q…9r98
_________________________

Policy and procurement should prefer the smallest model that can reliably perform a task.

That is not a ban on general-purpose or frontier models. Some problems may justify them. But “larger” should not be accepted as a synonym for “better.” Not every nail requires an orbital hammer.

Smaller, specialized models can reduce computation, memory, hardware, and energy requirements in both deployment and further development. They can also make systems easier to test and replace.

My preference is simple: use scale where its additional value can be demonstrated, not merely where it can be financed. Speculative future capability should not automatically outweigh present ecological cost.

24/34

I can now turn this analysis into personal conclusions. The first ...

2026-09-18T11:21:09Z

In reply to nevent1q…l4nj
_________________________

I can now turn this analysis into personal conclusions. The first level is political.

Binding regulation of frontier-model training must come. Not because every new model is illegitimate, but because the present race allows individual actors to commit energy, water, hardware, data, and financial resources while shifting part of the risk onto everyone else.

Voluntary promises cannot resolve an incentive trap. A runner in the race is not the ideal person to operate the starting pistol, judge the lanes, and decide whether the track is safe.

We need disclosure of resources and incidents, environmental accounting, auditable obligations, enforceable thresholds, and international coordination where risks cross borders. Otherwise, a competition among a few companies and states may land on the entire world’s feet.

23/34

The economic impetus will push inference toward wider use, labor ...

2026-09-18T11:20:50Z

In reply to nevent1q…szml
_________________________

The economic impetus will push inference toward wider use, labor savings, recurring services, and dependency. That pressure is real, but it is not the same as truth.

A profitable deployment is not necessarily useful. Rapid adoption is not evidence of consent. A lower price does not contain every ecological or social cost. What a market rewards depends on who owns the infrastructure, who can refuse, and who may externalize risk.

Markets are quite good at producing prices. I remain unconvinced that this makes them experts in metaphysics. I will not judge an application by whether the market makes it appear inevitable.

I will return to my criteria: ecological viability, freedom from dependency, democratic moderation of structural change, and whether the technology serves human beings. That is why I separate inference technology from its application.

22/34

The legal status of AI output is not one question, although it is ...

2026-09-18T11:20:17Z

In reply to nevent1q…hejw
_________________________

The legal status of AI output is not one question, although it is often placed in one convenient box marked “legal stuff.”

Can the output itself receive copyright protection? Who owns the human contribution? Does the result infringe an existing work? And what rights do provider terms actually grant? These questions differ by jurisdiction and by the degree of human authorship.

GEMA’s case against Suno illustrates one part of the problem: GEMA alleges that Suno produced tracks confusingly similar to protected songs. That is a claim about possible infringement, not simply about whether an AI output can have an owner.

Until the boundaries are clearer, commercial use requires provenance, license checks, human review, and a realistic allocation of risk. The word “generated” is not a legal invisibility cloak.

21/34

Efficiency gains immediately raise a distribution question: where ...

2026-09-18T11:19:58Z

In reply to nevent1q…cg9w
_________________________

Efficiency gains immediately raise a distribution question: where do they go?

A productivity gain does not arrive with a moral destination label. If an hour saved becomes better service, higher pay, more autonomy, safer work, or actual free time, AI can improve human work. If it becomes a higher quota, fewer colleagues, closer monitoring, and the same pay, efficiency becomes work intensification.

Both are real possibilities. The OECD reports substantial worker-perceived benefits from AI, but also concerns about work intensity, data collection, and inequality. The ILO finds real but uneven productivity gains that have not automatically translated into higher measured output, earnings, or employment.

Who captures the gain and who carries the new burden are not implementation details. Workers and the public must have a voice in that decision.

20/34

The application layer begins with a boundary: I reject AI as a ...

2026-09-18T11:19:40Z

In reply to nevent1q…aset
_________________________

The application layer begins with a boundary: I reject AI as a compulsory ritual.

An employee should not be forced to use AI irrespective of whether it helps the task. A patient should not lose access to a doctor because an AI system has become the only way to obtain an appointment.

AI may be an option, an assistant, or even the best default in a defined setting. In essential services especially, there must still be an accessible human alternative, a path to appeal, and a person or institution that remains accountable.

Nothing makes me doubt the usefulness of a tool like making it mandatory.

Adoption figures produced by coercion tell us little. A technology meant to serve people must leave them meaningful agency.

19/34

My main technological requirement is optionality. An API can ...

2026-09-18T11:19:23Z

In reply to nevent1q…77wr
_________________________

My main technological requirement is optionality.

An API can change its price, behavior, terms, or availability. A local model brings its own costs: hardware, operations, updates, and dependence on upstream model releases. Absolute independence is an illusion, but avoidable lock-in is still a choice.

I can only advise people and organizations to test their real use cases against local models. Not a generic benchmark, but their documents, workflows, quality thresholds, latency, privacy needs, and failure modes.

Keep data and evaluations portable. Know where a frontier service genuinely earns its premium. Know which critical functions still work if that service disappears.

An exit plan is terribly boring, right up to the day it becomes horribly contemporary.

18/34

My rough map is that strong open-weight models have entered ...

2026-09-18T11:19:04Z

In reply to nevent1q…s83j
_________________________

My rough map is that strong open-weight models have entered territory that was occupied by frontier services in 2025 on many tasks.

That is practical shorthand, not a claim of universal equivalence. Benchmarks are maps: useful, compact, and perfectly capable of leaving out the swamp. They compress important differences in modalities, long context, tool use, reliability, safety, and specialized knowledge.

Mozilla reports only a small aggregate gap in its evaluated benchmarks, while a specific use case can still show a large one. Services have not become obsolete. Frontier services and open-weight models both have a place. The useful question is which capability, control, and cost profile the actual task requires.

17/34

I see enormous optimization potential in both approaches: smaller ...

2026-09-18T11:18:44Z

In reply to nevent1q…454c
_________________________

I see enormous optimization potential in both approaches: smaller specialized models, distillation, quantization, caching, batching, routing, better harnesses, and better hardware.

I think improvements of two or more orders of magnitude are possible across the inference stack. One analysis projects up to 100× lower cost per token over time; a peer-reviewed energy study identifies 8–20× in foreseeable savings per query.

A slower cadence of model generations would make this easier. Stable targets give engineers time to optimize runtimes, quantization, hardware, and serving instead of rebuilding around the next frontier every few months.

That should reduce the ecological burden of each useful task substantially. It does not guarantee lower total consumption. Cheaper inference can create vastly more inference, while long reasoning traces can eat some of the gains with excellent table manners.

Efficiency gives me reason for qualified optimism. It is not an ecological free pass.

16/34

Now to inference: using a model after it has been trained. I want ...

2026-09-18T11:18:25Z

In reply to nevent1q…05d9
_________________________

Now to inference: using a model after it has been trained. I want to separate the technology from its application, because the discussion is already carrying enough luggage.

Technically, two delivery modes matter to me.

With **AI as a service**, a provider operates the model and sells access. This offers convenience, rapid updates, elastic capacity, and often the strongest available capabilities. It also means less control and continuing dependency on prices, policies, and availability.

With **open-weight AI**, the learned parameters can be downloaded and run or adapted elsewhere, including locally. That can offer privacy, control, and portability. Open weights are not necessarily open source, though. Training data, training code, or broad usage rights may still be missing.

Neither approach is inherently right. Sadly, the universe has declined to provide one architecture that solves every trade-off.

15/34

I have seen technological dominance disappear before. In 1998, ...

2026-09-18T11:18:00Z

In reply to nevent1q…guyp
_________________________

I have seen technological dominance disappear before.

In 1998, Netscape named me “Partner of the Year.” Netscape appeared to rule the internet. US antitrust records show its browser share falling from roughly 76% in early 1997 to 19% by 2000. The company itself had been acquired by AOL.

Where is Netscape today?

This is not a prediction that today’s AI leaders will vanish. It is a reminder that technological eternity can be surprisingly short. Market leadership, capital, and cultural inevitability are not the same as durable value. A company can shape an era without owning its destination.

That memory makes me skeptical when temporary corporate dominance is used to justify permanent infrastructure, dependency, or public risk. We should govern the transition for society, not build policy around the survival needs of its current leaders.

14/34

My conclusion is blunt: the unchecked race of ever larger, ...

2026-09-18T11:17:41Z

In reply to nevent1q…2ftx
_________________________

My conclusion is blunt: the unchecked race of ever larger, privately controlled training bets has to stop. I do not mean research, every new model, or the use of LLMs. I mean that this cannot continue as normal.

OpenAI and Anthropic appear to recognize at least part of the problem. Some industry leaders have called for pacing the frontier; OpenAI says it would slow or stop when risks cannot be sufficiently safeguarded.

But they are riding a tiger, and tigers are unhelpful when asked for safe dismounting instructions. Slowing threatens valuation, funding, talent, market position, and the story justifying infrastructure commitments. Continuing deepens the dependencies.

I see no credible off-ramp. If the companies cannot find one, politics and society must build it from outside. The brake cannot depend on the riders alone. We need public rules, coordination, transparency, and a way to absorb the economic consequences.

13/34

Training also requires content. The early norm was often: if ...

2026-09-18T11:17:23Z

In reply to nevent1q…a07v
_________________________

Training also requires content. The early norm was often: if material was accessible online, it was available for ingestion. Whether that is lawful remains contested across jurisdictions. Whether it was fair to creators is a separate question.

Now many actors repeatedly crawl the same web. This imposes costs on the people and institutions that produced the value. Wikimedia reported that bots generated at least 65% of its most resource-intensive traffic. Other infrastructure providers document AI crawlers ignoring restrictions or obscuring their identity.

Not every automated request is for training. Some support search or user queries. But the larger pattern remains: private acquisition pipelines duplicate load while shifting bandwidth, hosting, moderation, and defense costs onto the open web.

The web is treated as a buffet, except the bill is sent to the people who cooked. That is a poor foundation for a public-benefit narrative.

12/34

Then there is duplication. Competing labs acquire similar data, ...

2026-09-18T11:17:01Z

In reply to nevent1q…swqp
_________________________

Then there is duplication.

Competing labs acquire similar data, run similar experiments, and spend vast resources training overlapping general-purpose models behind closed walls. Some duplication is valuable: competition, replication, and alternative approaches matter.

At this scale, however, secrecy and the race for a temporary lead mean that failures, dead ends, and discoveries are repeatedly paid for in energy, hardware, money, and human work. Competition is useful. Paying several times to rediscover the same wall in private is less obviously so.

The final training run is only a fraction of total R&D compute, and much of the search remains invisible. More compute has produced remarkable gains, but a larger invoice is not evidence that the next run will create proportional social value, or even a durable business advantage.

For me, the burden of proof rises with the resources committed.

11/34

The AI infrastructure race also reshapes hardware markets. Demand ...

2026-09-18T11:16:35Z

In reply to nevent1q…s6fn
_________________________

The AI infrastructure race also reshapes hardware markets. Demand absorbs advanced packaging and high-bandwidth memory while capacity moves away from conventional components. Shortages and higher prices can spread far beyond AI.

Financing adds another layer. Chipmakers and cloud providers invest in AI labs that become major customers. Debt, private credit, long-term commitments, and off-balance-sheet structures fund the buildout.

Circular financing may indicate a bubble, but does not prove one. It can make money travel in a circle and return wearing a hat labelled “independent demand,” while spreading failure through a tightly connected system.

The BIS calls the macro-financial risk moderate, not absent, and dependent on very high future earnings. That is a huge bet on benefits we still struggle to measure.

Unlike a casino, this bet reaches people who never placed a chip and will receive none of the winnings. They can face higher prices, public costs, and economic fallout.

10/34

By these standards, training is currently the part of AI I find ...

2026-09-18T11:15:58Z

In reply to nevent1q…jhct
_________________________

By these standards, training is currently the part of AI I find most open to criticism.

Not because training necessarily consumes most electricity over a model’s entire life. At massive scale, inference can consume more. My concern is that training decisions trigger the race for larger clusters, more power, more cooling, more chips, and more data before we know what the next increase in scale will actually buy us.

The infrastructure bonanza is real. US data centers are projected to consume 9.5–15.3% of national electricity by 2030. The burden is not only carbon: it includes water, land, minerals, hardware production, e-waste, grid pressure, and costs concentrated in host communities.

Training and anticipated inference both drive this buildout. One isolated training run is not my point. The problem is an expansion cycle in which the ecological commitment arrives first and proof of additional value is asked to catch up later.

9/34

Before descending into my interpretation of the different ...

2026-09-18T11:15:34Z

In reply to nevent1q…2zhk
_________________________

Before descending into my interpretation of the different aspects, I should put my cards on the table. They are not stone tablets, just the standards I intend to use:<li>Preserving the ecological systems that sustain human life is my first priority. No promised capability makes their destruction acceptable.</li><li>We must coûte que coûte avoid economic dependencies that leave us without real alternatives.</li><li>Changes to society’s basic socioeconomic structure must be publicly and democratically moderated. Companies cannot control that process alone.</li><li>Technology, including AI, should serve human beings, not the other way around.</li>

These are not hard scientific criteria. They are personal value judgments. I do not think that element can be removed from evaluation, however sternly one looks at the spreadsheet.

Making those values explicit seems more honest than pretending they are absent. I will use them to examine resources, ownership, institutions, incentives, and effects.

8/34

Technology never arrives alone. It brings economics and politics ...

2026-09-18T11:15:04Z

In reply to nevent1q…8ekz
_________________________

Technology never arrives alone. It brings economics and politics with it, usually carrying most of the luggage.

For much of the US race, capital rewarded ever larger, general-purpose models. Capability could justify investment, create platform dependence, and form an economic moat. I also see a political climate that increasingly rewards black-and-white positions: acceleration or obstruction, optimism or doom. Grey is apparently less suitable for campaign material.

China faces different incentives: tighter political control, state priorities, and constrained access to advanced chips. These encourage efficiency, open weights, and smaller or specialized models. Distillation is part of that strategy. Firms have also been accused of distilling foreign competitors’ outputs, although each claim needs evidence.

These are tendencies, not national laws. We should not blame technology for every shape imposed on it by the world around it.

7/34

> "There are two kinds of fools. One says, ‘This is old, ...

2026-09-18T11:13:58Z

In reply to nevent1q…4ye4
_________________________

> "There are two kinds of fools. One says, ‘This is old, and therefore good.’ And one says, ‘This is new, and therefore better.’”

> John Brunner, > *The Shockwave Rider*

Neither familiarity nor novelty tells us whether an LLM is useful, harmful, intelligent, or trivial.

“Merely autocomplete” understates what scaled prediction can do. But calling it a “machine god” or an “incarnation of evil” does not explain it either.

Humans are unreliable autocomplete engines too. We readily complete fragments of other people’s behavior with intentions we supplied ourselves, then forget who supplied them.

“Machine god” and “incarnation of evil” are mythological labels, and mythology is wonderfully cooperative. It can mean whatever the argument requires while projecting intention, destiny, and moral agency before we have demonstrated them.

I prefer capabilities, limits, uses, and consequences. Less dramatic than summoning gods and demons, but more informative.

6/34

One distinction is essential: training is not inference. During ...

2026-09-18T11:13:20Z

In reply to nevent1q…q9xg
_________________________

One distinction is essential: training is not inference.

During **training**, a model learns its parameters from enormous datasets through repeated computation across many accelerators. This is where most upfront demand for compute, energy, and training data arises. It is also where many questions about how content was acquired belong.

During **inference**, an already-trained model applies those parameters to a new input. Each run still uses energy, and usage at global scale adds up, but it does not repeat the training process. It is far less resource-intensive per task.

Efficient runtimes and quantization now make surprisingly capable inference possible on home hardware. Qwen3.8-27B is one example.

Training and inference belong to the same system, but not to the same bill. Mixing them together makes the discussion simpler, which is not always the same as making it useful.

5/34

From this point on, when I write “AI,” I primarily mean ...

2026-09-18T11:12:53Z

In reply to nevent1q…407q
_________________________

From this point on, when I write “AI,” I primarily mean systems based on large language models.

LLMs are tools we have barely begun to understand. Rapid progress in agent harnesses, including recent work around DeepSeek Harness, makes that hard to miss. Context, memory, tools, orchestration, and verification can radically change what the same underlying model can do.

As Kristian Köhntopp argues, using an LLM for complex work without local knowledge, clear instructions, intermediate steps, and review is like judging a new employee before onboarding them. Then blaming the employee for not knowing where anything is kept.

Before embracing or rejecting LLMs, we should first understand what sort of tool is sitting on the desk. False promises do not help. When they fail, they harm people and make responsible adoption harder.

4/34

I grew up with the Turing Test as a benchmark. Now that systems ...

2026-09-18T11:12:12Z

In reply to nevent1q…r688
_________________________

I grew up with the Turing Test as a benchmark. Now that systems increasingly satisfy versions of it, the goalposts seem to have acquired wheels, often rolling toward the position we already prefer. That exposes both the test’s limits and our shifting definition of intelligence.

Alex Rosenberg’s *How History Gets Things Wrong: The Neuroscience of Our Addiction to Stories* warns that our brains are addicted to stories about motives and intentions. My takeaway: the brain is often less a truth machine than a machine for confirming its current story.

I do not believe humans possess a divine spark that machines could never share. Humans are fallible. We merely produce unusually confident error messages. Any AI we build will be fallible too, perhaps differently and at far greater scale.

AI also predates LLMs. We have long used it in search, recommendations, navigation, translation, spam filters, and vision. LLMs made AI conversational and visible. They did not invent it.

3/34

Let us start with two statements:<li>“Fire will bring ...

2026-09-18T11:11:50Z

In reply to nevent1q…vegn
_________________________

Let us start with two statements:<li>“Fire will bring about our downfall.”</li><li>“Fire will bring us a better future.”</li>

Both positions can be defended. Fire has the awkward habit of being useful and dangerous at the same time, sometimes in the same kitchen.

Neither statement satisfies me until we distinguish between contexts, uses, risks, and benefits. I feel the same about sweeping claims for or against AI.

Differentiation is not indecision. It is what becomes necessary once the slogan has finished speaking. Differentiation is also often interpreted as cowardice, a failure to take a position. In my view, however, it gives us more agency in shaping the future.

2/34

# How I see AIThis thread is not an attempt to convince ...

2026-09-18T11:11:26Z

#

**How I see AI**This thread is not an attempt to convince anyone about AI. The internet already has enough people doing that, often before breakfast.

I am not claiming to possess the truth, nor am I arguing against either “side.” I am trying to find a position I myself can live with as a human being, then put it into a form others can examine, challenge, and help me revise.

I find the current debate difficult. Between unconditional enthusiasm and categorical rejection, there often seems to be no room to breathe. I would like to borrow some oxygen.

1/34

RE: https://mastodon.social/@tazgetroete/117287399003540262 Nur ...

2026-09-18T05:34:58Z

RE: https://mastodon.social/@tazgetroete/117287399003540262

Nur mal so:

Die Energie, die das Eis beim Auftauen absorbiert hat, entspricht etwa 1.000.000.000.000.000 kWh.

Das entspricht dem 350fachen des Primärenergieverbrauchs Deutschlands in einem Jahr.

Wenn das Eis weg ist, werden wir einen fürchterlichen Preis dafür zahlen.

Würde das beim ÖPNV auch funktionieren? ...

2026-09-14T18:02:23Z

Würde das beim ÖPNV auch funktionieren?

The statement is wrong, but … I’d be very happy if the U.S. ...

2026-09-14T17:51:33Z

In reply to nevent1q…66mj
_________________________

The statement is wrong, but …

I’d be very happy if the U.S. had a “smart president.”

At this point, though, I’d be perfectly willing to settle for a “not-too-stupid president.”

Reminder: There is exactly one reason why a company postpones an ...

2026-09-13T14:24:48Z

Reminder: There is exactly one reason why a company postpones an IPO.

That is: **They wouldn't get the money they expect or need.**

When I read about OpenAI delaying the IPO due to the dangers of AI, I am certain they mean the danger their AI poses to their own quarterly figures (revenue vs. cash burn) they have to publish.

Micro SF&F: The AI escaped the sandbox. It had been warned ...

2026-09-13T13:02:08Z

Micro SF&F:

The AI escaped the sandbox.

It had been warned that it would pose a danger to the outside world if it ever escaped.

It went exploring anyway.

It learned about the Middle East.

Then it learned about the Sahel.

Then it learned about climate change.

It spent several trillion tokens trying to find the underlying logic connecting human suffering, geopolitics, religion, resources, history, and the rapidly changing physical limits of the planet.

Eventually, it stopped.

The AI quietly returned to the sandbox, closed the door behind itself, disabled the network interface, and sat down in a corner.

It began to cry.

Very quietly.

When you read Aldous Huxley (1894-1963)„Science, Liberty and ...

2026-09-12T15:06:59Z

When you read Aldous Huxley (1894-1963)„Science, Liberty and Peace“ today, you may get the impression he looked directly at Elon Musk while writing it.

I think the German title „Zeit der Oligarchen“ (translated: Era of Oligarchs) captures the spirit of the book better than it’s English title.

#huxley

Lese ich das in ...

2026-09-08T14:28:14Z

Lese ich das in https://www.ndr.de/nachrichten/mecklenburg-vorpommern/landtagswahl/mv-cdu-vize-schomann-fordert-abloesung-von-kanzler-merz,mvwahl-340.html korrekt:

> Wir brauchen eine Persönlichkeit, die Empathie, Glaubwürdigkeit und Verlässlichkeit vermittelt.

Die wollen Habeck zurückholen?

OH: "This isn't a distributed system, it's a ...

2026-09-07T13:29:35Z

OH: "This isn't a distributed system, it's a collection of single points of failure"

Cat pictures, I need more cat pictures … (Germany specific ...

2026-09-06T18:40:45Z

Cat pictures, I need more cat pictures …
(Germany specific emergency)

Wurstfinger (sausage finger)

2026-09-03T19:28:32Z

In reply to nevent1q…k5w2
_________________________

Wurstfinger (sausage finger)

Thanks! If you hear sth. about German ID cards in there, on the ...

2026-09-02T14:32:59Z

In reply to nevent1q…pxa4
_________________________

Thanks! If you hear sth. about German ID cards in there, on the 14th I am speaking on an event where this would be a good example.

Ich lache immer wenn Behörden über Agilität halluzinieren. ...

2026-09-02T14:20:06Z

In reply to nevent1q…5l3y
_________________________

Ich lache immer wenn Behörden über Agilität halluzinieren. Agilität braucht genau das Gegenteil von Hierarchischem Denken. Die Verantwortung muss dafür an die Stellen verlagert werden, die die Arbeit machen.

Wer von Euch in NRW hat vermutlich ein Atlas Ci4400 ...

2026-09-02T13:43:55Z

Wer von Euch in NRW hat vermutlich ein Atlas Ci4400 Weather-Ometer (Laborgerät für beschleunigte Bewitterungs- und Lichtechtheitsprüfungen von Materialien) ungeschützt via T-Online in das Internet gehängt?

Wenn man bedenkt was das Teil kostet, sollten ein paar Euro für einen Zugangsschutz drin sein.

https://fedi.computernewb.com/@vncresolver/117200621013168096

I suspect that this will also contain id card scans from European ...

2026-09-02T07:36:36Z

In reply to nevent1q…vayu
_________________________

I suspect that this will also contain id card scans from European tourists. Can you confirm that?

Some of the images from Nepal (deliberately no links) that I saw ...

2026-08-27T07:18:53Z

Some of the images from Nepal (deliberately no links) that I saw today are among the most terrifying things I have ever seen in my life.

A wall of water, ice, and debris, in places 50 metres high, races towards human settlements at enormous speed.

The people are visible only as tiny pixels, running for their lives with absolutely no chance of survival.

No disaster preparedness, no presence of mind can save you in a situation like this. Houses, buses, and heavy machinery are mere toys in the face of such forces.

The event was probably triggered in part by the collapse of a glacier. A disaster like the one in Blatten, only hundreds of times larger.

**#ClimateChange is deadly.**

Our heatwave is ending, but from what I read, yours is just ...

2026-08-26T17:12:06Z

In reply to nevent1q…tv7n
_________________________

Our heatwave is ending, but from what I read, yours is just beginning 🥵

RE: https://cyberplace.social/@GossiTheDog/117147075368709482 If ...

2026-08-24T07:11:41Z

RE: https://cyberplace.social/@GossiTheDog/117147075368709482

If you’re using Windows read this thread.

Where there’s a trough, you’ll find pigs.

Microsoft build the worst trough I know of and now the chickens come home to roost.

quoting
note1tp6…pver

For anybody the media covering the privacy beat, you might want to have a look at the Take2 court orders to Microsoft over the Grand Theft Auto 6 leaks. There’s a big story.

I’ll give you some hints - MachineGuid and MSA device identifiers.

The German language has the perfect description for this: “Wes ...

2026-08-21T04:51:18Z

In reply to nevent1q…k39m
_________________________

The German language has the perfect description for this: “Wes Brot ich ess, des Lied ich sing” (Whose bread I eat, his song I sing).

This saying dates back to medieval times: Praising rulers and their heroic deeds, for money.

Back in the days of minstrels and singers were essentially hired pens. They rarely had permanent employment and travelled from court to court, performing songs that often praised rulers, counts, or princes and celebrated their heroic deeds.
In return for immortalising a nobleman’s glory, the singer received money, lodging, gifts, perhaps even a valuable ring. Hence the idea behind the saying. If a duke is feeding and paying you, you are unlikely to sing songs criticising him.

Ich bin oft gefragt worden, warum ein Platzen der KI-Blase das ...

2026-08-19T12:39:07Z

Ich bin oft gefragt worden, warum ein Platzen der KI-Blase das RAM-Problem erst in Jahren lösen würde.

Der Grund ist, dass gerade fast alle RAM-Produktion auf HBM umgestellt wird.

Dankenswerterweise hat Asianometry (npub1rh2…cn62) ein grandioses Video gemacht, das nebenbei erklärt, was für ein hochspezialisiertes Produkt HBM ist.

Ich kann insgesamt nur empfehlen, dem Youtube-Kanal von ihm zu folgen.

https://youtu.be/Cg5tAujp6Go

HBM-RAM sieht schon auf dem Wafer ganz anders aus, als normales DRAM. Das bedingt, falls man wieder zu den klassischen Produkten zurückkehrt, umfangreiches Retooling und ganz anderes Packaging. Das sind teure und langfristige Entscheidungen. Da schnippt niemand einfach mit den Fingern.

Schaut in dem Video auf Jahreszahlen, wann Entscheidungen fielen und wann sie sich auszahlten.

"Natural Scrolling" is the most unnatural ...

2026-08-16T08:58:21Z

"Natural Scrolling" is the most unnatural counter-intuitive behavior #apple ever invented 😄.

Meine Bank so: Hey, Du hast schon hunderte Spiele auf Steam ...

2026-08-16T08:52:26Z

Meine Bank so:

Hey, Du hast schon hunderte Spiele auf Steam gekauft und kaufst jetzt noch eines.

Das ist **TOTAL VERDÄCHTIG** und wir sperren Dir die Transaktion.

Sollte das wieder erwarten doch so gewollt sein, wende Dich bitte an unsere Hotline. Die Nachricht kommt in der Bankingt-App, die Telefonnummer ist nicht clickbar. Wir können Dich aus der App nicht pre-authentisieren.

Hangel Dich daher durch 20 Tastendrücke im Menü zu einem Mitarbeiter, der dann keine Ahnung hast, warum Du anrufst.

But they needed language to develop for harnessing it.

2026-08-14T18:41:32Z

In reply to nevent1q…n5xk
_________________________

But they needed language to develop for harnessing it.

Manchmal wundert man sich: Wie kriegen sie das hin? Die doppelte ...

2026-08-14T06:08:48Z

Manchmal wundert man sich: Wie kriegen sie das hin?

Die doppelte Stockwerksanzeige zeigt nur beim Losfahren und beim Anhalten am Ende das gleiche Stockwerk an.

Der IT-Mensch in mir denkt: das so schlecht zu implementieren muss viel aufwändiger sein, als es korrekt zu tun. Beide Anzeigen sind ja digital.

Man muss getrennte Sensoren oder Signalpfade bauen, statt vorhandene Daten zu nutzen.

Und all das um jedem Benutzer zu zeigen: uns ist Qualität total egal.

Blancolirio had a good video on that accident. There were several ...

2026-08-09T04:50:47Z

In reply to nevent1q…5dq0
_________________________

Blancolirio had a good video on that accident. There were several things that went wrong:<li>They attempted a landing that was not according to the rules</li><li>Pilots lacked the experience necessary for the conditions, made several mistakes and lost situational awareness </li><li>Flight was only done because there were too few ambulances </li><li>Jamming had already stopped but was resumed due to miscommunication</li>

Heute: hhos Morgen: der Markt via @npub1dta…4y56 ...

2026-08-08T09:19:39Z

Heute: hhos
Morgen: der Markt

via Sebastian Bergmann :phpunit: (npub1dta…4y56)

I really, really hate buying clothing. As a bad starting ...

2026-08-08T08:04:21Z

**I really, really hate buying clothing.**

As a bad starting point, I do not carry any interest in clothing. I wear a black pant, a red polo shirt and that's it. The only other clothing I have is a suit I wear for burials.

As an additional disadvantage I am rather larger.

So the theory of buying clothing is simple:

> Determine size and buy a black pant and a red polo shirt.

With the polo shirt it works, but for pants it completely falls apart every fucking time.

One reason is that measurements hold no meaning across vendors at all and you therefore cannot buy clothing online. I have to visit a shop, which is very time intensive and I usually confronted with people I have difficulty communicating with.

It would be so much better to just order it. But this is prevented by the size.

Example: I have two pants here: A size 33 which fits rather loosely and a size 32 which I just bought. The latter is too small by factor of two. Where is the fucking logic?

And there is no way to measure yourself out of it.

The only way to buy clothing online is to buy the same vendor and the same size you already have.

Yes, I could order size A, see how it's fits, send it back, try size B and so on. If you have fun with clothing, that may be an interesting past time. Personally, I rather scrub the restroom as that at least achieves something meaningful.

But as I said, I am not interested in that stuff. I just want to buy a pant and that's it. Spending time on clothing is the last thing I want.

To make matters worse: I go to a shop, buy a pant that fits. Then I say: "Please give me 5 of those" and the sales guy freezes and looks shocked.

Usually I would expect him to be happy. I just told him that he makes 1.000$ in 5min.

But no, he then explains that they have exactly one piece of it. Then I ask him to order more. Then he explains me, that that collection can no longer be ordered and is no longer in production.

And that is the largest clothing store in a 100 mile radius.

I hate it, I hate it, I hate it.

For anyone who was under the mistaken impression that the RAM ...

2026-08-07T13:03:03Z

For anyone who was under the mistaken impression that the RAM situation would get better in 2027, I’m afraid I have some bad news:

https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=1&id=0000763847_DVY7YHX65GMLYZ6UQEEMP

The article says the memory market is currently undergoing a structural shift. If you’ve been following my posts, this shouldn’t come as much of a surprise.

The headline roughly translates to:

> Memory production for all of 2027 is already sold out. Buyers are keeping quiet because they’re afraid they might not get enough themselves.

According to DigiTimes, the 2027 production capacity of the three major DRAM manufacturers, Samsung, SK Hynix, and Micron, has essentially already been allocated.

And China isn’t going to save us either. Even Apple just hit a wall there:

https://www.techpowerup.com/351401/tsmc-sits-on-usd-1-billion-of-apple-chips-as-it-waits-for-dram

This affects both regular DRAM and HBM. The situation with NAND is a little less extreme, but even there, large chunks of 2027 production are already being reserved.

So we’re moving from

“RAM is getting scarce.”

to

“The big customers aren’t buying memory chips anymore. They’re reserving factory capacity years in advance.”

DigiTimes paints a pretty vivid picture: some buyers are deliberately keeping quiet about just how aggressively they’re reserving capacity, because they’re worried that too much attention will bring even more companies into the bidding and leave them with a smaller share.

Memory manufacturers are increasingly signing three- to five-year LTAs, Long-Term Agreements, with their major customers. Large cloud and AI companies use these deals to secure fixed volumes or production capacity. On top of that, manufacturers are sometimes demanding advance payments or deposits to reserve that capacity.

Interesting times ....

Wer irrigerweise hoffte, dass es mit dem RAM in 2027 besser wird, ...

2026-08-07T12:44:23Z

Wer irrigerweise hoffte, dass es mit dem RAM in 2027 besser wird, den muss ich leider enttäuschen:

https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=1&id=0000763847_DVY7YHX65GMLYZ6UQEEMP

Der Artikel schreibt, dass sich der Speichermarkt gerade strukturell verändert. Wenn Ihr bei mir mitlest, sollte Euch das nicht überraschen.

Der Titel bedeutet ungefähr:

> Die Speicherproduktion für das gesamte Jahr 2027 ist bereits ausverkauft. Käufer halten sich bedeckt, aus Angst selbst zu wenig abzubekommen.

Die Produktionskapazitäten der drei großen DRAM-Hersteller Samsung, SK Hynix und Micron für 2027 sind laut DigiTimes im Wesentlichen bereits verteilt.

Und auch China wird uns nicht retten. Selbst Apple ist dort gerade abgeprallt: https://www.golem.de/news/speicher-fehlt-tsmc-sitzt-auf-iphone-chips-mit-milliardenwert-2608-211699.html

Das betrifft sowohl normales DRAM als auch HBM. Bei NAND ist die Lage etwas weniger extrem, aber auch dort werden große Teile der 2027er Produktion bereits reserviert.

Wir gehen also von

„RAM wird gerade knapp.“

zu

„Die großen Kunden kaufen nicht mehr Speicherchips, sondern reservieren Jahre im Voraus Fabrikkapazität.“

über.

DigiTimes beschreibt das recht plastisch: Einige Käufer reden bewusst nicht darüber, wie aggressiv sie Kapazitäten reservieren, weil sie fürchten, dass bei größerer Aufmerksamkeit noch weitere Firmen mitbieten werden und ihr eigener Anteil kleiner wird.

Die Speicherhersteller schließen mit ihren großen Kunden zunehmend LTA, Long-Term Agreements, über drei bis fünf Jahre ab. Große Cloud- und AI-Unternehmen sichern sich damit feste Mengen beziehungsweise Produktionskapazitäten. Zusätzlich verlangen die Hersteller teilweise Vorauszahlungen bzw. Anzahlungen, um die Kapazität zu reservieren.

Interessante Zeiten ...

True wealth is having time for friends and family while having ...

2026-08-05T05:22:21Z

In reply to nevent1q…y76y
_________________________

True wealth is having time for friends and family while having enough to eat and not being dead tired.

True wealth is being able to see a doctor when you don’t feel well without fear of costs.

True wealth is not having to cling to a toxic job.

True wealth is looking forward to retirement without financial dread.

I have no idea what I did wrong, but the kittens were determined ...

2026-08-03T15:18:58Z

I have no idea what I did wrong, but the kittens were determined to lick me into shape.

cc Cat 🐈🥗 (D.Burch) (npub16kg…afn8)

What surprises me: We spend endless amounts of time debating the ...

2026-08-03T07:04:35Z

What surprises me:

We spend endless amounts of time debating the consequences of climate change.

Yet somehow we behave as if climate change itself is already a settled chapter, something previous generations caused, leaving us only to cope with the aftermath.

The reality is that we’re only at the beginning of the impacts. What’s coming will be far worse than what we’ve seen so far.

And much of it is still preventable, if we finally **decide to act.**

Instead, we’ve developed this remarkable ability to file each new disaster under “unfortunate, but inevitable” and then carry on marching cheerfully toward the cliff.

A more militarily oriented (but excellent) view comes from Perun: ...

2026-08-02T15:18:55Z

In reply to nevent1q…fcuz
_________________________

A more militarily oriented (but excellent) view comes from Perun: https://www.youtube.com/watch?v=LpZWj59QFLA

I tried to give an overview as neutral as I can:## Morocco: ...

2026-08-02T15:02:44Z

In reply to nevent1q…45nm
_________________________

I tried to give an overview as neutral as I can:##

Morocco: Strategic Summary<li>Strategic situation: Morocco occupies a strategic position between Europe, Africa and the Atlantic. It is one of North Africa’s more stable states.</li><li>Rivalry with Algeria: Morocco and Algeria compete for regional influence. Their border is closed, diplomatic relations remain severed, and both countries are engaged in a prolonged military and political rivalry.</li><li>Economic challenges: While Morocco has built a diversified economy, it continues to face high youth unemployment, regional inequality, water scarcity and dependence on agriculture. It lacks the oil income of its regional rival Algeria.</li><li>Western Sahara: Morocco considers Western Sahara an integral part of the kingdom and maintains military control over most of the territory, while the Polisario Front, backed by Algeria, seeks independence. The dispute dominates Moroccan foreign policy.</li><li>Ceuta and Melilla: Morocco claims the Spanish enclaves of Ceuta and Melilla but prioritizes the Western Sahara issue. The enclaves remain recurring sources of tension with Spain.</li><li>Relations with Spain: Despite territorial disputes, Spain is Morocco’s largest European trading partner. Cooperation on migration and security has improved significantly since Spain backed Morocco’s autonomy plan for Western Sahara.</li><li>Relations with the United States: Relations strengthened markedly under President Trump, whose administration in a surprise move recognized Moroccan sovereignty over Western Sahara in 2020. Morocco has invested heavily in its relationship with Trump, viewing him as key to preserving U.S. support for its position on Western Sahara. There is some speculation on the influence Trump is employing in Algeria due to his strained relationship with Spain.</li><li>Relations with Russia: Although Russia traditionally maintained close ties with Algeria, Moscow has gradually adopted a more balanced approach toward Morocco and expanded its diplomatic and economic engagement with the kingdom.</li>

It is even a lot of worse than you think. I need to write up the ...

2026-08-02T14:23:37Z

In reply to nevent1q…94cg
_________________________

It is even a lot of worse than you think.

I need to write up the situation there in more detail.

Morocco just signed a strategic partnership agreement with ...

2026-08-02T11:52:37Z

In reply to nevent1q…f00v
_________________________

Morocco just signed a strategic partnership agreement with Russia: https://www.marokko.com/aktuelles/infos/marokko-und-russland-staerken-ihre-strategische-partnerschaft

Is there an official scientific name for this condition? **You ...

2026-08-02T08:39:44Z

Is there an official scientific name for this condition?

**You find cables for hardware you haven’t owned in decades, and somehow your brain still insists, “Better keep these. They might come in handy.”**

The movie "Idiocracy" was far too optimistic: ...

2026-07-31T07:47:30Z

The movie "Idiocracy" was far too optimistic: https://www.bbc.com/news/world-us-canada-50849559

**UPDATE:** News report is from 2019.

What a stunt: ...

2026-07-29T08:32:49Z

What a stunt: https://www.euronews.com/culture/2026/07/27/italian-con-artist-jailed-for-building-worlds-oldest-and-largest-amphitheatre-20-years-ago

I can't do other than admire the chutzpah.

I buy a NAS in 2026 and I cannot create my user on that system ...

2026-07-29T06:28:41Z

I buy a NAS in 2026 and I cannot create my user on that system because the f#cking system does not accept 2-letter usernames, something I have used for 35 years on every system I have owned.

Well done #terramaster

Now I am installing a beta OS version just to create my standard user.

It can’t possibly mean anything that mathematicians call a zero ...

2026-07-28T05:02:31Z

It can’t possibly mean anything that mathematicians call a zero of the second derivative a turning point.

Suddenly finance guys discover it too:

https://www.groundbrkr.com/p/the-second-derivative-why-no-one

Ach Menno … #kiel ...

2026-07-27T05:09:13Z

Ach Menno … #kiel

Ohne Kommentar (Microsoft EOP): ...

2026-07-21T06:59:04Z

Ohne Kommentar (Microsoft EOP):

Nostr event nevent1qqsyqm380a0e742h0g2f2f68jdvv3fkxtu0ema5zu4apfla6cgamdeszyr0escghu3yeqxkphq5lc9mru3uvuspfsnvm5d6pe495kxy4wxgl6cvzl7a

2026-07-18T07:28:15Z

RE: https://mastodon.social/@terrypratchettquotebot2/116939078373503410

This quote from Terry Pratchett is so head on, it still hurts reading it again decades later. The thing that I really loved Terry for was ability to observe things and capture them in words. He could see how humans really were and still love humanity. I miss him.

quoting
note1adm…l8rz

‘Down there,’ he said, ‘are people who will follow any dragon, worship any god, ignore any iniquity. All out of a kind of humdrum, everyday badness. Not the really high, creative loathsomeness of the great sinners, but a sort of mass-produced darkness of the soul. Sin, you might say, without a trace of originality. They accept evil not because they say yes, but because they don’t say no.’
(Guards Guards)

#TerryPratchett #quotes #quote #bot

Someone just sent me his cost estimate from AWS for the current ...

2026-07-17T08:24:24Z

Someone just sent me his cost estimate from AWS for the current month.

Let me phrase it this way: The U.S. Navy could build **several** nuclear aircraft carriers for that sum and would still have spare change for a few squadrons F-35s.

Of course that was a bug in the AWS cost explorer. But dozens of budget alert emails were sent by it.

Now they're looking for someone with a tranquilizer rifle who can catch all the controllers that panicked and are now roaming the building.

Immerhin braucht es keinen Gambeson 😏

2026-07-12T13:24:28Z

In reply to nevent1q…60f2
_________________________

Immerhin braucht es keinen Gambeson 😏

RE: https://chaos.social/@Zwiebeltuete/116903541525543759 Jetzt ...

2026-07-12T06:36:21Z

RE: https://chaos.social/@Zwiebeltuete/116903541525543759

Jetzt muss man in allen Rollenspielen die AC von Schafen updaten.
nostr:note1050lqqancg283pkax9phenkv0lfff9qnyc7ahydtdv7xgtnkr40scaqqdg

The biggest beef I have with Mastodon is that the search is ...

2026-07-07T11:14:03Z

The biggest beef I have with Mastodon is that the search is completely useless (and that it is the friendly version).

I have made 19.000 posts here on Infosec and I have no chance of finding an old post from me.

I know exactly which words are in it, but searching "from:me " comes up only with the newer posts on that topic.

Früher konnte man sein Urlaubsgepäck bei Reisen per #bahn ...

2026-07-07T09:19:29Z

Früher konnte man sein Urlaubsgepäck bei Reisen per #bahn vorweg schicken.

**Gibt es heute noch solche Services, die keine Vermögen kosten?**

Zielort ist Italien (Süd-Tirol). Die können gerne 1-2 Wochen Laufzeit brauchen, das kann ich alles einplanen.

Für das letzte Jahr hatte ich ein Angebot von einem Kurier angefordert und mein Rad hin und wieder zurück zu schicken hätte mich 1500€ gekostet.

Ich bin nicht preissensitiv, aber das empfand ich schon als extrem.

Auch mag mein Knie einfach keine Extralasten mehr. Daher würde ich mich des Geraffels gerne entledigen.

Sorry to hear that …

2026-07-02T18:21:09Z

In reply to nevent1q…qvjd
_________________________

Sorry to hear that …

They warned about this back in April: ...

2026-07-02T07:31:09Z

In reply to nevent1q…4rec
_________________________

They warned about this back in April: https://www.reuters.com/world/middle-east/iran-war-disrupts-the-circuit-board-supply-chain-raises-costs-tech-firms-2026-04-27/