THIS. Did you know sickle cell is curable, now? FULLY curable. If you're feeling down or despairing about what FEELS like technology/advancement/science "standing still" or being fully malicious, I have great news: You're not paying attention to actual science. Just tech industry marketing.

But to be honest when I actually watched MTV youtu.be/8SbUC-UaAxE

#39c3 Vortrag @nicosemsrott.de was PRÜF! anders macht als bisherige Kampagnen #Pruef „Alle Parteien, die vom Verfassungsschutz als rechtsextremer Verdachtsfall oder gesichert rechtsextrem eingestuft werden, sollen durch das Bundesverfassungsgericht überprüft werden.“ media.ccc.de/v/39c3-pruf#...

Nur noch 16 Tage. Es sind gerade mal knapp unter 7000 Unterschriften für die Petition gegen ein Register von Trans- und nichtbinären Personen. Das Quorum liegt bei 30.000. Bitte unterschreibt. Bitte Rehpost🦌💌! #trans #queer #register epetitionen.bundestag.de/petitionen/_...

Interesting talk. Thank you for that and the Signal app. media.ccc.de/v/39c3-ai-ag...

It's a fun concept at least pivot-to-ai.com/2025/11/24/d...

Abgebrannter Kältebus in Berlin: Deutsche Bahn Stiftung spendet 70.000 Euro an die Berliner Stadtmission zur Finanzierung eines neuen Fahrzeugs. www.deutschebahn.com/de/konzern/I...

Soon there will be LLM sommeliers. They can recognize not only the brand, but the vintage of slop produced by any LLM. "This has the bouquet of oil and fresh compost, like the bottom of a lawn mower. And I'm detecting a strong undertone of white nationalism. Grok 2025?"

Learning from this would mean AI pushers like Ilya and Lilian Weng actually bothered to read the history of their own field, so we’re doomed to suffer as they perpetually reinvent the wheel. It’s like Silicon Valley is some arcane punishment in a circle of hell, like Sisyphus, but for all of us.

Did we forget years of usability best practices, for the sake of AI hype? I really like how @vitalyf.bsky.social explains it, in his Design patterns for AI interfaces course (and tend to agree): “I think that very often we kind of get a little bit too excited about the hype again...

Nachdem das französische Skigebiet Céüze den Betrieb einstellen musste, erobert die Natur das Gebiet zurück. Die Skilifte wurden abgerissen. Sträucher und Bäume beginnen wieder zu wachsen. ⛷️⛷️⛷️ Es gibt zu viele negative News, deswegen poste ich jeden Tag eine #guteNewsdesTages.

"Never tried therapy before but this is probably it?" 9 words summarising the attitude of Big Tech to all things of which they have no knowledge or experience - with GenAI, they can do [thing] without [thing] experts or even educating themselves on [thing]. My eyes rolled a grand total of 720° 🙄

IHR SEID DER WAHNSINN! Unser zweites Ziel ist erreicht und unsere neue Redaktionsstelle ist damit für 2026 finanziert! DANKE! Das Crowdfunding läuft noch bis morgen Abend - falls wir noch das letzte Ziel erreichen, überlegen wir uns was ganz besonderes für euch 😁 👉 crowdfund.volksverpetzer.de

"Checking the output is not sufficient as humans are notoriously poor at checking automated processes while the overhead of checking bullshit is often the same or even higher than creating the underlying content." This is true in all sorts of contexts beyond teaching and computer science.

A good essay on this aeon.co/essays/your-...

Hatten wir alles schon: Zahl d Arztbesuche wurde nicht verringert; Geringverdiener wurden v notwendigen Arztbesuchen abgehalten; es belastete einseitig Patienten; Zusatzeinnahmen waren gering u deckten nur knapp 1% der Ausgaben der GKV; verursachte unnütze Bürokratie de.wikipedia.org/wiki/Praxisg...

Wir HATTEN bereits die Praxisgebühr. Wir WISSEN, dass sie soziale Ungleichheit verstärkt, aber den eigentlichen Effekt nicht erreichen wird. Warum müssen wir die immer gleichen sinnlosen, kontraproduktiven Schnapsideen immer wieder durchkauen?

Apparently I'm doing this -- my open letter to that open latter about AI in writing and publishing. I emerged from Hibernation Week to write it, so god only knows how much sense it makes. But at least a human wrote it, so that's nice. Bonus: picture of my "cat," sweet baby Boomba.

By any reasonable historical standard — including that of the technology industry! — ChatGPT should be pulled from the market and its product managers and executives held accountable for creating a product that ROUTINELY tells teens to kill themselves. This is a basic, common sense standard.

Remember: When Tylenol was poisoned *by an outsider* and killed people, the company recalled all their products & redesigned them. When Intel’s Pentium had a bug so obscure it affected 1 in _9 billion_ long division calculations, they recalled their chips. ChatGPT was made deadly *by its team*.

Letzte Nacht hat irgendjemand den Kältebus der Berliner Stadtmission angezündet. Falls ihr noch was spenden wollt und könnt: www.berliner-stadtmission.de/spenden/onli...

Hier kann man spenden ⬇️ www.berliner-stadtmission.de/spenden/onli...

nicht alt genug für social media, aber alt genug für den knast

Some of you have forgotten that only three years ago you were perfectly capable of writing a text, writing an email, telling a bedtime story to a child, and it should worry you that powerful companies have convinced us we can’t do things we’ve been doing since the dawn of time.

Consider a programmer looking at code they themselves wrote a year ago, now having to identify and fix a bug. They have to rebuild the mental model of what is going on, which is easier if they have already had a mental model of what is going on.

This is deskilling *the* most vital, essential parts of writing code. The part of writing code that actually makes software work, that supposedly justifies being paid programmer salaries.

The result is a goddamn nightmare for the phase of software development that accounts for the vast vast majority of any code's lifespan: maintenance. The code is unmaintainable. Even less maintainable than Mel's code, because at least Mel had a thought process, something llm's entirely lack.

When, not if, it has to be maintained, there is very literally no connection between the documentation (the plagiarism machine regurgitates what looks statistically similar to documentation) and the actual code (which was generated to be statistically similar to the average of all of Stack Overflow)

"These two equations are not the same equation" is a very large detail. It is made up of smaller details, like the individual characters used to express that equation. LLM's allow a programmer to avoid paying any attention at all to any of these details.

It's now possible for someone to shit out a whole program without paying attention to *anything*. Even large details, like the fact that the program isn't doing what they think it's doing, or the fact that it's trying to use libraries that don't exist, or that it's doing something a very stupid way

It's funny that everyone who has *ever* programmed, knows how important attention to detail is. Hours wasted because you made a typo, forgot to add a character, etc. The tiniest thing you literally don't notice is wrong, and your program screams at you and dies. This is a universal experience.

We recorded a podcast episode about the research on gender and math ability yesterday. I did so much research for it I felt overwhelmed trying to summarize it. These issues and discriminatory experiences are too massive -- like they destabilize you with their gravity.

From October 2025: “I might invoke Edward R. Murrow, but they remember him. They don’t have any goodwill toward you. They don’t know who you are at all. And they are not going to defend you when you screw up.” Read more from @lopatto.bsky.social:

The statistical averageness of AI is the problem. It requires higher levels of education and professional scepticism to spot when it is going wrong.

This is 404 Media's @jasonkoebler.bsky.social waving at himself through a Flock camera; one of 60 we learned was left exposed to the open internet. Not only could anyone with a link livestream it, but some admin portals were open with no login credentials required. www.404media.co/flock-expose...

CBS paid the president of the United States millions in bribe money and now they censor news reports so it does not annoy him. I don’t know how much plainer we can put this.

If you're not studying it, then it's not your area of study. This isn't remotely difficult to understand.

Schon bitter. Die Züge der Deutschen Bahn waren 2025 insgesamt länger verspätet als die Züge der anderen Länder zusammen. chuuchuu.com/2025wrapped

When I write a summary, the summary is not the work product. The work product is me having read it and put it into structure and making decisions about what to include at what level. The summary is a signifier of me having done that work.

There are always empty words about "human in the loop", that you should always check the work of an LLM. But the entire value of a summary is as a replacement for reading. Nobody uses an LLM to generate a "summary" *and* independently verifies.

Professional societies keep beclowning themselves buying into a lie about what an LLM "summary" is. They are inherently counterfeit: not an epistemic product of the ideas in the source, but summary-shaped text linguistically based on *other* works (in the training corpus) that use related language.

I'd like to also raise the alarm about this planned chat interface. This impulse brought us a slew of horrific whoopsies this year * chat with pacifist John Brown about compassion for enslavers * chat with Harriet Tubman about the upsides of slavery * chat with Anne Frank about not blaming the Nazis

Accessibility without verifiability isn’t clarity—it’s risk transfer. AI summaries don’t simplify truth; they abstract accountability. If readers can’t validate claims, the output degrades from aid to liability.

I don't know how many times this needs to be said, but apparently a lot: If you are not in a position to verify the accuracy of some synthetic text, the synthetic text is not useful/has a high potential to be misleading. This is one of those cases.

ACHTUNG: 28 heimliche Änderungen ‼️ Das Wirtschaftsministerium hat ein unabhängiges Gutachten zur Energiewende vor Veröffentlichung verändert, z.B. kritische Warnungen zu Gaskraftwerken gezielt abgeschwächt. Zur Greenpeace-Recherche 👉 www.mynewsdesk.com/de/greenpeac...

„We will soon add chat-style interactivity to our content and search functionality (…) we conduct regular audits for hallucinations and other errors.“ Besides that this is like whackamole: with novel papers where your ideas are not in the training data, how would that even work?

ACM shows LLM summary as default instead of the author written abstract „to enhance your research experience“ while warning „it cannot replace the experience of reading an article in full“ but „we do believe it will help you find, understand and use ACM content both more quickly and more deeply“ 🤬

I think it's also important to recognize that LLM "summaries" are not epistemically grounded in the ideas of the source material. When they're correct, it's more because *other* writing (in the training corpus) contains similar language. That makes it especially bad when applied to novel results.

*taps mic* The abstract is the summary of an academic paper. It doesn’t need AI slop summary. Well done ACM for making my work as a researcher *harder* as now I have to hunt for the actual abstract that humans had an input into writing.

Another good moment to ask "what problem are we solving" when we add an AI summary to articles that already include an abstract but you don't know it's an AI summary until you read to the bottom of it and also it shows up as the default?