Tootfinder

Der Skandal ist nicht der Angriff, sondern ein System, das daran scheitert, ihn abzufangen. Signal ohne Regeln ersetzt keine staatliche Infrastruktur. Caspar Clemens Mierau hat in mancherlei Beziehung Recht, aber es ist schon bezeichnend, wenn jemand wie Klöckner u.a. auf einen simplen Phishing-Angriff reinfallen. Das zeigt etwas über deren Digitalkompetenz aus.
#Golem

Featuring headliners such as
Robert De Niro,
Minneapolis Mayor Jacob Frey
and journalist Don Lemon,
the “State of the Swamp” address
is set to continue through Trump’s address with live rebuttals.
Attendees were encouraged to dress in green frog attire
as a symbol of defiance,
honoring the frog costumes worn by many anti-Immigration and Customs Enforcement protesters during its occupation of the city.
It is also intended to reference the “…

State of the Union live updates: Epstein survivors to attend Trump’s address
The speech comes as the president faces an angry and divided Congress, plummeting approval ratings, and a major Supreme Court loss on his tariff policy.

Why Pass@k Optimization Can Degrade Pass@1: Prompt Interference in LLM Post-training
Anas Barakat, Souradip Chakraborty, Khushbu Pahwa, Amrit Singh Bedi
https://arxiv.org/abs/2602.21189 https://arxiv.org/pdf/2602.21189 https://arxiv.org/html/2602.21189
arXiv:2602.21189v1 Announce Type: new
Abstract: Pass@k is a widely used performance metric for verifiable large language model tasks, including mathematical reasoning, code generation, and short-answer reasoning. It defines success if any of $k$ independently sampled solutions passes a verifier. This multi-sample inference metric has motivated inference-aware fine-tuning methods that directly optimize pass@$k$. However, prior work reports a recurring trade-off: pass@k improves while pass@1 degrades under such methods. This trade-off is practically important because pass@1 often remains a hard operational constraint due to latency and cost budgets, imperfect verifier coverage, and the need for a reliable single-shot fallback. We study the origin of this trade-off and provide a theoretical characterization of when pass@k policy optimization can reduce pass@1 through gradient conflict induced by prompt interference. We show that pass@$k$ policy gradients can conflict with pass@1 gradients because pass@$k$ optimization implicitly reweights prompts toward low-success prompts; when these prompts are what we term negatively interfering, their upweighting can rotate the pass@k update direction away from the pass@1 direction. We illustrate our theoretical findings with large language model experiments on verifiable mathematical reasoning tasks.
toXiv_bot_toot

Google rolls out Gemini 3.1 Pro, which it says is "a step forward in core reasoning", for AI Pro and Ultra subscribers; the .1 increment is a first for Google (Abner Li/9to5Google)
https://9to5google.com/2026/02/19/google-announces-gem…

Google announces Gemini 3.1 Pro for ‘complex problem-solving’
In November, Google introduced Gemini 3 Pro in preview. Google today announced Gemini 3.1 Pro "for tasks where a simple answer isn’t enough."

The sales analyst leans back and types a question into her company’s shiny new AI data agent: “Which leads from last quarter should we follow up on?”
Simple enough. The kind of question that used to eat up half a junior analyst’s afternoon. The agent spins for a few seconds. Then it either returns a confidently wrong answer, crashes without a word, or does the thing I find most telling of all: it gives up before it even tries.
https://levelup.gitconnected.com/why-sql-agents-fail-62-of-the-time-on-real-enterprise-queries-5da9eeb48ede

from my link log —
U 237C ⍼ RIGHT ANGLE WITH DOWNWARDS ZIGZAG ARROW is a symbol for azimuth.
https://ionathan.ch/2022/04/09/angzarr.html
saved 2026-03-11 …

Power loom built by T. Larmuth & Co., in Manchester, around 1860 - now at the Manchester Museum of Science and Industry.
> "By the early 19th century, new machines like this power loom could make cloth more quickly and cheaply than people. Groups of angry handloom weavers raided cotton mills at night. They burnt and broke power looms to protest against the new technology."
This aspect of the Industrial Revolution is now being breathlessly repeated by those who s…

I don’t know the answer to my original question, but maybe it’s as simple as “seize the means of repetition.”
/end

If you think #vibecoding is fine, let me ask you a single question: would you use a medical device whose software was vibecoded? And by "medical device" I mean something where a bug could literally kill you.
If you answered "oh, gawd, no!" then consider that anytime you use an #LLM to contribute to or develop an #OpenSource project, there's a chance that this code will end up powering such a device. And even if it doesn't, you're setting a trend, and it will be even more likely that the software used by these devices will be vibecoded.
I have type 1 #diabetes. I also lead a physically active life. This is both a blessing and a curse. My doctors keep suggesting Constant Glucose Monitoring systems and insulin pumps to me. And I do realize that such hardware would likely improve my blood glucose, and definitely make my life much easier (especially with a closed loop system).
So why do my fingertips look like crap, and I keep using a glucometer and insulin pens? Because I don't want to risk my life to an unnecessarily complex technology.
Admittedly, I occasionally get things wrong and suffer consequences. Or I suspect I got them wrong and worry. Or meet an unexpected situation and need to figure out a way out. Or even accept having elevated glucose levels (as in nearing 200 mg/dl) because there's just no way to safely fit insulin doses on a particular day.
But still, I prefer having control and risking my own mistakes to a device that could suddenly start pumping insulin because of a bug. And that was even before the story of the application that stripped the decimal point and gave people ten times the dose. Or the one about CGMs giving wrong high glucose alerts. Or the whole vibecoding fancy.
Back then, I could have considered such a device. Now, I'm more worried than ever. And honestly, I'm hoping that relatively simple glucometers will remain available. To think that my worst fear used to be of a mechanical fault…
#AI #NoAI #NoLLM

“Asking an AI chatbot a question consumes a great deal more energy than finding the answer via simple web search or calculator. It adds extra demand for no good reason... a bit like driving to the shops in an SUV instead of riding your bike.”
Or driving any kind of car. AI, like cars, aims to reconfigure everyday life in a way that’s vastly more energy-hungry and polluting… and profitable for a few.