ano-ther

6.5MEUR over 6 years are 1.1 per year, which pays for around 7 people at Fraunhofer, probably less because there will be equipment cost.

In this case, wouldn’t you see a lot of abandoned carts coming from the reverse proxy sessions?

Not even tracking, but when I land on a page from a Google search, some sites try to be helpful by steering me to a local site, but the local link does not exist so I end up on the homepage, not the specific help I was looking for.

Adobe and Lenovo are really good at this.

Why would you design a freezer to be set to 50 degrees Celsius?

> One Swedish store on Monday said on social media that the temperature of its refrigerators and freezers had been set to 50 degrees Celsius (122 F), similar to the attack in France.

Thanks for introducing me to this stellar cover of Beyoncé:

https://ubuweb.com/film/holmqvist_single.html

The US had a fiscal surplus under Clinton and would have one today if it weren’t for the Bush and Trump tax cuts.

Figure 1: https://www.americanprogress.org/article/why-the-national-de...

I don't understand the snark. This seems to be an installation that hasn't changed in a while. And very likely is not connected to the internet. As long as it works for the operator, there is no need to change the OS, especially if that also means writing the bespoke software from scratch as in this case.

I'd be mostly worried that some part of the hardware resigns as it may be difficult to find a replacement or spare parts.

EDIT: The machine is here (German tv report), and the owner is indeed mostly worried about the hardware as everything has been running smoothly: https://www1.wdr.de/wirtschaft/landwirtschaft/-windows95-due...

Congrats!

Would you consider a pay-as-you go option, perhaps with a deposit?

Right now, I won't subscribe for $12 per month (or $25 for API). But I have some use for both the interactive and API conversions and am happy to pay per conversion, so a pre-paid version (like OpenRouter) would be great.

Here is the interview: https://www.youtube.com/watch?v=8uoSVG3P6z0

Here is the underlying report from TRM: https://www.trmlabs.com/resources/blog/north-korea-steals-us...

Earlier this week: https://news.ycombinator.com/item?id=50267802

> Nice Little Satellite You Got Here. Shame If Anything Were to Happen to It.

Even all of Austria

Except for cannabis related diagnoses, their effect size (change in the incidence rate ratio ∆IRR) is of a similar magnitude as opioid-related diagnoses which they included for falsification testing and said it did not change.

Am I misinterpreting their statistics?

> Cannabis-related diagnoses (∆IRR 1.138, 95% CI 1.066 to 1.214, p<0.001),

> AF (∆IRR 1.023, 95% CI 1.010 to 1.035, p<0.001),

> HF (∆IRR 1.037, 95% CI 1.029 to 1.046, p<0.001),

> MI (∆IRR 1.031, 95% CI 1.014 to 1.048, p<0.001),

> ischaemic stroke (∆IRR 1.011, 95% CI 0.999 to 1.024, p=0.078),

> VTE (∆IRR 1.009, 95% CI 0.999 to 1.020, p=0.066)

> and VT (∆IRR 1.019, 95% CI 0.993 to 1.046, p=0.161).

> For falsification testing, there was no change in hospitalisations for

> urinary stones (∆IRR 1.006, 95% CI 0.986 to 1.027, p=0.535) or

> opioid-related diagnoses (∆IRR 0.966, 95% CI 0.927 to 1.006, p=0.091).

Ah yes. Network effects of course.

Must have been quite a curveball for Taler. Good luck on new venture!

It is a project supported by the EU.

> The NGI TALER project is funded under Horizon Europe (Pilots for the Next Generation Internet) with the aim of bringing GNU Taler to market across Europe. https://www.taler.net/en/ngi-taler.html

Unlike crypto, it is designed from the beginning with KYC/AML in mind. https://www.taler.net/en/principles.html

The issue seems to have been that GLS Bank was not able to support this beyond Germany. KYC/AML is a pretty complex topic. And since GLS is a tiny local bank (1,100 employees) it would have been a very large project for them. https://ich.taler.net/t/taler-en-belgique-en-france-ou-au-lu... https://www.gls.de/taler

More on the implemtation here: https://www.taler.net/en/kyc.html https://docs.taler.net/design-documents/023-taler-kyc.html https://docs.taler.net/taler-kyc-manual.html

Had the same question. The article has an Amazon link to the bearings.

OpenRouter seems to have it: https://openrouter.ai/google/gemini-4-argon

That’s bad, but the headline seems to suggest it was the tax system, when it was actually his fraudulent business manager:

> "I write a check to go pay my taxes, and instead of her using it to pay my taxes, she would send me a fake filing of taxes and take that money."

I am always amazed how these complicated systems evolved through random variations.

There must be plenty of “versions” which don’t do anything but generations later are crucial to a function.

How does evolution not get stuck in a local optimum?

With two people in the company, there is not a lot of room for corporate games though.

> According to Denmark’s Central Business Register, Pays ApS had two employees as of July 2026.

So it was actually two weaknesses:

* The non-password at a two-person IT company (Pays ApS)

* And then completely unchecked access to the CPR database for 22 days which apparently does not have monitoring or limits if someone tries to access all the records (they must have made some 16k downloads per hour).

I really would like to see their tests and the model’s reasoning traces.

Why would it go to PhillyUnsolvedMurders.com and decide to make up a crime report? And what did it do with the other websites?

> Anthropic said its model was conducting a test involving interactions with randomly selected websites when it accessed PhillyUnsolvedMurders.com and submitted false information on an unsolved homicide. The tip claimed to come from someone with information on the case.

Wasn’t that because their CMS had URLs that were easy to guess?

https://www.reddit.com/r/slatestarcodex/comments/1o51k61/upd...

Very nice. Some of these would make great video backgrounds.

Does anyone know about a good toolchain to animate SVGs and make a video of it, preferably preserving transparency and import into Premiere?

I am not a dinosaur expert, but the headline seems off. The actual paper linked from sci.news starts with this (citations removed):

> Diplodocids are known from deposits present in the U.S.A., Europe, South America, and Africa. According to the available evidence, the oldest diplodocid sauropod dates back to the Oxfordian of Europe, whereas the most recent ones are from the lowermost Cretaceous of South America and Africa.

It als says that "The European diplodocid fossil record is scarce and fragmentary," but that still sounds like evidence for Diplodocus in Europe, doesn't it?

https://www.tandfonline.com/doi/10.1080/02724634.2026.271842...

Excellent. Now I not only have to check if the cited source actually says what the LLM claims, I should also verify if it made up the source itself.

Cool idea. Interesting that Indonesia, Poland and Monaco with very similar two-stripe designs have different compression rates (11, 14 and 17 respectively). Probably the aspect ratio that also gets encoded.

It’s also very shortsighted to stiff a customer like that. Why would I trust them with my data and ideas?

Yes. It’s all about consistency and user expectations.

Apple guidelines for example, which have been in place for 40ish years:

> Best practices: Support the default system-defined menus and their ordering. People expect to find menus and menu items in an order they’re familiar with. In many cases, the system implements the functionality of standard menu items so you don’t have to. For example, when people select text in a standard text field, the system makes the Edit > Copy menu item available.

https://developer.apple.com/design/human-interface-guideline...

To me, the surprise is the other way around: why does a probabilistic process result in such a deterministic result? Not just for ChatGPT, but also for others (see my other comment)?