Wieviel gefühlten/praktischen Nutzen bringt das Ding denn gegenüber einem Rückspiegel?

Ah, habe gerade den Unterschied zwischen nominalem und realen Bruttoinlandsprodukt gelernt. Letzteres ist inflationsbereinigt.

Halbdumme Frage: steigert Inflation das Bruttosozialprodukt?

Zu niedrige Inflation.

Abbildung 2 aus der verlinkten Pressemitteilung. Zwei Karten von Deutschland (links für Diesel, rechts für Benzin). Gelbere Bereiche bedeuten niedrigere Senkung des Preises, violettere bedeuten höhere Senkung. In der linken Karte ist die südliche Hälfte Deutschlands gelblich, die nördliche violett. In der rechten Karte ist fast nur Bayern gelblich, der Rest violett.

Interessant: der Tankrabatt wurde 2022 regional sehr unterschiedlich weitergegeben. Insbesondere in Bayern ist nur die Hälfte der Entlastung bei den Tankenden angekommen - im Norden mehr als 100% (was mMn auch ein Artefakt der Messmethode sein kann).
www.rwi-essen.de/presse/wisse...

…
So I'm going to experiment with constraints that both make it harder to use LLMs and force them to really think hard about architectural optimizations.

Yeah, that's what I'm already doing. My main issue is that any "explain this piece of code" interview only happens after the students have finished their project. Sure, I can give them bad grades for bad work - but I'd prefer that they learn a lot and get a good grade.
…

Nice! Automated local backups of anything that is important enough to print.

…
On the other hand,
- LLMs might actually work better,
- students might be overwhelmed by the limits, and
- maybe it is more important for students to learn how to LLM-write much code than to write little code themselves.

Do you have any suggestions?

…
So, I might try out another thing next time: a hard limit on code size, number of dependencies, complexity metrics, etc.
My hope would be that students will be required to think harder and understand all the code, and that LLMs will have a harder time working with such dense code.
…

…
In some cases this is due to over-engineering and auto-generating code (DB wrappers and migrations and views) but in other cases it seems to be due to (illegal) LLM "support".
Of course, this might come out when we do the 1:1 code review together but then it is too late for the student to learn.
…

…
Not really hard but a few weeks of work for a student who has never done this before.
Typically, 1000 lines of Python are sufficient for the backend, but recently a few students have submitted 10 times as much code.
…

Oh, thanks for posting this. It gave me the push to rethink an exercise for a software development course I teach.

For the final project, students are required to build a small, relatively trivial piece of software (e.g., an asset management system) based on a few pages of requirements.
…

Heute ist Sonntag, da hat der natürlich zu.

Time series showing Southwest U.S. maximum temperatures from 1895 to 2026, which is a long-term warming trend and record for 2026.

Sooooo, the Southwest U.S. was more than 5°F warmer than the previous March record high for maximum (daytime) temperatures.... 🫠

Graphic from www.ncei.noaa.gov/access/monit...

Oh, that's just Mr. Trump's 4D chess strategy.
He is using reverse psychology to oust a Kremlin ally. There is no other plausible explanation.

Im Ernst: ich glaube, dass die Security-Leute bei MS grundsätzlich ethisch handeln, und dass es für MS im eigenen Interesse liegt, auch Linux sicherer zu machen.
Aber für den Vertriebler ist das natürlich ein Argument: "Bei MS ist alles auf Sicherheitslücken gescannt worden - bei Ihrem Linux auch?"

Vor allem jetzt, wo Microsoft Zugriff auf Anthropics Mythos hat, ...

"Schönes Linux-Ökosystem haben Sie da. Wäre ja schade, wenn jemand da Zero-Days suchen würde..."

Could it be that you just notice this more easily the more of it you read? Once you've noticed certain things ("It's not X, it's Y", etc.) you can't unsee them.

I'd be more concerned if the LLM-generated texts felt better and better written even if they are not.

@danieldrepper.bsky.social @hatr.bsky.social
Gibt es evtl. im deutschsprachigen Raum oder generell irgendwo eine Handreichung für nicht-journalistische Rechercheure, wie sie sicher/produktiv/befriedigend mit Journalisten zusammenarbeiten?

bsky.app/profile/raph...

FWIW, here's a scientific paper on the general topic:
journals.sagepub.com/doi/pdf/10.1...

It cites a master's thesis by Josie Cochrane on motivations of non-professional OSINT researchers:
www.diva-portal.org/smash/get/di...

I'd assume that something like this already exists somewhere - but I actually couldn't find anything that's close to what I want.

I'm certainly not the most qualified person to write such an explainer, so maybe someone else might want to do it. I'd be happy to contribute to any efforts.

So, what I had in mind was a small document that helps OSINT researchers and journalists understand where there might be different expectations, and how they can successfully collaborate.
Something like this would have helped me (and still might).

Another thing: because precision and correctness is important for me as a scientist, I want to make sure that an article that incorporates my research mentions uncertainties and alternative explanations. Some journos see researchers more like a source than a collaborator, however.
Etc.

On the other hand, some investigative journalists don't seem to understand the motivations of OSINT researchers - which are certainly quite diverse. For example, I don't usually want credit - the most interesting collaborations are when I get access to confidential information and can learn.

Yeah, I agree about the obvious essentials. What I'd love to read (and base collaborations on) would be some document that makes the different motivations and constraints more explicit. E.g., it took some time for me (as a scientist) to grasp the importance of exclusivity and speed for journos.

Is there any best-practice guide for OSINT research collaborations between journalists and private (or university) researchers? E.g., outlining motivations and constraints. I have only had pleasurable experiences so far but also rejected requests that felt like one-way streets.

Frankenthal-class minehunter - Wikipedia On 21 February 2007, Grömitz ran onto a reef in the Floro fjord while on tour in western Norway and remained stranded in a spectacular way until being salvaged.[4]

FWIW, the German navy has 10 mine-clearing ships - maybe Chancellor Merz had them in mind when he mentioned that Germany would help making the Strait safe again.
One ship was sold to the United Arab Emirates in 2006 but sunk by Huthi rebels in 2017 (says Wikipedia).

en.wikipedia.org/wiki/Franken...

Frankenthal-class minehunter - Wikipedia

BTW, the German navy has 10 mine-clearing ships. I guess Chancellor Merz had them in mind when mentioning that Germany would help making the Strait safe again.
One ship was sold to the United Arab Emirates in 2006 but sunk by Huthi rebels in 2017 (says Wikipedia).

en.wikipedia.org/wiki/Franken...

I always check hormuzstatus.com - the site is obviously broken, but you know that saying: a broken dashboard shows the correct status of a broken system twice a day.