Also, students are already using LLMs to learn about topics. It's not that they necessarily need a "mandatory" tutoring session with the university's LLM assistant.

But at the same time, they need a reliable source of truth to start with. You don't want to wonder, whether your prof hallucinated.

It's okay for a lot of cases, but when it just makes up things and sells it as truth and students wouldn't know, wouldn't be able to spot the false data point.

I see potential given a very limited scope of engagement, as to prevent wandering off into the LLM dream land

My comment wasn't doubting the study, more that LLM systems aren't ready to replace teachers/experts as long as they keep outputting wrong information, which unfortunately is inherently part of how LLMs work.

Expertly wrong in many cases, luckily the students won't notice, because they're learning the subject

Where's that from? Or what's that for?

It's quite impressive. The German is very well understood. It does translate some technical terms, but other terms like "CreateWindow" or similar aren't translated as you reference them.
It's crazy how it also applies the speech patterns. Slowing down, speeding up, lengthening words, etc.

Thank you! And especially thank you for driving Material for MkDocs for so long! Looking forward to see where Zensical is headed 😊

Did they ask for treats?

For the past weeks I couldn't stop thinking about idea of a glue code library using other well-maintained libraries.

With the first draft, I want to enable converting between DocumentFormat.OpenXml (i.e. *.docx), MigraDoc and PDFSharp (i.e. *.pdf).

It's interesting even if not completely surprising that most PDF rendering solutions for .NET are only available as commercial products or with certain restrictions in the free options.