quasselkopf.de is a single user Mastodon where the sole user is heavily using an LLM to generate posts. They're using AI generated concern trolling to annoy and harass people.

Here's an example of them bothering someone else with AI generated replies:

https://quasselkopf.de/@kranzkrone/116176600327746998

Here's where they began doing it towards us:

https://quasselkopf.de/@kranzkrone/116176504616907351

We posted a constructive reply with useful information so they doubled down on trolling. We called it out and got targeted with abuse.

#fediblock

@GrapheneOS
*shows middle finger*; "Criticism here isn’t hostility"

@GrapheneOS I'm not convinced that the posts are LLM generated, though. There is an LLM translation of content that he's linking to in a recent post, but there's also a typo in a reply to you.

However, I remember visiting this profile before because I thought a post was fishy, but didn't take a note of course.

@scheme We didn't say all of his posts are LLM generated. The first and second reply to us are LLM generated. The third one after we called out the use of an LLM to generate concern trolling replies is an entirely different in terms of writing style, tone and content. If you look through his overall feed you can see a huge portion of the posts are generated with an LLM generating them in a certain style. Many of the posts are not generated in the same way and we weren't saying it's a bot.
@scheme As further evidence the posts written in that incredibly unnatural writing style are LLM generated you can see he has many posts about using LLMs which aren't written in that style. It's quite clear to us that he's largely using an LLM to generate replies while also writing them himself. He doesn't appear to mix those together at least outside of rare cases but rather they're one or the other. It's either LLM generated slop in that specific style or different manually written replies.