Post B1FbcpOSDpeHMboiw4 by Nicro@fedi.absturztau.be
 (DIR) More posts by Nicro@fedi.absturztau.be
 (DIR) Post #B1Fayepc4WGT1UMjvk by kaia@brotka.st
       2025-12-14T20:40:01.756797Z
       
       4 likes, 0 repeats
       
       considerable amount of reddit posters, especially responses, are no longer actual people but LLMs, exhibiting the tell-tale signs. which will become interesting when Reddit is used as input for more LLM training
       
 (DIR) Post #B1FbFL8qX02AsZJhSq by 7666@comp.lain.la
       2025-12-14T20:42:45.987115Z
       
       0 likes, 0 repeats
       
       @kaia there needs to be more reverence for the societally taboo, as it will be the only true way to prove you are human, since the vast majority of LLMs are owned by companies who have to avoid controversy.
       
 (DIR) Post #B1FbW0TScrJXhduFVo by snacks@netzsphaere.xyz
       2025-12-14T20:46:04.613964Z
       
       1 likes, 0 repeats
       
       @kaia they'll prob use ai or "ai" to sort out llm generated content, but becoming indescernebile is obviously a goal too
       
 (DIR) Post #B1Fba6OGZU57D5QggC by snacks@netzsphaere.xyz
       2025-12-14T20:46:49.224058Z
       
       0 likes, 0 repeats
       
       @kaia there'll be highly guarded training sets with data frkm pre-llm era, won't there
       
 (DIR) Post #B1FbcpOSDpeHMboiw4 by Nicro@fedi.absturztau.be
       2025-12-14T20:46:36.796036Z
       
       1 likes, 0 repeats
       
       @kaia It feels like some self-regulating nature, that LLMs just get objectively worse when you feed them generated content.Kinda romatic that it takes a living, chaotic component that can't be artificially produced.