OpenAI's roon argues AI writing fatigue is caused by repetitive user exposure rather than declining model quality
A post on X argues that people are tiring of AI prose because they now see its favorite habits constantly, not because the writing itself suddenly got worse.
In a post on X, @tszzl argued that AI prose now feels more irritating to many readers not because language models used to write better, but because heavy use has made their mannerisms hard to miss. The post says models need a "superhumanly diverse idiolect" to avoid becoming grating, and it points to an earlier post from @nabeelqu that critics in this conversation treated as an example of the kind of stock phrasing now associated with AI writing.
hypothesis: the writing styles of language models are basically fine, they weren’t better in some halcyon before times. we just use them so much that we get annoyed by their mannerisms. they need to have a superhumanly diverse idiolect to not become grating
Commentary on X
Highest rankedEvery time I read something AI written I want to commit a violent crime. That is not to say it is an LLM's fault or makers of models, but it's a far cry IMO from being basically fine. Things were absolutely better when you as a human were forced to think and write with your own mind to another human* *I mean this mostly for direct human to human communication https://x.com/joshpuckett/status/2074991032526364729
Combined views
67.8K
11 posts, first seen 2h ago