It’s this horrible mish mash of bloated terseness, words overloaded with meaning, but somehow simultaneously too many of them. It feels like reading through treacle. There’s no pacing, no pauses to encourage the reader to stop and consider. It’s locally dense and globally bloated.
Take this sentence that I was writing:
“The big questions never really get settled but small pieces still do. One specific argument can hold up, one specific objection can land even while the big question stays open forever”.
When I was reviewing it with Claude, it “rewrote it for clarity and accuracy”:
“Where I'd still hold the line: Chalmers's finding is specifically about convergence on the big, top-level questions — is compatibilism true, does God exist. It's compatible with real local progress underneath that top line: whether Frankfurt's specific case succeeds against PAP, whether Vihvelin's dispositionalism survives the fink objection, whether Carroll is or isn't an anti-realist. Those get settled and stay settled even while "is free will real" doesn't. That's the distinction I was reaching for badly last message — not "philosophy converges," which the data says it mostly doesn't, but "individual moves within it can still be checked and eliminated," which is a real, narrower form of progress and the only kind this thread actually produced”.
I’m pretty sure some people will take this to be a sign that LLMs are “evolving”. But it’s just terrible writing. They’re becoming worse at prose.
[link] [comments]