tl;dr: I used to really like Claude (Opus/Fable) for writing but think Astra might be better for what I do. Going to try using Astra as the first choice model. I use AI for content writing mainly (do lots of other stuff, but it's more for fun or side quests). I've been getting quite cheesed off with Opus and Fable output lately, when once it was really good. I think I might've polluted my Claude.md and other instruction files, so it could just be a "me" problem. But they've both been writing stuff that doesn't sound like me or any human I know! (They don't include the obvious AI tells, as those have been easy to document and tell it to avoid via the rules etc). But I think they're just writing funny now for some other reason. Plus each time Claude replies to something, it's like an essay. Not only does that take ages to read but it must be inflating the context window and costing tokens. I do have tl;dr turned on but the main output is still there (and probably worth having so I can get the full picture). I've been using OpenAI as a second opinion for everything I write for a while now. I just tell Claude in VS Code to "get a second opinion," and OpenAI/ChatGPT does a harsh teardown of the article via a connection I set up. But recently, since Claude got a bit weak, I've been "chatting" with Astra in VS Code, in the same folder as Claude, and getting it to talk me through why I don't like the Opus/Fable draft. It's been really good with that. But all these extra layers/steps/models are really taking up a lot of time. I thought AI was supposed to make me more productive? Now I've got loads of versions of the same draft, trying to cobble them together to make one good one that sounds like me, matches the brief/outline, and is factually correct! 😭 So yeah, I might switch to Astra as my main worker and have Opus/Fable as the checker...