(You still use Claude like it's a stranger) Most people "personalize" AI by typing a few preferences into a chat box. That barely scratches your voice. Real process: open Cowork mode, build a dedicated voice folder, run Opus 4.8 with extended thinking on, then dictate (don't type) answers to all 100 interview questions. Push past vague replies, document what you refuse to write about too, your limits define your voice as much as your opinions. One final prompt compresses those 20,000 words into a single 4K token file. Save it under your name. That file now carries the way you think and talk into any Claude session, anywhere. Save it for later Share it to help others

I almost jumped on the Claude Fable hype. For about 24 hours it felt like everyone was talking about it. People were sharing demos, posting threads, showing what they built with it on X, and making it sound like the next thing every founder should be using. Then almost as fast as the hype arrived, the headlines changed. Watching it unfold made me realize how risky it can be to build core parts of a product around something you've only known for a day or two. Maybe I move too slowly, but moments like this make me glad I didn't rush to integrate it into anything important. Now I'm curious about the founders who actually did. Did anyone here use Claude Fable in a real project, even as a test? What did you build with it, and was it actually better than the alternatives or was the hype bigger than the results?