Claude built me playlists from new songs based on me Spotify Export

I have an extensive 10 year Spotify history that is great raw data for AI to analyze and make a playlist out of. This is a really great use for Fable that I wouldn't pay AI credits for, but would certainly use for while it's still around on subscription. What's really nice about the Spotify export is that it even has skip information, so Claude can know which ones I didn't really listen to all that much. First I had it clean the exported files. LLMs can't really read JSON efficiently, so I had it create an "LLM friendly" version of the exports and put them all in a folder. I just told it to optimize for analysis. It knew what to do based on those two things. Fable wasn't necessary for this task, Opus would do just fine. It created CSVs and MD files that are easier to scan, and in my case even a table of most listened to artists, summaries, etc. Next, Fable (in Claude Code) ran in the folder with the cleaned up export. I first told it to just read it and give cute analysis, which is just a bonus. Stuff like what can you infer about me based on the data, which genres do I like, characteristics of the music I enjoy, my listening eras, etc. I also had it build a website to visualize the analysis which was fun to explore. Finally, I had Fable extract what I enjoy from the music I tend to binge, told it to connect to free Music APIs available on the web to get even more music info (BPM, genre, key signatures, etc) and devise a plan to create a 30-song playlist filled with music I have never heard before but guaranteed to enjoy because of his analysis. Fable automatically added "double check API to make sure the track is not a hallucination", and it was a nice add because ChatGPT and Opus hallucinated tracks when I gave it this task in the past. Fable found bangers. I have new favorites now. I can't recommend this enough.