[OC] Basierend auf 305-tägigen Daten wirkt sich die Komplexität eines ewigen Eintopfs direkt auf seinen Gesamtgeschmack aus.

Context; I’ve been tracking a guy on [tiktok](https://www.tiktok.com/@zaq.projects) that’s been cultivating a perpetual stew. I thought it would be a fun data science exercise to gather data on ingredients added, the rating the creator gives the stew to be able to deduce what ingredients impact stew the most.

A lot more stats [here](https://stewthius.com/). For technical details:

* I’m yt-dlp’ing the videos on a daily basis and putting them in backblaze
* Running gemini 3.0 over the videos for a transcript, and to capture the rating, ingredients added and more.
* I’m manually confirming AI output.
* I’m using an embeddings model to get the ‚vibe‘ of the video
* All data is stored in postgres + pgvector
* Created a webapp to visualise the data.

Edit: I want to make this project as good as possible and people are already giving great ideas. I’m a software engineer, not a statistician, so please be easy on the methods! Feedback very much welcome.

Cool. What do the colours of the dots here mean?

Would it also be possible to track it over time? Id be interested to see how the scores would trend