Great question! One of the big changes between 2020 and now is the popularization of generative AI. If I were to make this project now, I'd probably feel pressured to either use genAI or justify why I am not using it for the music-making aspect. Though honestly I'd probably keep it the same since being a child of the 80's and 90's I am nostalgic for samplers and sequencers.
Another big change since 2020 is the Music Modernization Act (MMA) which started to actually move (quite old) music recordings into the public domain. This adds significantly more audio to our collective crate of free sounds (I actually added a bunch more audio to Citizen DJ since my residency ended: https://citizen-dj.labs.loc.gov/public-domain-2022/ ). Given the leaps in easier-to-use ML tooling in the past few years, I could probably do a lot more around discovery and visualization of large amounts of audio.