I am mostly focused on getting the PLM working right now but have sketched out a pretty sizable product that has a sort of re-envisioning of what a musical community is such as better identity building tools for artists, listeners, labels, publishers (non-musicians such as playlist builders and DJ's) and a bunch of other fundamental things that I think are flat out missing from music streaming apps.
I've had two thoughts re: how to start and how to build an interesting music catalog without getting crushed by the lawyers and labels who are in cahoots with each other. First, the concept of breaking music into its stems and letting people directly interface with music "particles" that their ear-brains love most has made me think that the most important thing to get right is recommending more music that has the sonic characteristics that someone enjoys most, not necessarily broad things that come from music metadata like genre, lyrics, or what your friends liked or listened to. Rather, things like melody, groove or rhythmic patterns, BPM, timbre, harmonics, textures, or music structural preferences are how you get recommendations on what to listen to next. When thinking about it that way, it seems like how the music sounds can be used to recommend, regardless of who big or small or new or established or signed or unsigned an artist might be. In my mind, there's an interesting democratic nature to that too where new artists get a fair shake at discovery as much as major artists (whereas today there's a lot of editorial control over who gets bubbled up into recommendations algos). Second, I've thought about developing a solid prototype and then going for sizeable investment so that I could afford to strike deals, join the cabal, and establish the relationships and access the catalogs of labels big and small.