Make the playlists actually good

[markkohdev]

I think I math'd wrong somewhere along the way pretty bad.
Reasons why it might be bad:

• My math for vector sim is actually just totally wrong
• Cosine similarity just doesn't work for this jawn
• We should be normalizing the preference vector into a histogram, not in a scaled window (this is probz pretty big). That is, don't place the values in min--vec--max, place it that percentage of the way through a histogram or something
• Drew didn't help us build it :(

[markkohdev]

Idea: Put the min and max values next to the tracks to get an idea of the window range (for debugging only)