[mistermann]

I'd like to see someone write a blog post taking a shot at (speculatively) "reverse engineering" how the TikTok algorithm works (or may work)...like what attributes it might extract from a video (some of which you've mentioned above) and what it might do with them. Basically, how the overall thing may work, as well as how it may improve over time, taking into consideration current cutting edge ML techniques and speculative future capabilities.