Comment by pfraze

Genuinely curious, how would you build it differently? Would you stick with atproto's indexing model and try to simplify it, or would you use another approach?