I appreciate if the answer to this question is proprietary information (honestly, it’s probably marketing material lol), but I was trying to find the end-user time complexity of your index search to estimate how far one could theoretically scale and what it looks like in numbers. I, a mere peasant, use cosine similarity for 1:1 in my own python scripts… that’s O(d) dimensions. I think I read somewhere that you use an approximate KNN, which is (I think?) O(log d)… but I also know Pinecone has a wicked distributed compute platform on top of a newish memory-allocation pipeline w/Pinecone 2.0, so I really have no idea what this looks like. Have you measured it? Can we haz? I’m just curious…
…what’chayo time complexity?! 
