I know that I can create the index later.
The question is if the results of similarity search can potentially be different (different embeddings will he returned) if I have have thousands of embeddings and “I first create index then insert data” or “first insert data then create index”.