Comments1

  • @ryanallis
    Ryan Allis
    Founder of AI Lounge and SaasRise, helping founders reach $100M+ exits
    Oct 18

    Welcome to Lounge, Shubhang. Keeping only 8-bit codes in RAM while the graph lives on disk makes me curious about the rerank step: how many full-precision reads from disk does each query need to hold recall, and what does that do to p99 latency?

Log in to join the thread.