Last time I posted SeekStorm here, it was lexical-only. Since then, SeekStorm has gained native vector and hybrid search, a sharded, multi-core index architecture (think of a miniature data center inside one machine), faster phrase search with an N-gram index, multiple document compression schemes, result sorting, typo-tolerant search and autocomplete, PDF ingestion, SDKs for multiple languages, and now an MCP server.

A lot has changed in two years!

What's your take on sub-millisecond latency if nobody will notice the difference anyway? I don't really see the gain for 99% of use cases