How to keep a second brain cheap enough to actually run: cost bound by query volume, not corpus size.
Loading…