Sanity check on Milla Jovovich’s MemPalace: Mixed metrics, bypassed judges, and that 96.6% LongMemEval score
Disclosure up front: I work on a different open-source memory system (bitterbot-desktop, ~125 stars vs MemPalace's ~40k so calibrate accordingly). We're trying to solve the same problem from different angles, and I went and read MemPalace's…