What if MemPalace got 100% in the metric it was going for all the while still not using any LLMs or APIs or GPUs...
Well, wonder no more, I tidied up an experiment from this past weekend and am sharing the results as well as the code (links in replies)