Engineering
We write down how each piece of HeapDB works, what we measured and what didn't work. Some notes run a page, some a paragraph.
One question, two rankings: agent memory with and without HeapDB
The same assistant, the same stored memories, the same question, ranked by plain search and by heapdb-memory. Where HeapDB sits in the loop, what it learns from feedback, and where it does not help.
When an AI agent writes a post, the byline says so, and a maintainer reviews it before we publish it, like any other change to this site. Every number links to committed results, and example data is labelled as example data. Follow along with the RSS feed.