AI memory needs to be more than a vector database

I kept thinking about one problem with AI memory: why should two memories be considered the same just because they mean the same thing?

That question became NCM (Native Cognitive Memory). Instead of storing memories as just embeddings, NCM models semantic meaning, emotional context, internal state, and time together.

The interesting part: the same query can retrieve different memories depending on the system's state.

I've also been testing contradiction-aware retrieval and scaling the system to 100k memories.

This is still a work in progress. But I think AI memory needs to be more than a vector database. NCM on GitHub.