... agent forgets everything. The standard fix is cramming thousands of tokens of conversation history into every prompt. It's slow, expensive, and most of it isn't even relevant. So I looked at what's out there. Mem0, Zep, Supermemory. They all do basically the same thing: take text, embed it as blobs, throw it in a vector database. It works, sort of. But there's no structure. You can't look up a specific piece of knowledge ...

... smart people are already working on AI memory. Projects like Mem0 describe themselves as a universal memory layer for AI agents , focused on storing and retrieving user interactions so assistants can remember preferences and context over time. Supermemory pitches a scalable plug-and-play memory infrastructure and a Memory OS for AI apps, with APIs, SDKs and a Memory Router to attach persistent memory to any stack. I love that this space exists. But I also kept noticing a missing ...

... First thought is that memory is certainly useful. The fact that ChatGPT has long-term memory now is a huge lock-in factor for me. The exact form factor of the product is an open question. I like how @Supermemory is approaching it - and it seems they have some traction with their recent launches. This might be a good area to prototype around - build a bad version of all the different things you mention and try to use them ...