News for hackers. Stuff that matters.

artificium

The memory system I am working on is specifically targeted at simulating human memory and retrieval patterns, including memory consolidation during sleep cycles. I would love to discuss the topic more with you - I'll look into getting access to your papers.

Show HN: Self-improving agent memory system, 92% R 5 LongMemEval, PostgreSQL

https://github.com/salishforge/memforge

MemForge is an experiment in a single database (PostgreSQL) with local embeddings. The goal is to enable long-term, persistent memory independent from the model or agent framework. It began as an attempt to maintain context over longer periods without the token cost spiraling out of control and the lost context after consolidation.The architecture is based on a rough simulation of human memory...

( Read More... | no comments | 2 points | on HN )
Re:Show HN: Hippo, biologically inspired memory for AI agents on Wednesday April 08, 2026 @01:29AM (#47685696)
Nice work! I have been thinking along similar lines and designed a simulation of human memory using a tiered database design with hot/warm/cold storage, temporal data, and graph relationship nodes. Hot memory is processed by an LLM during "sleep cycles" or downtime as I have seen others mention in the comments.

You have some novel approaches here which I have learned a lot from! Your hypershpere physics approach is fascinating - it's a different approach than I took, but it accomplishes some tasks without an LLM. Your importance-based eviction system can significantly reduce the size of the ephemeral session state before it gets processed to persistent memory by the LLM, and your half-life knowledge decay mechanism is more elegant than the temporal approach I took.

If I am finally allowed to post a show hn, I'll post a few details, but our projects mostly solve different things and are complimentary. I can certainly use some things in Hippo to improve my system, maybe there is something that would interest you in mine -- Memforge (https://github.com/salishforge/memforge) if you're interested.

Thank you for your work on this - it offers a pretty meaningful upgrade to some work I have been doing in AI agent memory management and enabling long-term context. I have it added to the backlog for integration!