Cool project. I like the neuroscience analogy with decay and consolidation.
I've been working on a related problem from the other direction: Claude Code and Codex already persist full session transcripts, but there's no good way to search across them. So I built ccrider (https://github.com/neilberkman/ccrider). It indexes existing sessions into SQLite FTS5 and exposes an MCP server so agents can query their own conversation history without a separate memory layer. Basically treating it as a retrieval problem rather than a storage problem.
hmm the repo doesnt mention this at all but this name and problem domain brings up HippoRAG https://arxiv.org/abs/2405.14831 <- any relation? seems odd to miss out this exactly similarly named paper with related techniques.
Aren't tools like claude already store context by project in file system? Also any reason use "capture" instead of "export" (an obvious opposite of import)?
no open code plugin? This seems like something that should just run in the background. It's well documented that it should just be a skill agents can use when they get into various fruitless states.
The "biological" memory strength shouldn't just be a time thing, and even then, the time of the AI agent should only be conformed to the AI's lifetime and not the actual clock. Look up https://stackoverflow.com/questions/3523442/difference-betwe... monotonic clock. If you want a decay, it shouldn't be related to an actual clock, but it's work time.
But memory is more about triggers than it is about anything else. So you should absolutely have memory triggers based on location. Something like a path hash. So whever an agent is working and remembering things it should be tightly compacted to that location; only where a "compaction" happens should these memories become more and more generalized to locations.
The types of memory that often are more prominent are like this, whether it's sports or GUIs, physical location triggers much more intrinsics than conscious memory. Focus on how to trigger recall based on project paths, filenames in the path, file path names, etc.
Memory links to location but that's largely because humans are localised. Isn't that also a weakness. We should be trying to exploit the benefits of non-locality [of ML models and training data] too.
I feel like much of my life is virtual, non-localised. Writing missives to the four corners of the wind here and elsewhere; gaming online; research/chats with LLMs or on the web, email with people.
My physical location is often not important - a continuing context from non-physical aspects of my existence matters more.
That said, one of the things that's hard for me about digital life is the lack of waymarks - I used to be quite "geographical" in my thinking. Like "oh the part I found interesting was on the left page after the RGB diagram", I'd find that and also find my train of thought and extend it. Now, information can be in any myriad of freeform places across at least 3 devices and in emails, notebooks, bookmarks, chat histories, and of course my brain. When some ready syncretism of those things happens it feels like we'll make better advances. Personal agents can be a part of that.
How does it select what to forget? Let's say I land a PR that introduces a sharp change, migrating from one thing to another. An exponential decay won't catch this. Biological learning makes sense when things we observe similar things repeatedly in order to learn patterns. I am skeptical that it applies to learning the commits of one code base.
Cool project. I like the neuroscience analogy with decay and consolidation.
I've been working on a related problem from the other direction: Claude Code and Codex already persist full session transcripts, but there's no good way to search across them. So I built ccrider (https://github.com/neilberkman/ccrider). It indexes existing sessions into SQLite FTS5 and exposes an MCP server so agents can query their own conversation history without a separate memory layer. Basically treating it as a retrieval problem rather than a storage problem.
Cool to see others on this thread.
Here's a post I wrote about how we can start to potentially mimic mechanisms
https://n0tls.com/2026-03-14-musings.html
Would love to compare notes, I'm also looking at linguistic phenomena through an LLM lens
https://n0tls.com/2026-03-19-more-musings.html
Hoping to wrap up some of the kaggle eval work and move back to researching more neuropsych.
hmm the repo doesnt mention this at all but this name and problem domain brings up HippoRAG https://arxiv.org/abs/2405.14831 <- any relation? seems odd to miss out this exactly similarly named paper with related techniques.
Aren't tools like claude already store context by project in file system? Also any reason use "capture" instead of "export" (an obvious opposite of import)?
> Aren't tools like claude already store context by project in file system?
They do, the missing piece is a tool to access them. See comment about my tool that addresses this: https://news.ycombinator.com/item?id=47668270
cool project mate, gj
yegge has a cool solution for this in gastown: the current agent is able to hold a seance with the previous one
no open code plugin? This seems like something that should just run in the background. It's well documented that it should just be a skill agents can use when they get into various fruitless states.
The "biological" memory strength shouldn't just be a time thing, and even then, the time of the AI agent should only be conformed to the AI's lifetime and not the actual clock. Look up https://stackoverflow.com/questions/3523442/difference-betwe... monotonic clock. If you want a decay, it shouldn't be related to an actual clock, but it's work time.
But memory is more about triggers than it is about anything else. So you should absolutely have memory triggers based on location. Something like a path hash. So whever an agent is working and remembering things it should be tightly compacted to that location; only where a "compaction" happens should these memories become more and more generalized to locations.
The types of memory that often are more prominent are like this, whether it's sports or GUIs, physical location triggers much more intrinsics than conscious memory. Focus on how to trigger recall based on project paths, filenames in the path, file path names, etc.
Memory links to location but that's largely because humans are localised. Isn't that also a weakness. We should be trying to exploit the benefits of non-locality [of ML models and training data] too.
I feel like much of my life is virtual, non-localised. Writing missives to the four corners of the wind here and elsewhere; gaming online; research/chats with LLMs or on the web, email with people.
My physical location is often not important - a continuing context from non-physical aspects of my existence matters more.
That said, one of the things that's hard for me about digital life is the lack of waymarks - I used to be quite "geographical" in my thinking. Like "oh the part I found interesting was on the left page after the RGB diagram", I'd find that and also find my train of thought and extend it. Now, information can be in any myriad of freeform places across at least 3 devices and in emails, notebooks, bookmarks, chat histories, and of course my brain. When some ready syncretism of those things happens it feels like we'll make better advances. Personal agents can be a part of that.
Sparse distributed memory is what you’re looking for.
https://en.wikipedia.org/wiki/Sparse_distributed_memory
coming right up, adding it as we speak
yep came here to say this. great to hear it's in process.
How does it select what to forget? Let's say I land a PR that introduces a sharp change, migrating from one thing to another. An exponential decay won't catch this. Biological learning makes sense when things we observe similar things repeatedly in order to learn patterns. I am skeptical that it applies to learning the commits of one code base.