Show HN: Mdlens – Reduce token spend and boost retrieval on Markdown-heavy repos
3 by dreeseaw | 0 comments on Hacker News.
Hack Nux
Watch the number of websites being hacked today, one by one on a page, increasing in real time.
New Show Hacker News story: Show HN: AI memory with biological decay (52% recall)
Show HN: AI memory with biological decay (52% recall)
11 by SachitRafa | 5 comments on Hacker News.
Most RAG setups fail because they treat memory like a static filing cabinet. When every transient bug fix or abandoned rule is stored forever, the context window eventually chokes on noise, spiking token costs and degrading the agent's reasoning. This implementation experiments with a biological approach by using the Ebbinghaus forgetting curve to manage context as a living substrate. Memories are assigned a "strength" score where each recall reinforces the data and flattens its decay curve (spaced repetition), while unused data eventually hits a threshold and is pruned. To solve the "logical neighbor" problem where semantic search misses relevant but non-similar nodes, a graph layer is layered over the vector store. Benchmarked against the LoCoMo dataset, this reached 52% Recall@5, nearly double the accuracy of stateless vector stores, while cutting token waste by roughly 84%. Built as a local first MCP server using DuckDB, the hypothesis is that for agents handling long-running projects, "what to forget" is just as critical as "what to remember." I'd be interested to hear if others are exploring non-linear decay or similar biological constraints for context management. GitHub: https://ift.tt/wdhVMnl
11 by SachitRafa | 5 comments on Hacker News.
Most RAG setups fail because they treat memory like a static filing cabinet. When every transient bug fix or abandoned rule is stored forever, the context window eventually chokes on noise, spiking token costs and degrading the agent's reasoning. This implementation experiments with a biological approach by using the Ebbinghaus forgetting curve to manage context as a living substrate. Memories are assigned a "strength" score where each recall reinforces the data and flattens its decay curve (spaced repetition), while unused data eventually hits a threshold and is pruned. To solve the "logical neighbor" problem where semantic search misses relevant but non-similar nodes, a graph layer is layered over the vector store. Benchmarked against the LoCoMo dataset, this reached 52% Recall@5, nearly double the accuracy of stateless vector stores, while cutting token waste by roughly 84%. Built as a local first MCP server using DuckDB, the hypothesis is that for agents handling long-running projects, "what to forget" is just as critical as "what to remember." I'd be interested to hear if others are exploring non-linear decay or similar biological constraints for context management. GitHub: https://ift.tt/wdhVMnl
New ask Hacker News story: Ask HN: Anyone want to collaborate on a local-first AI-based research assistant
Ask HN: Anyone want to collaborate on a local-first AI-based research assistant
2 by venkatram-s | 0 comments on Hacker News.
Hi HN Community, I'm Venkatram, a sophomore who's on a mission to build a local alternative to proprietary third-party AI-based research assistants. The idea is to turn documents into researchable assets that contain as much as information as the original information does, but it's more reusable. Well, quite frankly, this is still under a WORK IN PROGRESS, so i'm still figuring on how it can be properly used, and I got to be honest here, i definitely need some help to build this, so if you wish, you are welcome! TlDR: NotebookLM, but Locally with your OWN AI Model Github: https://ift.tt/MQuYZ8S
2 by venkatram-s | 0 comments on Hacker News.
Hi HN Community, I'm Venkatram, a sophomore who's on a mission to build a local alternative to proprietary third-party AI-based research assistants. The idea is to turn documents into researchable assets that contain as much as information as the original information does, but it's more reusable. Well, quite frankly, this is still under a WORK IN PROGRESS, so i'm still figuring on how it can be properly used, and I got to be honest here, i definitely need some help to build this, so if you wish, you are welcome! TlDR: NotebookLM, but Locally with your OWN AI Model Github: https://ift.tt/MQuYZ8S
New Show Hacker News story: Show HN: Good AI Task – a tool for asking AI what it can and can't do
Show HN: Good AI Task – a tool for asking AI what it can and can't do
3 by jmt710 | 1 comments on Hacker News.
Describe a task, and AI will give you a breakdown of whether it can do your task well, poorly, or somewhere in between. I built it mostly because I kept getting asked "what is AI even good for" and fumbling the answer. The most fun use is testing it on things you already know it can't do and seeing how it explains why it can't be done.
3 by jmt710 | 1 comments on Hacker News.
Describe a task, and AI will give you a breakdown of whether it can do your task well, poorly, or somewhere in between. I built it mostly because I kept getting asked "what is AI even good for" and fumbling the answer. The most fun use is testing it on things you already know it can't do and seeing how it explains why it can't be done.
New ask Hacker News story: Ask HN: How did the industry settle on weekly limits?
Ask HN: How did the industry settle on weekly limits?
2 by saratogacx | 1 comments on Hacker News.
I understand that the cheap compute ride wouldn't last forever but something that feels somewhat unique seems to have come about from all of this AI belt tightening. Weekly Limits This policy seemed to come from almost nowhere but was quickly adopted across many products. Cutting off access for 25% of the time based product you've paid for feels like it is just incompatible with a subscription at the core conceptual level. Cooldown times and over-use back-offs are nothing new but there is a drastic difference in tone from 5 hours to 5-7 days. I'm at a loss to see how this became an acceptable practice with the most common answer being "Buy more subs"
2 by saratogacx | 1 comments on Hacker News.
I understand that the cheap compute ride wouldn't last forever but something that feels somewhat unique seems to have come about from all of this AI belt tightening. Weekly Limits This policy seemed to come from almost nowhere but was quickly adopted across many products. Cutting off access for 25% of the time based product you've paid for feels like it is just incompatible with a subscription at the core conceptual level. Cooldown times and over-use back-offs are nothing new but there is a drastic difference in tone from 5 hours to 5-7 days. I'm at a loss to see how this became an acceptable practice with the most common answer being "Buy more subs"