---
title: "Show HN: Experiments with Long-Term Memory Architecture for LLMs"
source: "Hacker News 热门（Algolia · 中文化）"
source_tier: "t2"
url: "https://github.com/faisalhussain-devs/MindCache/tree/collapsed_tree"
permalink: "https://feed.astralor.com/items/itm_86046420a26f21da1155faae"
published: "2026-08-12T20:03:11.000Z"
score: 41
channel: "ai"
tags: ["LLM","MindCache","Collapsed Tree","Long-Term Memory"]
---

# Show HN: Experiments with Long-Term Memory Architecture for LLMs

## AI commentary

> Though rough, such small experiments often inspire alternative approaches beyond mainstream solutions. For developers working on RAG or memory systems, MindCache's collapsed-tree structure is worth a quick look, even just as a comparison point.

## At a glance

A Hacker News submission introduces MindCache, an open-source experimental project that explores long-term memory for large language models using a collapsed tree structure. The approach aims to create a scalable memory organization method, enabling models to retain and retrieve key information across sessions. Still in early exploration with minimal discussion, the project offers a fresh direction for LLM persistent memory design.

## Key points

- MindCache organizes LLM long-term memory via a collapsed tree, emphasizing hierarchical information storage.
- It aims to solve context loss across multi-session conversations.
- The repo is at an early stage with architecture demos and basic code, not yet a full framework.

## Full text

Show HN: Tried some experiments with architecture for Long term memory for LLM · HN points: 1 · comments: 0 · https://news.ycombinator.com/item?id=49277842

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_86046420a26f21da1155faae