Ask HN: Do you use LLM memory features?
I've been experimenting with AI assistants and their "memory" features, but I'm finding them opaque and unreliable.
Instead, I've switched to keep important context in .md files that I explicitly reference when I need or when I have to recall them. This gives me:
Full visibility into what's in context No mystery recalls from weeks ago Easier debugging when behavior drifts Predictable token usage The downside is manual maintenance, but it feels more reliable than hoping the system remembered correctly.
Do you use built-in memory features? Or do you manage context explicitly through files/RAG/other methods? What's working for you? I experienced this as well. The built-in memory features are a black-box. Can't see what's stored, can't control when it's recalled and takes random irrelevant examples from past conversations into unrelated conversations. Also, I am into using multiple models, so it was frustrating they can't talk to each other. The .md file approach is cool but like you mentioned, it is an overhead and doesn't scale well. I have been building something in this space (Maximem Vity [https://maximem.ai]) that tries to solve this more systematically: a cross-LLM, cross-app memory layer that sits in a secure cloud vault. The idea is you control what gets stored and you can summon specific pieces of it into any AI session, granularly. So instead of hoping GPT remembers that you prefer TypeScript and work at a fintech startup, you explicitly pull that context in wherever you need it. I've been doing exactly the same thing as you and i've found it to be the most reliable. I am more accustomed to letting AI summarize key points every once in a while, and then supplementing myself. I prefer flat files too. I'd rather have mostly manual control of context.