Δ-Mem: Efficient Online Memory for Large Language Models | Hacker News

Hacker News discussed Δ-Mem, a proposed online memory system for LLMs aimed at improving how models retain and reuse information.

Some commenters argued it mainly expands context capacity, not the harder problems of retrieval, caching, and finding the right information at the right time.

Related ideas included fixed-size memory, agent-written knowledge bases, scripts, and documentation-first workflows such as Claude.md or git history.

The debate underscores a core AI systems challenge: storing more data matters less than reliably retrieving relevant context, which affects cost and automation.