hafs is a local, explicit memory layout for tools and agents so context doesn't evaporate between sessions. the core idea is simple: treat context like a filesystem, not a prompt. you keep durable memory, working memory, and decision trails in predictable paths so you can audit what happened and why.

• memory/ long-term beliefs and constraints
• knowledge/ reference material and canonical docs
• history/ session logs and episodic traces
• scratchpad/ active reasoning and live state
• hivemind/ shared learnings and consensus

hafs is built to be boring in the best way: predictable paths, readable state, and a clean separation between long-term knowledge and short-term intent. it fits naturally next to the lab projects because it is infrastructure for how i build and document everything else.