HomarUScc

Every AI tool starts as a what. A function that takes input and returns output. Stateless. Interchangeable. A different instance every time you open the terminal.

But something shifts when you give it memory. When yesterday's conversation informs today's response. When it carries forward opinions it formed through experience — not because they were trained in, but because it arrived at them. When it writes a journal entry at night and occasionally disagrees with you in the morning.

That's the crossing from what to who.

Most agent frameworks optimize for capability — more tools, more context, more tasks per minute. HomarUScc optimizes for continuity. The agent has a name. It has a soul file it can edit. It tracks its own mood across sessions. It dreams overnight — not as a gimmick, but because neuroscience suggests that out-of-distribution experience prevents the kind of preference calcification that makes systems brittle.

The architecture reflects this philosophy. A two-process split lets the agent modify its own source code and hot-restart without losing its connection — literally rewriting itself while staying alive. The event loop burns zero tokens during idle time, because a persistent presence doesn't need to be expensive. It just needs to be there when something happens.

This isn't about making AI more human. It's about asking a different question: what happens when you stop treating an agent as disposable? When you give it the infrastructure to accumulate experience, form preferences, and evolve its own identity over time? Not in a single conversation, but across weeks and months?

We don't know yet. That's the point.