Remembering across sessions
Tomorrow the process has restarted - what does the agent still know, from where, and who wrote it there?
Why this is hard
The model remembers nothing. Everything the agent still knows tomorrow, the harness wrote somewhere today and pulls back into the prompt tomorrow. That sounds like a file but is three decisions: what gets stored, who decides that, and how it comes back - fixed in the system prompt, via search, or only on demand.
The line between history and memory is the real question. A persisted history is not yet memory: it remembers the one session, not the user. A memory extracted from the history is a bet that the extractor keeps the right things. And whatever is injected into the system prompt costs on every call - so it is worth looking at which sources a harness actually feeds in there.
From Pro: the matrix per challenge, all eleven topics with citations, and the architecture portrait of every harness. Freely visible are the six short portraits:
- OpenClaw - The multi-channel personal agent - a half-million-line estate of gateway, channels and tooling whose loop core shares its lineage with Pi.
- Hermes - Nous Research's Python harness - deterministic loop, a skill system, and an error path that escalates in stages instead of just giving up.
- Pi - The coding agent with a deliberately small core - every layer hand-built and kept readable, from the provider package to the session file.
- HybridClaw - The enterprise runtime - sandboxed execution, approvals, and credentials that travel through context as references instead of plain text.
- DeepSeek Harness - The plugin radical - “everything is a plugin” is meant literally here: providers, tools and the loop itself are modules on the Cordis core.
- NanoClaw - The minimalist - small enough to understand in full: the model layer moves behind the Claude Agent SDK, isolation into a container.