Directions, prototypes, archived experiments

Recurring patterns detected in experience packets are promoted into durable beliefs - but only under governance.

Every tool call, error, and milestone becomes an experience packet with derived internal state - cognitive load, frustration, energy, dominant emotion, stance, urgency. The agent has a mood.

An advisory pipeline that runs scan → update → consolidate before each session, producing a living-state block that reflects the agent's current internal state.

A prototype mechanism: treat external storage (files, DB) as an extension of the context window, not a separate retrieval system. The agent reads from disk the way a process reads from swap.

A specified architecture for preserving unresolved reasoning across sessions as an evidence-backed, branchable reasoning graph - not as a lossy narrative summary.

A research direction: can a small local model "apprentice" under a larger cloud model - learning preferences, patterns, and style from CSM memory - and eventually replace it for routine work?

The agent does not call tools. The agent creates tools. The model becomes a compiler. WASM generation sprint archived; capability-lineage research active.