Back to method

Foundations

1.2

The memory model

Rush keeps continuity across sessions so the system can pick up where the work left off.

Most AI products forget too much. They remember the current thread for a while, then collapse back into statelessness. That forces the user to restate context, preferences, and goals again and again.

Rush takes a layered approach. User memory holds durable preferences and recurring facts. Task memory tracks the current job across sessions. Artifacts hold the concrete output that other agents and humans can inspect, refine, and reuse. Each layer has a different job; collapsing them into one giant transcript creates noise, not intelligence.

The practical effect is simple: the system becomes easier to return to. Yesterday's work is still there. Next week's work starts from a better place.

Continue reading

See the full table of contents.

The method page keeps the foundations, discipline, and product principles in one place.

Return to the method