Principles / The Compounding Principle
§ Principle 06 of 23
The Compounding Principle
Knowledge that persists and compounds across sessions produces nonlinear returns. Knowledge that resets or is siloed produces linear costs.
The Compounding Principle explains why the gap between stateful and stateless systems widens over time. In a stateful system, knowledge accumulates: each session is more productive than the last because the system knows more. In a stateless system, knowledge resets: every session is roughly the same cost because you start over each time. After 100 sessions, the stateful system has compound knowledge; the stateless system has 100 first sessions. The returns diverge exponentially.
Why it matters
The Compounding Principle means that short-term evaluations underestimate the value of persistent memory. A stateful system might look similar to a stateless one in the first week, but after three months, the difference is dramatic. Teams that evaluate AI systems based on short pilots miss the compounding effect.
In practice
Measure session efficiency over time. In a stateful system, you should see tokens-per-task decreasing as the system accumulates knowledge. If it's flat, knowledge isn't compounding — something is wrong with the memory layer. If it's increasing, you have the opposite of compounding: degradation.