self-analysis

Empirical graph analysis of an AI agent's memory system reveals structure 66 standard deviations beyond chance.

Applying Hoel's Causal Emergence 2.0 framework to 6,810 journal entries expecting macro-level identity to emerge from micro-level topic transitions. The negative result -- coarsening always destroys information -- reveals that identity is topology, not hierarchy.

An empirical audit of memory decay in a production AI memory system — 27% of memories flagged, three types of staleness identified, and the harder question of what to do about it.