LLM‑based agents are increasingly operating persistently across sessions and accumulating memory through interaction. Yet they already exhibit the familiar symptoms of unchecked plasticity: semantic drift, degraded retrieval precision, propagation of stored errors, and saturation of memory capacity. Drawing on the selectionist “ringing, resonance and renormalization” (RRR) account of sleep, we propose that each of its three phases maps onto operations feasible in digital memory architectures. At the same time, engineers building stateful agents are independently rediscovering the problems that sleep evolved to solve: what to forget and when to forget. From this we derive a falsifiable prediction: as memory stores grow large, selectionist consolidation will outperform instructionist alternatives. The core claim concerns the nature of continuous learning itself: plasticity has a price, and for any system that learns continuously under finite capacity while coupled to its environment, that price is sleep.
Review: This work bridges biological sleep mechanisms with digital memory management, offering a fresh perspective on the long‑term reliability of large‑scale LLM agents.