Engram
An engram is a physical or chemical trace in the brain theorized to underlie the storage and retrieval of memories. The term was coined by German biologist Richard Semon in 1904 to describe the neural substrate of memory—the biological changes that occur when information is encoded in the brain. Semon proposed that sensory experiences leave persistent marks on neural tissue, which could later be reactivated to produce memory recall. Though Semon’s specific mechanistic proposals were not sustained by later research, the conceptual framework of engrams became central to neuroscience and cognitive science inquiries into how memories are physically instantiated.
Neuroscientific perspective
Modern neuroscience understands engrams in terms of synaptic changes, neural circuit modifications, and molecular processes including protein synthesis and gene expression. Research suggests that memories involve structural alterations in connections between neurons, particularly through mechanisms like long-term potentiation and long-term depression. While the exact physical basis of memory remains incompletely understood, contemporary neuroscience accepts that memories do have a material foundation in brain structure and function, validating Semon’s original intuition even as the details have been substantially revised.
Contemporary usage
In recent years, the term “engram” has been adopted by artificial intelligence researchers, particularly at DeepSeek, to describe context-aware knowledge retrieval systems in large language models. This application extends the original concept metaphorically—referring to how neural networks store and retrieve learned patterns and information. The dual usage reflects both the historical significance of engrams in memory science and their continued relevance as a conceptual framework for understanding how information persistence functions across biological and artificial systems.
Source Notes
- 2026-04-14: “But OpenClaw is expensive…”
- 2026-04-07: DeepSeek Engram Solving LLM Inefficiency Through Context Aware · ▶ source