Elliot Callender's 'Self-Control' reveals working memory limit is a neurological type error

Elliot Callender's LessWrong article 'Self-Control Is A (Neurological) Type Error' challenges the intuitive model of executive function. The author attempted to expand working memory by using brain stimulation to have an external system take the role of 'controller', expecting to gain free metacognition. This failed because working memory items are not discrete units stored by a central will; rather, they emerge from the coordination of many brain regions synchronizing on a common carrier frequency. This coordination is fundamentally limited by harmonic uncertainty, capping the number of working memory entries to a few tens. The feeling of limited control is a compression of this ground-up process, not a bottleneck that can be bypassed by external stimulation.

On timescales below ~500ms, cognitive improvement from such interventions is negligible. However, for tasks spanning seconds or longer (especially those requiring sleep for consolidation), engineering external tools like an index or efficient retrieval systems can effectively expand usable memory. Callender also notes that this refined understanding helps explain psychosis symptoms in digital telepaths—a phenomenon explored in a follow-up post. The core insight: 'control' is a generative effect of local neural computations, not a resource that can be directly augmented. This has significant implications for cognitive enhancement, AI safety, and intelligence amplification.