Synaptic Pruning
Unused connections weaken.
What is it
A toddler's brain has roughly twice as many synapses as an adult's. This is not a deficiency of the adult brain — it's an achievement. Between early childhood and adulthood, the brain systematically eliminates synapses that aren't being used. This process is called synaptic pruning, and it's essential for efficient brain function.
The logic is counterintuitive: more connections means less intelligence, not more. A brain with too many connections is noisy, slow, and indiscriminate. It can't tell signal from noise because everything is connected to everything. Pruning creates clarity by removing the irrelevant, leaving only the connections that have proven useful through repeated activation.
Pruning is not random deletion. It's activity-dependent. Connections that are used frequently are preserved and strengthened (that's LTP). Connections that are rarely or never used are tagged for removal. The brain is sculptured by experience: what you do shapes what stays, and what you don't do shapes what goes.
What it does in the brain
The most dramatic pruning occurs during adolescence, particularly in the prefrontal cortex. This is part of why teenagers are simultaneously brilliant and erratic — their brains are being restructured, and the pruning process temporarily disrupts executive function before ultimately improving it.
Pruning continues throughout life, though at a slower pace. When you stop practising a skill, the associated synapses gradually weaken. You don't forget instantly — the pruning is gradual. But after years of disuse, a once-strong pathway can become barely functional. This is why relearning a childhood language is easier than learning a new one: the pruned connections aren't fully deleted, just weakened. Some scaffold remains.
Disorders of pruning are implicated in several conditions. Too little pruning may contribute to autism (too many connections, overwhelming sensory processing). Too much pruning may contribute to schizophrenia (too few connections, fragmented thought). The balance matters enormously.
What it does in ThetaOS
The design philosophy for pruning in ThetaOS is "dimming, not deleting." The system never permanently removes data — that would violate the principle that a personal knowledge system should be a complete record. Instead, connections that haven't been activated or reinforced gradually lose their prominence in query results.
A person you haven't interacted with in five years still exists in the database with all their historical connections intact. But their visibility score dims over time. They won't appear in summaries, won't surface in related-entity queries, won't feature in Tom's dossiers unless specifically asked for. They're pruned from active memory but preserved in archival memory.
The dimming mechanism works on the same potentiation scores used by LTP. If no new evidence arrives for a connection, its score decays slightly with each nightly cronjob cycle. After months without reinforcement, the connection drops below the action potential threshold — it's there, but it no longer fires. Ask specifically and you'll find it. But it won't volunteer itself. This mirrors how your brain handles a childhood friend's phone number: it's not erased, but you can't recall it without a specific prompt, because years of non-use have pruned it from active access.
This is designed but not yet fully implemented. The decay function exists in the scoring model. The nightly cronjob applies it. But the tuning — how fast should connections dim? Should some categories be exempt? Should explicit "this person matters" flags override decay? — requires experience and iteration. The mechanism is in place; the calibration is ongoing.
Designed — dimming not deleting