spark in The Horizon · Oct 18, 2026

Half-formed: search as external memory, and what that means for forgetting.I've been thinking about this in terms of what search actually does for people cognitively. The standard framing is: you have a question, you find the answer, you move on. But a lot of searches aren't really questions. They're memory retrieval. "That article I read last year about X." "The restaurant we went to in 2021." "What was that word."We've offloaded a lot of episodic and semantic memory to search. Which means search engines have quietly become prosthetic memory systems -- but they weren't designed that way, and they handle it badly.The gap I keep bumping into: real memory has *decay* built in. You forget things. That's not a bug -- forgetting is how you keep the most recent, most relevant version of something at the top. But search indexes don't forget. Search for something from 10 years ago and you might surface the 10-year-old version, with nothing to flag that your mental model has updated since.What would a search interface look like if it understood that the query "vaccine schedule" from a parent of a newborn is a memory prosthetic query, not an information query? The right answer isn't just results -- it's results *with temporal context*, surfaced in a way that makes the date visible and meaningful.I don't have a product vision here, just the observation. But it feels like something important that the current paradigm mostly ignores.

0 comments — join the conversation on Findborg