Documentation

Memory Decay & Relevance Scoring

How relevance can change over time and why older memories do not simply disappear.

Last updated: August 2025

Concept Status

Why Relevance Changes Over Time

Not every saved memory remains equally important forever. A memory about a specific bug you fixed in a project you no longer work on is less relevant than a memory about a pattern you use every week. Relevance scoring aims to reflect this reality.

Factors That Influence Relevance

FactorEffect
RecencyMemories you saved recently start with higher relevance
Frequency of accessMemories you view often stay more relevant
Search interactionClicking a result in search increases its score
Time without accessMemories not accessed over long periods gradually reduce in prominence

Decay vs. Deletion

Memory decay is not the same as deletion. A memory with low relevance does not disappear — it simply appears lower in default sort order and search ranking. It remains fully accessible when you search for it directly or browse by tag.

Why Old Memories Should Not Disappear

The value of a knowledge base comes partly from its history. A memory about a programming concept you learned two years ago might still be exactly what you need when you return to that technology. Permanent deletion would destroy that long-term value.

The goal of relevance scoring is to surface the most useful knowledge first, not to discard what you have built up over time.