What does it mean that graph maintenance is ongoing?

Periodically review and clean your graph — remove dead links and add missing connections.

Example: A product manager builds a personal knowledge graph over six months — 400 nodes covering product strategy, user research, competitor analysis, and engineering trade-offs. She never reviews it. By month nine, a third of the links point to obsolete frameworks her team abandoned, two key concept nodes duplicate each other with slightly different names, and the entire 'market research' cluster is disconnected from 'product decisions' because she never created the edges between insights and the strategies they informed. The graph still looks impressive. It no longer tells the truth.

Try this: Open your knowledge graph or note system. Pick one cluster or tag you haven't touched in 30+ days. Walk through every node and every link. For each node, ask: is this still accurate? For each link, ask: does this connection still hold? Delete or archive anything that has decayed. Add any connection that should exist but doesn't. Track your counts: how many dead links did you remove? How many missing connections did you add? That ratio tells you how fast your graph is drifting from reality.