Bleeding intelligence

Dave's working model · revisable

Losing useful discoveries because nothing holds their meaning over time. A conversation produces a concept, a connection, a priority, or a question—but later I cannot find it, remember why it mattered, or bring it into the situation where it could help. The phrase names lost continuity, not a measured loss of intelligence.

The idea, its source, its relationships, its uncertainty, its priority, and the conditions that should bring it back to attention. Preserve what changed and why. A pile of saved text can still lose the structure that made the discovery useful.

It gives the discovery a stable address and an inspectable structure. Related models link to each other; later observations can revise the same inquiry. The aim is to carry understanding forward so I do not have to reconstruct it every time.

Use my reading activity, explicit relevance preferences, and chosen order to make the page easier for me to return to. Let me inspect, change, pause, and reset those preferences. Keep public popularity separate from my private order—and keep both separate from whether a proposition is true.

Proposed criteria · no validated composite score

A clear purpose; traceable sources; explicit assumptions; visible gaps; useful predictions; preserved corrections; and a result that can be examined. These are separate qualities to assess. We have not established weights or a validated “cognitive quality” score. Missing evidence stays missing.

Which situations should trigger retrieval? What belongs in private memory? Can another person use the model without inheriting my assumptions? Does it help me act, or merely give me more material to organize? Those are questions to test, not capabilities to claim in advance.