Product
A brain does not keep a separate copy of "red" for every red thing it has seen. It holds the primitive once and links to it from every memory that needs it. When it meets something new, it reaches for the primitives it already has and combines them. Clewd stores data the same way: thoughts built from sub-thoughts, sub-thoughts built from primitives, and every primitive held exactly once.
Every object you store becomes a node that points into a shared graph of primitives. Two objects that look nothing alike on the surface often share most of their primitives underneath, and Clewd keeps those connections explicit. The graph is the store; the objects are views onto it.
Compaction gains grow with the size of the collection, because larger sets repeat more. Clewd is designed for data sets where the cost of storage and comparison dominates, and where the collection keeps growing.