Everything below the waist is aggregation — mechanical, per-chain, and it needs nobody's agreement. Everything above it is universal. The waist is the only place in the system where independent parties must converge on a single value, and that is what a coordination layer is. It is also the one layer not yet decentralised.
A flat stack of nine equal bands is easy to draw and it hides the three facts that matter. Each correction below changes what the picture argues.
“Coordination layer” is usually a word people reach for when the thing is hard to describe. Here it has a test, and the diagram passes or fails it visibly.
Below the waist, nobody needs to agree with anybody. A twig root is a function of block headers — two honest machines compute the same one, and if they differ, one is simply wrong. That is aggregation, and it is not a consensus problem.
At the waist, sixty-two per-chain roots become one superroot, and there is exactly one of them. Every party downstream — a contract on Base, an agent on the CLOB, a hosted Cubicle strategy — inherits that single value. They are coordinated by construction, because there is nothing else to be coordinated on.
Today the waist is computed by a single indexer on a ten-second timer. Everything beneath it is independently checkable; everything above it inherits whatever that one process decided.
That is not a weakness discovered in review — it is the product. The lattice sits at L4 precisely because L4 is the last layer holding many independent values and the first producing one. It is the only place a quorum can do any work.
Which is also why the status lens is worth showing to a technical investor rather than hiding: eight layers read REAL, one reads PROPOSED, and the one that reads PROPOSED is the one being raised for. A stack diagram where everything is green is a diagram nobody believes.
Three audiences want three different things from the same picture, and shipping three diagrams means maintaining three diagrams and confusing anyone who sees two. The lens switcher is the cheaper answer: the geometry never moves, only the emphasis and the annotations do, so a reader who has seen one lens can read the next without relearning the shape.