How a lab works

A lab starts with a pattern, not a product. Each one takes a slice of the 4+1 stack, a single layer, a combination of layers, or the whole model, builds the smallest honest version of the workload, and reports where the seams actually fall.

The unit of the verdict is the seam map: each function in the loop is ready to run local, conditional with a named tradeoff, or it needs the cloud. The headline is the DAPM table, which places authority layer by layer and ties the lab back to the canon's assessments as they stood on the day the lab ran.

Labs are scoped narrow on purpose. One seam, one verdict. They do not assess a whole vendor and they do not declare winners. The brand is separating marketing from reality: what is ready, what is not, and where the boundary lives.

What gets tested

Authority placement, data movement, retrieval and serving readiness, cost shape, and the question every practitioner is quietly asking: what can I keep local, what still has to go up.

What does not get claimed

No universal superiority. Every lab is scoped to the layers and the workload it tests, and a defined seam. A finding may inform a future Layer2C assessment. It does not purchase one.