the long version
This post answers a reader's question in full. The short version lives in the quantities ledger on the home page.
The question: the framework converts intelligence into creation through recursion, but building anything real also consumes energy and matter. Is the framework missing a term?
No, and the reason it is not is the framework's own first row. Physical creation is fed creation. Anything that builds in physical space draws energy and matter through a transport network: blood vessels, root systems, road grids, supply chains. The geometry of that network sets the geometric scaling bound, and that is exactly what d/(d+1) says: how fast a physical system can grow is set by the dimensions of the space that feeds it, and the expression stays below one for any finite number of dimensions. Energy-and-mass-required is not an objection to the framework. It is the transport regime, and it is why nothing physical, organism, city or economy, has ever grown without bound.
The second row's claim is then narrow and precise: recursive information growth is the first kind of creation whose binding constraint is not supply geometry. Software still burns joules, and data centres are physical. But the recursion itself, information compounding on information, is not fed through a d-dimensional delivery network, so the transport ceiling does not bind it. Something else has to.
The third row proposes what: correction. In a system that improves itself, the correction has to grow at least as fast as the capability, and the proposed ceiling on stable recursive growth is that condition evaluated at the limit of how well any corrector living inside the system can scale. Read this way the three rows are one story. Creation bound by energy and mass obeys the transport law. Creation unbound from transport needs a new limiter. The proposed new limiter is governance, not plumbing.
One edge belongs in the open, because it is falsification-shaped. If AI capability growth ever turned out to be rate-limited by energy supply with a transport-like signature, the curve bending to the grid rather than to correction, that would drag software back under the first row and hollow out the ARC Ceiling's premise. Nothing published shows that today, and current scaling behaviour does not look like it. But the framework is falsifiable at its scoping, not only at its numbers, and this is where.
The full statement of the problem is at the open problem; how the pieces compose is at the synthesis; the dated record with kill-conditions is at the evidence page.