Paper 1: novelty as an exact resource theory
Paper 1 is where the framework’s central word stops being a slogan. Novelty, up to this point, had been used the way start-ups use the word disruption: constantly, and without a definition. The paper’s move is to treat novelty as a resource, in the exact sense that thermodynamics treats heat and quantum information treats entanglement, and once you do that, a surprising amount of structure appears for free.
What resource theory gives you
A resource theory starts with a distinction: which operations are free, and which state transformations cost resources. In thermodynamics the free operations are anything that neither creates nor concentrates energy; the resource is work. In entanglement theory the free operations are local operations and classical communication; the resource is entanglement itself. The payoffs are always the same shape: monotones, which are quantities that free operations can never increase, and these monotones become the theory’s currency, convertible along well-defined exchange rates.
Paper 1 defines the free operations for novelty as the compressive maps: any operation that cannot increase the description length of the data it acts on. The resource is then the discoverable structure in a data stream, measured by the drop in description length that absorbing that structure buys. The monotone is novelty flux density, the rate at which this drop occurs per unit resource, and the first theorem is that no free operation manufactures it. You cannot get discoveries from a scheme that only compresses. Discovery has to come from somewhere: new data, new instruments, or new physics, and the paper’s accounting makes that somewhere explicit.
The two results that matter downstream
Two theorems carry the framework, and both are proved rather than asserted. The plateau theorem: if the truth about a domain lies inside your model class, the novelty rate decays to zero and the total budget of future discovery is finite. The proof is the two-part code arithmetic from the walkthrough series: each discovery buys a smaller residual drop than the last, and the sum of drops converges. The second theorem covers the case where the truth is outside your class: the rate still dies, but it floors at a level set by the model class itself, not by your effort.
That floor changes how you read every saturation curve in the empirical literature, and I think it is the paper’s real contribution. A field that looks like it has stopped discovering may be in one of two very different states: genuinely exhausted, or working below a floor its own methods impose. The theorem tells you these are distinguishable, by checking whether the residual cost at the floor is compressible by a larger class. It also tells you the two escapes: buy a bigger class, or buy data with different physics. Both are expensive. Neither is optional.
What the paper does not claim
It does not claim that human science is currently at a plateau; the empirical case for that lives in the Discovery Plateau work and is contested. It does not claim the resource theory is unique; other monotones could be defined, and the paper flags that the choice of free operations is the substantive assumption. What it claims is narrow and useful: there is a consistent mathematics in which novelty is scarce by theorem rather than by lament, and in that mathematics, the question of where future discovery comes from has a forced answer. The rest of the series is built on that answer.
The next paper puts the measure to work on the architecture question, and settles which level the quantum mechanics is happening in: Paper 2, the two-level architecture.