What the Discovery Plateau Hypothesis Actually Claims
The Discovery Plateau Hypothesis makes one claim, and it is narrower than the name suggests. Within any bounded system, the amount of utility-rich knowledge that can be discovered is finite, and the rate at which it is discovered decays to zero.
That is the whole thesis. Everything else in this programme — the topological substrate, the cosmological constant work, the simulated multiverse — is downstream of taking that sentence seriously.
What it does not claim
It does not claim intelligence is bounded. It does not claim effort stops mattering. It does not claim the space of true statements is finite, because it plainly is not: you can generate new mathematical theorems forever. The claim is about accessible, useful, non-redundant knowledge in a domain with fixed tools and fixed resources.
That distinction does a lot of work, and most objections to the idea attack the wrong version of it.
The evidence is boringly empirical
Three sources, none of which come from physics:
- Citation networks saturate. Growth in a field slows as the field fills in.
- Papers and patents have become measurably less disruptive since the mid-twentieth century, on large datasets, against several different disruption metrics.
- The cost per discovery rises. You need more researchers, more money, more time per result. Moore’s law did not save productivity; the effective research workforce grew roughly an order of magnitude faster than output.
None of this says discovery has stopped. It says the shape of the curve has a ceiling, and we are somewhere on the flat part of it in several classical domains.
Why the ceiling is structural rather than temporary
A domain is a combinatorial space. Chemistry is pairs and triples of elements under conditions. Physics is models and couplings. Biology is sequences and folds. Once the useful region of that space has been visited, the remaining unvisited region is still enormous — and empty of anything you have a use for.
That is the difference between the two numbers this programme keeps separate. The accessible plateau and the total combinatorial possibility space are different quantities, differing by hundreds of orders of magnitude. Confusing them produces either complacency (“discovery is over”) or dismissal (“the space is infinite, so the whole idea is silly”). Both are wrong.
What would falsify it
The claim predicts that sustained novelty in a bounded domain decays. It is dead if a domain shows undamped discovery growth against rising cost, or if the “plateau” turns out to be an artefact of measurement — for instance if disruption metrics are tracking citation conventions rather than substance.
I hold the empirical case loosely, because it rests on bibliometrics rather than on anything I can compute from first principles. The formal part of the programme — the part I can actually prove — starts in a different place: what novelty is as a measurable quantity, and why a bounded agent’s rate of acquiring it must fall to zero.
That is the next post.
This is the first in a series on the Discovery Plateau Hypothesis. The papers are collected at /dph.