The World as Instrument: Simulated Realities that Carry the Frontier of Physics

Procedurally generated worlds are usually built by painting outcomes: a designer encodes the result directly, as data or a tuned constant. This position paper argues for the opposite discipline. A world can instead be derived from a small generative ontology of continuous fields and thresholds, where a threshold is the point at which the ratio of two competing physical terms crosses one; its discrete structure then follows as a consequence rather than being listed by hand. The paper's contribution concerns what such a world should do at the edges of physical knowledge. Rather than dropping a poorly-measured quantity (setting it to zero) or fitting it until the output looks right, every physical unknown is carried as a named quantity, held at its measured bound, with the uncertainty of that bound recorded alongside it. A world built under this rule becomes an instrument: because the frontier of physics is represented explicitly, a discrepancy a participant encounters while acting in the world can be read back as information about the underlying physics — but only when the discrepancy exceeds the model's own expected error, a condition the paper makes central. The stance is distinguished from four neighbouring bodies of work: axiom-evolving machine discovery, games as testbeds for open-endedness, the philosophy of simulation as experiment, and citizen-science games. The last is the sharpest contrast: there the science is a labelling task bolted onto an inert world, whereas here the discrepancy lives in the world's own derived physics. The confinement of the quark-gluon plasma into hadrons serves as a worked example that, under the discipline's own rule, correctly cannot yet be built — three required inputs have no measurement standing between assumption and use — which the paper treats as the method working rather than failing. Finally, the ontology is argued to be falsifiable in the strong sense that it must rule some worlds out: if changing a threshold's terms can be made to produce any outcome, the ontology is decorative rather than generative, and the paper states the test that would decide this.