The Opacity of the Substrate: Can a discrete simulation determine, from within, that it is discrete?
In this essay, I argue that we cannot know whether a discrete simulation can determine that it is discrete. I do this by eliminating the possibility of observation and formal proofs through reasoning about existing mathematical literature.
At the foundation of my argument is the epistemic structure that a simulation must be a child process operating within a parent substrate. The child has access to its own internal states: its models, its observations, its mathematics. But it would have no channel to the type-level properties of its container. Assuming discreteness is one such inaccessible property, the question becomes: are there consequences of discreteness that propagate inward and become visible from within?
Note: I proceed with this structure of a child process operating within a parent substrate, as it aligns with the conventional simulation theory people know.
At first, you might think a discrete grid has a characteristic scale below which the simulation cannot resolve. Probe there, and you find quantization: space appears granular, information is bounded, certain measurements behave anomalously. A cubic lattice would produce anisotropies — physics subtly different along grid axes than along diagonals. This is also where numerical artifacts could accumulate. These are real signatures, and our own universe exhibits striking analogs: the Planck length, quantum information limits, the Bekenstein bound. The simulation hypothesis draws exactly this inference — that our observed quantization is a computational fingerprint.
However, this argument fails on a critical ambiguity. Even granting all three signatures, the observer would not be able to distinguish between two explanations:
- the quantization is a genuine feature of the territory being modeled, which would exist regardless of how it is computed.
And the Shannon-Nyquist theorem expresses this conundrum formally: a discrete system sampling at a sufficiently high frequency can perfectly reconstruct a band-limited continuous signal. Therefore, a fine-enough discrete simulation of a continuous universe is not merely practically indistinguishable from continuous — it is formally indistinguishable. The observer inside literally cannot tell which substrate is real.
So we cannot rely on observation for an answer. No matter how much evidence you gather, evidence alone cannot decide between two complete pictures of what’s generating it — this is the Duhem-Quine problem. Formal proof doesn’t rescue you either, because a proof is only ever as good as the assumptions you feed it, and here those assumptions are exactly what is in dispute.
Here’s what that looks like concretely. There are two ways the underlying fabric of reality — space, time, whatever the substrate’s state actually is — could be laid out:
- As a chain of beads: one moment, then the next, then the next, each with an immediate successor and nothing squeezed in between. This is discreteness.
- As a continuous line: between any two points, however close, there is always another point in between. No smallest step, no beads, no gaps to speak of. This is continuity.
Now notice something. Each of these pictures, taken on its own terms, is a complete mathematical theory with no loose threads. The “chain of beads” picture is old, settled mathematics — it answers every question you could ask about its ordering, with nothing left over. The “continuous line” picture is equally old and equally settled, and it too answers every ordering question completely.
That’s precisely what makes this harder, not easier. The core thesis is that this is not a system straining against a flaw in its own logic — it is a choice between two flawless, mutually exclusive descriptions, and nothing observed and nothing proven from inside either one tells you which of the two you’re actually standing inside. Picture two windows, both perfectly clean, each looking out onto a different room, with no door between them — no amount of staring through either glass tells you which room you’re really in.
I’ve made my argument. Still, I like to look beyond sound arguments or what is proven and ask whether any other barriers remain. Here are some extra points:
- Assume we had a complete and consistent formal language; what barriers would come next for answering that central question? I list other reasons our knowledge must be limited in my The Limits of Truth-Seeking essay.
- Given Wolfram’s Computational Irreducibility, even if discreteness signatures exist in a simulation’s physics and are a genuine feature of the territory being modeled, detecting them might require computation equivalent to running the full simulation. So the answer is not just hard to reach; it may be unreachable in principle without resources that cannot exist inside the system being questioned.
- Turning to philosophy of mind and phenomenology, introspection is no better than measurement. When I think about continuity, I access the concept, not the machinery computing it — a discrete mind and a continuous mind entertaining ‘continuous’ are entertaining the same thing, and nothing in the entertaining shows which one is doing it. Therefore, the simulation’s inner experience gives no leverage on the substrate question either.
- Given things like Turing Completeness, various models of computation, the inference of LLMs, and the Curry-Howard Correspondence, a discrete system can do inference and formal logic about real numbers symbolically, execute calculus, and represent limits and continuity, but cannot prove that its instantiation is discrete. If a discrete computation can contain all of classical analysis without contradiction, then the ontological question ‘is my substrate discrete?’ is orthogonal to the epistemic question ‘do I have access to continuous structures?‘.
This essay may be limited to claims about the structure of self-knowledge in formal systems. But if we want an answer to whether a discrete simulation can determine, from within, that it is discrete, and want that answer to be provably verifiable and communicable, its construction must be via a formal system—the only method we have for creating proof systems. The core point is that a meaningful answer you could share with others is inaccessible, not because of our current observational instruments, but because of the logical relationship between a system and its own computational foundation.