Papers · Preprint · v2

Hiddenness as a Structural Law of Emergence

Once the machine's internal state shows in what you can see, no change can alter the future while leaving the present untouched. The hidden information is still there.

In plain words

When does seeing more of a computation's past add something to what you can see now? This paper studies finite sets of histories made by a procedure with an internal state, and groups them two ways: by tests that read only what is visible now, and by tests that may also read witness and continuation data. Exposure means some current test reads a part of the internal state that actually varies.

Two panels with the same two histories, A and B, whose visible records match, same formula, trace and audit, while their internal states differ: 0 1 and 1 1. Hidden: current tests read only the visible records, so current A equals B, while predictive tests read the states, so predictive A is not B: a hidden difference. Exposed: a current test reads the first state bit, 0 versus 1, so current A is not B as well: told apart now. Note: under the paper's premises, no claim about P versus NP.
An illustration with two-bit internal states. While no current test reads the state, these two histories look the same now and still predict differently. Once a current test reads a part of the state that varies, they are told apart now. Under the paper's premises, no loop can then change what a history predicts without changing what it shows, though the predictive grouping can remain finer than the current one.

Under six stated premises, the main lemma has four clauses. With exposure, no loop can change what a history predicts without also changing what it shows now; the predictive grouping can still be finer. Without exposure, and with predictive witness content, some loop does change the prediction unseen. Without witness content the two groupings agree, and which branch holds depends on exposure, not on how strongly the procedure couples to its state. The clauses are short consequences of the premises; the contribution is their formulation, not a derivation of the premises for any class of algorithms.

A second theorem is a minimal example: two histories with different predictions and the same current view give the same yes or no outputs, so the difference can vanish when one passes to outputs. No link to P and NP is built.

What it shows

  • The Candidate Structural Law: four clauses proved from six stated premises.
  • Exposure as the single predicate that decides between the branches.
  • A minimal example where a predictive difference disappears in the outputs.

What it does not claim

Nothing about P and NP. The premises are assumed, not derived for any class of algorithms, and the law holds on finite carriers that meet them.

Cite

Tsiokos, I. (2026). Hiddenness as a Structural Law of Emergence. Zenodo. https://doi.org/10.5281/zenodo.23086991