Appendix P-1: Informational Normality via M-Randomness

Original Task P-1: Informational Normality Problem: Currently a foundational axiom analogous to Borel normality, lacking formal derivation. Deliverable: A theorem-level derivation leveraging algorithmic information theory (Martin-Löf randomness).

Closure status: RECURRENCE CONJECTURE + COMPUTATIONAL REALISM POSTULATE (June 2026 demotion). Earlier versions of this appendix presented the recurrence component of P-1 as a mathematical theorem (“the physical recurrence of any finite informational sequence … appears infinitely often in \tilde{M}-almost-all sequences”). An external review (opt-theory-review-2026-06, findings F7/F8) identified two gaps that the printed derivation does not close: (i) a discrete/continuous coding-theorem conflation — the §2 construction induced the continuous measure from a halting prefix-free machine, on which no output mass survives on infinite sequences; on the standard monotone-machine repair the discrete identity K(x) = -\log \tilde{M}(x) + O(1) fails, since only the one-sided K\!M(x) \le K(x) + O(1) holds and the gap K - K\!M is unbounded (Gács 1983); and (ii) an atomic-measure obstruction — the a priori measure has atoms on individual computable sequences, on which the recurrence claim is false outright, so the almost-sure recurrence statement cannot hold over the full substrate. This revision demotes the recurrence result from theorem to Conjecture P-1a (Recurrence), stated over the non-atomic observer-compatible component (cf. Preprint §3.1, which already excludes simple repeating sequences from the observer-compatible class), and relabels the §4 derivation as an explicitly heuristic argument with both gaps named in place.

What survives the demotion: The §2 construction of the a priori semimeasure M (now correctly defined via a universal monotone machine). The §3 measure-one status of \tilde{M}-Martin-Löf randomness. The unbundling of Informational Normality into a mathematical component plus the Computational Realism Postulate (§5), which is untouched. The Borel-Cantelli recurrence strategy as the candidate proof route, conditional on repairing the conditional lower bound with monotone-complexity tools.

What changes: “Proposition P-1” / “Corollary+Postulate P-1” is now Conjecture+Postulate P-1; “formally guaranteed infinitely many times” becomes “conjectured to recur infinitely many times, conditional on Conjecture P-1a”; the §2 discrete coding-theorem identity is replaced by the monotone-machine relation K\!M(x) = -\log M([x]) with the one-sided bound K\!M \le K + O(1); the bound mislabelled “near-submultiplicativity” is relabelled as the (unproved) supermultiplicativity lower bound it actually is.

1. The Epistemic Boundary of “Axiomatic” Normality

Within the Ordered Patch Theory (OPT), “Structural Hope” relies structurally on the principle of Informational Normality: the proposition that the algorithmic substrate (\mathcal{I}) is densely populated not merely with noise, but with every finite structural functional pattern. The ethical weight of OPT—the mandate to maintain the stability of the shared patch (Survivors Watch Ethics)—demands that the counterpart observers we interact with have distributed, fundamentally real functional equivalents elsewhere in the substrate.

Historically within the OPT framework, this proposition was treated formally as a single monolithic Axiom—an untestable, foundational assumption layered onto the physics to avoid solipsism.

This appendix resolves the mathematical ambiguity of that stance. We unbundle Informational Normality into two distinct components: an algorithmic recurrence conjecture (Conjecture P-1a below — demoted from theorem grade in the June 2026 review; see the closure-status banner above), bound together by a single metaphysical postulate necessary to bridge mathematical existence into ontological reality.

2. From Semimeasure to Universal Measure (\xi to M)

OPT’s foundation (Preprint §3.1) relies heavily on the Solomonoff algorithmic probability prior. Under this formulation, the generative substrate operates as an infinite algorithmic space executing on a universal prefix-free Turing machine U.

The algorithmic probability or universal semimeasure of a finite string x is:

\xi(x) = \sum_{U(p) = x*} 2^{-|p|}

Where the sum is taken over all minimal programs p whose execution output begins with x. Crucially, \xi is a lower semi-computable semimeasure over finite strings.

To formalize the substrate as a continuous generative space, we transition to the continuous setting over the Cantor space 2^{\mathbb{N}}. The halting prefix-free machine U cannot induce this directly: every halting program produces a finite output, so no output mass survives on infinite sequences and the naive cylinder construction is ill-defined as a distribution on 2^{\mathbb{N}}. Following the standard repair (June 2026 correction, review finding F7), we define the a priori semimeasure M via a universal monotone machine U_m: M([x]) = \sum_{p \,:\, U_m(p) \succeq x} 2^{-|p|}, summing over minimal programs whose (possibly infinite) monotone output extends x. The associated complexity is the monotone complexity K\!M(x) = -\log M([x]). Its relation to the discrete prefix complexity is one-sided only: K\!M(x) \le K(x) + O(1), while the gap K(x) - K\!M(x) is unbounded in general (Gács 1983). The earlier claim that the continuous construction is multiplicatively equivalent to the discrete semimeasure (M \asymp \xi, so that M-null and \xi-null sets coincide) is therefore withdrawn: the discrete coding theorem does not transfer to the continuous setting. This brings the appendix into line with the v3.6.0 §3.1 semimeasure sharpening in the main preprint.

(Note: M is a lower-semicomputable semimeasure on the Cantor space — M([x]) \ge M([x0]) + M([x1]), with the deficit corresponding to program mass whose monotone output halts at a finite string — and the total mass surviving on infinite sequences satisfies M(2^{\mathbb{N}}) < 1. We explicitly define the normalized probability measure \tilde{M} = M / M(2^{\mathbb{N}}). While \tilde{M} is only lower-semicomputable up to the non-computable normalization constant M(2^{\mathbb{N}}), all subsequent “almost surely” statements operate with respect to \tilde{M}. The discrete coding-theorem identity K(x) = -\log \tilde{M}(x) + O(1) asserted in earlier versions is not available here; the correct relation is K\!M(x) = -\log M([x]) together with the one-sided bound K\!M(x) \le K(x) + O(1) (Gács 1983).)

3. M-Martin-Löf Randomness

To formalize the nature of the generative space, we invoke Martin-Löf (ML) Randomness. However, one must distinguish between continuous measures. A sequence \omega that is ML-random with respect to the uniform (Lebesgue) measure \lambda behaves entirely differently from a sequence that is ML-random with respect to M.

Because the OPT substrate evaluates probability by algorithmic simplicity, the relevant formalism relies on \tilde{M}-Martin-Löf randomness. The foundational theorem of AIT states that for any computable probability measure \mu, the set of \mu-ML-random sequences holds \mu-measure 1. Extending this result to lower-semicomputable semimeasures (cf. Nies 2009, §3.2 “Randomness for arbitrary measures”), the set of all \tilde{M}-Martin-Löf random sequences successfully retains measure 1 with respect to \tilde{M}.

Therefore, \tilde{M}-almost-all infinite substrate sequences are strictly \tilde{M}-ML-random.

(Note: Utilizing \tilde{M}-ML-randomness structurally guarantees that the typical outputs of the substrate are drawn self-consistently from the biased, highly structured algorithmic measure \tilde{M} rather than uniform noise, providing the mathematical scaffolding for the structural-frequency conjecture below (Conjecture P-1a).)

4. M-Normality vs. Borel Normality

A highly significant — now explicitly conjectural — consequence of M-ML-randomness relates to structural frequency. Under uniform Lebesgue-ML randomness, a sequence is strictly Borel normal—generating every finite binary string of length k with an identical, uniform frequency.

However, since \tilde{M} is decidedly non-uniform—skewing heavily to assign massive probability weight to algorithmically simple, compressible, lawfully structured patterns—\tilde{M}-almost-all sequences are NOT uniformly Borel normal. Instead, we define their structural limits via \tilde{M}-normality.

Because the measure \tilde{M} is fundamentally non-stationary (algorithmic probability depends on absolute prefix position), we cannot rely on standard ergodic frequency-convergence limits. Formally, we define \tilde{M}-normality by the weaker but strictly sufficient property of infinite recurrence.

Conjecture P-1a (Recurrence — June 2026 demotion from theorem grade). Restricted to the non-atomic, observer-compatible component of the substrate (cf. Preprint §3.1, which already excludes simple repeating sequences from the observer-compatible class), \tilde{M}-almost every sequence contains every finite string x — in particular the discrete formal configuration of a conscious observer (K_{\text{obs}}) — infinitely often.

Heuristic argument (formerly presented as a proof; relabelled per review findings F7/F8). The intended derivation ran as follows. Since \tilde{M}([x]) > 0 for all finite strings x, the chain rule for prefix Kolmogorov complexity, K(sx) \le K(s) + K(x) + O(1), would yield the lower bound M([s \cdot x]) \ge M([s]) \cdot M([x]) \cdot 2^{-O(1)} — a supermultiplicativity lower bound, mislabelled “near-submultiplicativity” in earlier versions. The conditional probability of x appearing in any window, given any prior prefix s, would then be uniformly bounded below (\tilde{M}([x] \mid [s]) \ge \tilde{M}([x])/c > 0), and the conditional Borel-Cantelli lemma applied to non-overlapping windows of length |x| would deliver infinite recurrence almost surely. Two explicit gaps block this from being a proof. (i) The discrete/continuous conflation: the step from the prefix chain rule to the cylinder-measure bound requires the discrete coding-theorem identity K(x) = -\log \tilde{M}([x]) + O(1), which fails in the continuous monotone-machine setting of §2, where only the one-sided K\!M(x) \le K(x) + O(1) holds and the gap is unbounded (Gács 1983). (ii) The atomic-measure obstruction: the a priori measure has atoms — individual computable sequences (e.g., 000\ldots) carry strictly positive \tilde{M}-mass — and on such an atom the recurrence claim is false outright, so no almost-sure recurrence statement can hold over the full space 2^{\mathbb{N}}; this is why the conjecture is stated on the non-atomic component. A proof would have to both repair the conditional lower bound with monotone-complexity tools and control the conditional structure of \tilde{M} on the non-atomic component; neither step is currently available.

5. The Computational Realism Postulate

Conjecture P-1a asserts — it is no longer claimed as an AIT guarantee (June 2026 demotion) — that the finite representation of any observer (K_{\text{obs}}) appears as a structural sequence infinitely many times within the non-atomic \tilde{M}-ML-random substrate.

However, mathematical information theory cannot inherently cross the boundary into physical ontology. A finite string occurring on the output tape of a Turing machine is a static artifact of execution—a snapshot. A coherent observer requires continuous internal dynamics, relational coupling, and active inference looping. The string itself does not “feel” anymore than a brain scan stored on a hard drive is conscious. The execution belongs to the generating program, not to the resulting snapshot code.

To assert that the uncomputable continuous limits governing the mathematical substrate structurally give birth to ontologically real, causally active phenomenological universes, OPT must make a single explicit metaphysical commitment.

Postulate (Computational Realism): In an infinite uncomputable substrate governed by identical mathematical dynamics, abstract mathematical computation formally equivalent to the causal description of an observer (where formal equivalence is defined as computational isomorphism of the observer’s causal state-transition structure) possesses causally efficacious, ontologically real existence. Furthermore, structurally discrete computational instantiations across the substrate possess independent ontological individuation, constituting distinct subjective counterparts (and by the foundational phenomenality axiom in Preprint §8.1, such causally efficacious observer-equivalent computations constitute genuine subjects of experience).

6. Conjecture P-1 (Informational Normality)

By uniting Conjecture P-1a with the Computational Realism Postulate, the case against solipsism is stated in its sharpest available form — now explicitly conditional rather than theorem-grade.

Conjecture+Postulate P-1 (Informational Normality — June 2026 demotion): Under the generalized algorithmic prior, the continuous substrate operates via \tilde{M}-Martin-Löf randomness almost surely. If Conjecture P-1a (Recurrence) holds on the non-atomic observer-compatible component, the mathematical occurrence of every finite structural observer description K_{\text{obs}} recurs infinitely many times. Operating upon this scaffolding, the Computational Realism Postulate bridges these generating mathematical artifacts into ontological physical reality. Providing that both Conjecture P-1a and computational realism hold, the existence of structurally equivalent, causally active, and uniquely individuated counterpart observers across the substrate follows.