What These Chapters Owe

Before the act turns to reference, an accounting. The chapters just finished make the loudest claims in the book and rest on the thinnest construction in it, and Remark 57.2 promised that this arrangement would be stated rather than disguised. This is the statement.

63.1 Open gaps

1. Colimit at limit ordinals. Definition 58.3 defines \(\fiber{\lambda}\) as the colimit in \(\mathbf{GSLT}\) for limit ordinals \(\lambda\). The existence and properties of this colimit require verification. In particular: does the colimit GSLT have a well-defined bisimulation quotient, and is it the union of the bisimulation quotients of the stages below?

2. Subject reduction for schedulers. This gap originally asked for a construction of the oracle as a persistent receive in the rho calculus. Chapters 59 and 60 argue the gap was the wrong shape: the oracle need not be constructed at all, because the resolution of a contended channel is already a choice function and the scheduler that supplies it already spends a graded choice resource. What is owed instead is the metatheorem. Does the term assignment of Chapter 61 satisfy subject reduction — can a scheduler, merely by running, come to realize a stronger choice principle than it was typed for? If it can, the graded factorization is not a reduction invariant, an island can crack while it runs, and that cracking is what one agent perturbing another amounts to. This is Question 60.1 and Question 60.2, which are the same question, and it is the load-bearing gap of these chapters.

3. Strict refinement over the lattice order (Proposition 62.2). The proof sketch uses a diagonalization argument, and it was restated in this pass for the lattice order \(v < w\) rather than for an ordinal successor. The full proof requires specifying the encoding of the scheduling problem characteristic of \(w\) as terms and verifying that the diagonalization goes through in the GSLT setting. The restatement makes the gap slightly wider than it was: for a successor there was an obvious candidate problem, and for a general pair \(v < w\) one must be produced.

4. New currents (Proposition 62.4). The claim that new conserved quantities appear higher in the lattice requires identifying concrete symmetries of the cost map at \(w\) and verifying that they give rise to conserved quantities via the Noether argument of Part Part IV.

5. The standing charge of a position. Section 62.5 asserts that occupying a lattice point is a recurring cost and not a one-time acquisition, and the unfavorable cycle turns on it. Chapter 60 prices individual resolutions; it does not price the maintenance of the capacity to resolve. Nothing in the book supplies that quantity, and Remark 62.11 needs it.

6. Effective resolution. Definition 62.1 requires that the agent can form and evaluate the queries a position resolves. A precise definition of effective access, in terms of the agent’s account and the expressiveness of its hypothesis language, is needed.

7. Approximate resolution (Definition 62.2). Remark 62.12 restates this on the stochastic instance, correcting a squared amplitude that should not have survived the withdrawal in Chapter 33. What the restatement leaves open is whether approximate resolution places an agent at a genuine intermediate point of \(\mathfrak{W}\), or merely grades its reliability at a point it already occupies. These are different claims and the earlier presentation ran them together.

8. Internalization is a conjecture (Conjecture 26.1). Chapter 26 states the selection argument for internalization in the form it would need to be checked in — a threshold in the variance of inflow — and does not check it. The machinery to check it exists: Chapter 13 supplies propensities and a well-posed simulator. The experiment has not been run, and until it is, the price of a self-model is located rather than known.

9. The hard problem. These chapters offer structural analogues: mathematical objects with the right functional properties. They do not address why there is something it is like to be an agent reading its own drift at a point in the lattice. The explanatory gap between functional description and phenomenal experience remains open and is not claimed to be resolved here. Chapter 26 chooses a heavily freighted word for a structure, deliberately; this gap is the price of that choice and we do not pretend the word pays it.

63.2 Relation to existing work

The book’s comparisons with neighboring programs — integrated information theory, the free energy principle, global workspace, constructor theory, the ruliad, AIXI, and assembly theory — are gathered in Chapter Related Work, and Where This Differs at the end of the Pledge, where a reader can find them before committing to six hundred pages rather than after. What belongs here is only what is specific to these chapters.

Turing’s oracle hierarchy [12] is the direct inspiration for the tower of Chapter 58, and the departure is the move off the chain: first to a family indexed over a category of theories, and then, once Remark 58.2 has conceded that this family is not the fibration it was called, to a coordinate in the Weihrauch lattice.

The account of sentience is where the largest change from earlier drafts lies, and it is worth saying what it was. Sentience was a distinguished component of the vectorial account driven by a reinforcement signal on predictive accuracy, which made it formally a reinforcement-learning construct with an internal reward [13]. Chapter 26 withdraws that and replaces it with a formula the learner holds about its own reserves. The gain is that the reward function stops being a free parameter: there is nothing left to choose, because the signal is the learner’s own assay of its distance from a goal it carries. The loss is that the construction now depends on the instrument discipline of Chapter 20, which is itself conditional on two unproved hypotheses. We think that is the right trade — a derived quantity resting on stated hypotheses is better than a posited one resting on nothing — but the reader should see it as a trade.

One correction, since earlier drafts made the claim and it should not stand. The treatment of consciousness was described as ordinal-valued. That description was superseded by Chapters 59 and 60, which locate the vertical coordinate in the Weihrauch lattice — emphatically a lattice and not a chain. Consciousness here is a position an agent occupies, not a threshold it crosses, and the ordinal tower survives only as the chain-shaped shadow used to introduce the intuition.

63.3 Where the argument stands

Taken together with what precedes them, these chapters complete the framework’s third leg.

• Part Part IV establishes the physics: causal structure, dynamics, and a semiring-parametric path weight from the bare structure of a GSLT. The complex instance of that weight was proposed and then withdrawn in Chapter 33; what the framework carries is the Boolean, tropical and stochastic instances, and the last of those is what everything downstream actually uses.

• Part Part V establishes the epistemology: why situated agents in massive populations perceive a hierarchical false ontology rather than the true bisimulation quotient.

• These chapters establish the phenomenology, in the deflationary sense the book can support: why a learner that watches its own reserves has states that behave like affect, and why a learner’s position fixes not only what it can do but what there is.

The unifying theme is that the category of GSLTs is not merely a setting for studying computation. It is a setting rich enough to ground physics, epistemology, and — if the arguments here have merit — phenomenology.

What remains is the question the Prestige exists to ask. All of the above was got by restricting scope to computation, and the restriction bought an ontology and a grounded language free. Chapter 64 asks what would be required to ground a symbol system without that restriction, and Chapter 65 asks what the answer implies about the world we are in.