The maze is a working example of the AI of the Future based on the Vacuum.net-theory.

Have a look at the Maze push here

J.Konstapel,Leiden,24-8-2026.

Short Summary

The Maze proposes a self-addressing architecture in which knowledge receives computed addresses using balanced ternary rather than assigned identifiers.

Its inhabited net has been tested with 65 disciplines, about 1,100 Wikipedia articles, and the complete periodic table.

A quaternion-based transformation provides the first mathematically verified candidate for movement through the knowledge space, preserving depth and structure.

The key unresolved question is whether this mathematically valid movement also corresponds to meaningful relationships between pieces of knowledge.

The central problem addressed by The Maze, Inhabited is deceptively simple: how can knowledge be given an address without requiring an institution to decide where that knowledge belongs?

The Challenge

Most systems we use today rely on assigned addresses. Books receive classification numbers. Documents receive identifiers. Web resources receive URLs. Academic publications receive DOIs. Modern AI systems take a different approach, representing knowledge as positions in continuously changing vector spaces. Despite their differences, these systems share an important characteristic: the address is assigned by a system outside the object itself. When the system changes, the address can change with it.

The Maze proposes a different principle. Instead of assigning an address, it computes one. A repeatable rule examines a piece of knowledge and produces a sequence of three possible distinctions: −1, 0, or +1. These are represented as balanced ternary digits. The resulting number is not merely an identifier. Its structure contains information about scale, opposition, distance and refinement.

This distinction is the foundation of the project.

A conventional identifier answers the question, “What label have we given this object?” The Maze attempts to answer a different question: “Where does this object belong according to a rule that can be applied again, independently, and at any scale?”

That change has significant consequences.

A coordinate system for knowledge

The use of balanced ternary is central because three-valued distinctions provide a natural middle state between two opposites. Each step in a route can move above a reference, remain on it, or move below it. A sequence of such decisions becomes a coordinate.

The mathematical advantage is that the resulting coordinate system has a built-in hierarchy. Removing the final digits does not merely shorten an identifier. It coarsens the address. The more detailed position is reduced to a broader one. In this sense, truncation behaves like rounding.

The opposite operation is equally simple. Reversing every sign changes +1 into −1 and vice versa. The address therefore has an intrinsic antithesis: its mirror position is obtained without consulting a database.

Distance can be obtained through subtraction. Refinement is represented by extending the address. Two independently constructed archives can therefore, in principle, be merged because they are not dependent on a common numbering authority.

This is what makes the proposal more ambitious than a new classification scheme. The address is intended to be an element of the structure rather than a label attached to it.

The paper formalizes these properties as propositions concerning truncation, antithesis and scale.

From theory to an inhabited space

A classification theory becomes considerably more interesting when it is populated with actual material. The paper therefore reports three forms of population.

The first consists of sixty-five disciplines drawn from Paths to the Knot. The second is a random sample of approximately one thousand English Wikipedia articles. The third is the periodic table.

The results suggest that the architecture does not merely provide an abstract coordinate system. It produces repeated locations. Approximately 1,100 coded pieces occupy 248 windings. In other words, many different pieces of knowledge arrive at the same structural location when subjected to the same questions.

This is presented not as information loss but as the discovery of shared form.

The periodic table provides a particularly clear demonstration. Ring 5 has a capacity extending to ±121, while the periodic table contains 118 elements. The elements therefore fit into the ring with three positions remaining. The empty positions can be represented explicitly rather than requiring them to be added later by an administrator.

This is reminiscent of Mendeleev’s use of gaps in the periodic table. The historical importance of those gaps was not that they represented missing labels, but that they represented missing members of an underlying structure.

The Maze attempts to generalize this principle: an empty location can itself become informative.

The experiment is therefore important, although its evidential scope should remain clear. The periodic table demonstrates that the architecture can represent an existing natural ordering particularly cleanly. It does not by itself establish that the system can discover unknown knowledge. That stronger claim requires predictive experiments.

The importance of failure

One of the most revealing results in the paper is a zero.

When the two independent encoders were required to agree in the proving configuration, none of the sixty-five articles passed the admission gate. At first sight this appears to be a failure of the system. The paper argues, more convincingly, that it is also evidence of a useful safety property.

The architecture requires every move to carry a literal ground: a span of text that actually occurs in the source material. A plausible interpretation is not sufficient. A system cannot simply invent evidence for a classification decision.

This produces an unusual asymmetry with generative AI. A language model can always provide an apparently plausible answer. The Maze’s admission mechanism can instead return nothing when the required evidence is unavailable or when its independent readings disagree.

That is an important architectural idea: uncertainty is represented not by increasingly fluent language, but by refusal to make an unsupported move.

At the same time, the zero exposed a design problem. The original agreement mechanism was being applied across routes substantially longer than the four-ring unit for which it had been conceived. The paper consequently proposes block-wise gating as the next specification revision. The failure therefore functions as an experiment on the architecture itself.

The missing ingredient: movement

At this point the Maze has an address space, but an address space is not yet an active architecture.

This is the problem identified in the paper as O2: the core can address, but it cannot move.

Truncation changes resolution but loses depth. Antithesis preserves depth but merely moves back and forth between two states. Other obvious transformations, such as reversing or rotating the route, preserve some properties but interfere with the scale structure.

The paper therefore asks a deeper question: can knowledge move through the Maze according to a lawful transformation that preserves its structural properties?

The proposed answer comes from quaternion algebra.

The four rings naturally form a block of four moves. These four balanced-ternary digits can be treated as the four components of a quaternion. Left multiplication by the quaternion unit (i) then produces a transformation that rotates the components within the block.

The resulting operation preserves depth and the allowed digit values. It is reversible. Applying it twice produces the complete antithesis, and applying it four times returns to the starting point.

The significance is not simply that the operation is mathematically elegant. It satisfies the three requirements that the paper has established for a candidate movement: it preserves depth, remains closed within the architecture, and actually goes somewhere.

The transformation was also exhaustively checked over all 6,561 routes of depth eight.

Symmetry or law?

This is where the paper reaches its most important unresolved question.

The quaternion transformation is a valid mathematical operation on Maze addresses. But a mathematical operation is not automatically a law of knowledge.

For the transformation to become a genuine law of motion, the address reached by moving from one object must have some meaningful relationship to the original object. The fact that the transformation is elegant and closed is not sufficient.

This distinction is crucial.

The paper itself states that the candidate is not yet a law. The next experiment must therefore test whether quaternion neighbours are semantically or structurally related in the inhabited net.

That experiment would represent a major transition. Until then, the quaternion construction demonstrates that the Maze can move without destroying its architecture. The next question is whether the world itself recognizes that movement.

A broader mathematical pattern

The paper extends the construction through the Cayley–Dickson sequence:

real numbers, complex numbers, quaternions and octonions.

The resulting observation is that richer number systems impose increasingly strict requirements on coarsening. The complex construction operates on blocks of two, the quaternion construction on blocks of four, and the octonion construction on blocks of eight.

This produces an intriguing relationship between representation and scale. As the algebraic structure becomes richer, it becomes less tolerant of arbitrary truncation.

In the language of the Maze, coarsening is therefore not a neutral operation. It interacts with the internal structure of the representation.

The paper proposes this as a possible deeper principle rather than presenting it as an established law of nature. That distinction is important. The computations establish the stated closure properties; they do not yet establish why the same hierarchy should govern knowledge or physical systems more generally.

What has actually been demonstrated?

The achievement of The Maze, Inhabited is best understood as a sequence of increasingly demanding demonstrations.

First, it specifies a computable address system.

Second, it shows that the basic arithmetic of balanced ternary supplies several useful structural operations without requiring an external indexing authority.

Third, it populates the resulting space with real material.

Fourth, it demonstrates that the admission mechanism can reject unsupported classifications rather than manufacture them.

Fifth, it supplies a mathematically verified candidate for movement through the space.

What has not yet been demonstrated is equally important. The paper has not established that the computed address is universally superior to existing information architectures. It has not established that structural collisions always reveal useful relationships. And it has not established that quaternion movement corresponds to meaningful movement through knowledge.

Those are not defects in the present paper so much as the precise experiments that define the next stage of the project.

The significance of the experiment

The larger significance of the Maze is therefore not that it proposes another way of indexing documents.

Its more radical proposition is that knowledge might be organized by a rule that generates its own coordinates.

If that principle works, an archive no longer needs to know the future in order to accommodate it. New material does not necessarily require a new classification authority. It can be evaluated by the same rule and assigned a position within the existing mathematical space.

The result would be an architecture in which empty space, distance, opposition, refinement and movement are not additional metadata imposed on knowledge. They would be consequences of the addressing rule itself.

That is the central idea worth testing.

The present paper moves the Maze beyond a theoretical numbering system. It gives the architecture inhabitants, measures its behavior, exposes a real failure in its admission mechanism, and proposes the first operation that allows the inhabitants to move.

The decisive experiment now lies ahead: determine whether the movement generated by the mathematics corresponds to relationships that exist independently in the knowledge being mapped.

If it does, the Maze will have demonstrated something considerably stronger than a clever coordinate system. It will have begun to demonstrate a computational geometry of knowledge.

If it does not, the quaternion construction will still have achieved something valuable: it will have identified, with unusual precision, the boundary between mathematical symmetry and meaningful dynamics.

That is the point at which The Maze, Inhabited leaves the reader: not with a finished system, but with a testable architecture and a sharply defined question about whether an address can become a path.