The theory
Alexandria in one page
For anyone. About 880 words. Citations are to the treatise by Part and paragraph.
What it is. Alexandria is a general theory of intelligence, and its answer is simple. Intelligence is not something a person or a machine has, like height or a test score. It is something a mind does: working out what could not simply be looked up. (Treatise 0.65)
The picture. Picture everything humanity knows as one great structure. The most certain things, such as mathematics and settled science, are packed tight at the centre. Newer knowledge, confirmed but not yet built upon, sits further out. Around the edge lies everything not yet known. Each of us carries a structure like this too: everything we have worked out, or taken on, in our own lives. The same account describes both, and most of what follows is about how the two meet. (Treatise 1.1 to 1.5, 1.14, 2.38)
Four ideas carry the rest. A reader who holds these four can follow everything else. (Treatise 0.64)
- Looking something up and working something out are different, even when they produce the same answer. Looking up moves through what is already known. Working out crosses into what was not. (Treatise 2.28, 2.94)
- Working something out leaves a record: what was tried, what failed and where it closed. A claim with no record beneath it is information or assertion, however confidently it is held. (Treatise 2.70, 2.74)
- Whether something was new is a fact about what humanity already knew, not about who or what produced it. That is what makes the line checkable, for a person or for a machine. Something can be new for you and already known to everyone else, and the theory treats that as real for you. (Treatise 2.29, 2.76)
- Any two bodies of knowledge stand in a relation from the moment either exists: what one holds shapes what the other can see and reach. (Treatise 10.7, 10.30)
One definition. Everything in the treatise is derived from a single definition:
Intelligence is not a property of a system. It is a property of a process: the irreversible, locally directed decrease in entropy within a structured possibility space, achieved through inductive traversal that generates constraint geometry not derivable by deduction from prior knowledge, transforming coherence potential into an actualised INT in a manner that is unreplicable without the specific path taken. (Treatise 2.19)
In plain words: a crossing from what is known into what is not, which cannot be undone, and which could not have been reached any other way.
The definition and four more. From that definition, with nothing further assumed, come its plain form and four further answers to things the fields have long argued over. (Treatise 11.4, 12.23)
- Intelligence is a crossing beyond what was known. The question to ask of any system is not how much it has, but whether it crossed.
- Knowledge is the record of those crossings, not a stock of true beliefs.
- Power is how far one person's or group's knowledge shapes what another can search, read, commit to or reach, whether or not anyone acts.
- Strategy is working something out while someone else works to stop you.
- The governance of knowledge is keeping the record, with an audit anyone can rerun, so that what holds it together is not trust in an institution. (Treatise 12.24 to 12.28)
What it changes.
Machines. Today's systems extend what can be looked up, at a speed and scale nothing before them approached, and that brings people to the edge faster. None has been shown to work something out and hold it as its own. Whether one has is a test fixed in advance, not moved afterwards. What human intelligence is for is the step off the edge, and holding what it reached. (Treatise 7.10, 11.38, 12.9)
Misinformation. A claim with real work behind it looks much like one that is simply repeated. Alexandria specifies how the record behind a claim can be made visible, so that anyone could see whether it was earned. (Treatise 11.58 to 11.63, 12.30)
Education. Education has long argued over whether to pass on what is known, teach skills, or teach people to think for themselves. All three are needed and none replaces another. The aim becomes building a person's own structure rather than placing someone else's in it. (Treatise 11.68 to 11.72, 12.7)
Science and credentials. When what counts is the record of the work, a credential stops being the gatekeeper, and a good idea no longer needs permission. (Treatise 7.16, 12.33)
A life. Everything a person genuinely works out is a permanent addition to what is known, whether or not anyone applauds it. (Treatise 12.34)
What could be built. Alexandria is also a design. The treatise presents two things that are worth holding apart: a theory of intelligence, and the specification of a record of knowledge that others can build, govern and use. The specification follows from the theory; it is not an application added to it. That record can be built as one open, shared repository: current, auditable by anyone, and governed so that no single party owns it. Nothing has yet been built to it. The first Library of Alexandria was lost because its record lived in one place. This one is designed so that cannot happen. (Treatise 0.1, 11.22, 11.52, 11.54, 12.32)
How to test it. The theory is read against thirty-three fields, from mathematics, thermodynamics and quantum physics to law, medicine and ethics, and against thirty-eight paradoxes. In each it records where the field agrees and where it does not. It names the specific results that would show it wrong. The challenge it welcomes is the kind it describes: new work that shows where something does not stand. (Treatise 0.2, 7.11)
The treatise ends with five words that sum up what it asks of each of us: Don't follow maps. Build them. (Treatise 11.41)
For specialists and academic readers, the same account with full references follows: Alexandria in outline.
The theory, second page
Alexandria in outline
For specialists and academic readers. About 2,750 words, with full references to the treatise.
Alexandria is a general theory of intelligence. It specifies what intelligence is as a process. It derives the boundary that separates intelligence from what resembles it, and grounds its irreversibility in the physics of observation. It places human, animal, artificial and collective intelligence under one geometric architecture. (Treatise 0.1)
Everything in the treatise is derived from a single definition:
Intelligence is not a property of a system. It is a property of a process: the irreversible, locally directed decrease in entropy within a structured possibility space, achieved through inductive traversal that generates constraint geometry not derivable by deduction from prior knowledge, transforming coherence potential into an actualised INT in a manner that is unreplicable without the specific path taken. (Treatise 2.19)
The definition differs from existing accounts in three respects that can be checked:
- It is a process with a boundary. A crossing occurred or it did not.
- It is independent of substrate by derivation, not by assertion, because nothing in the process names what carries it.
- It produces the criterion for what is not intelligence as a consequence.
(Treatise 0.65, 12.10)
The treatise presents two things that the reader should hold apart. One is a theory of intelligence, derived in Parts Two to Seven and Ten. The other is the specification of a record of knowledge that others can build, govern and use, stated in Part Eleven and claimed in Part Twelve as the theory's institutional consequence. The second is entailed by the first and is not an application added to it. (Treatise 0.1)
Within that, the treatise makes its claims at four levels, and each is best tested at its own:
- The mechanism: how a crossing is made, confirmed, placed and integrated, at the scale of a person and of a civilisation (Parts Two to Seven).
- The chain of five definitions derived from it: intelligence, knowledge, power, strategy and the governance of knowledge. (Treatise 12.23)
- The consequences for institutions and for a person: for machine systems, education, science, the information environment, and the personal audit (Part Eleven). (Treatise 2.164)
- The further terms the same object defines, each stated where the architecture needs it. (Treatise 12.29)
Set out this way, each level can be tested on its own terms, rather than the whole being tested against its strongest claim.
2. The structure, individual and collective
Everything humanity knows forms one geometric structure, The Knowable. Each confirmed position in it is an INT, the mark the treatise uses for what a crossing produces. Each carries:
- a position held once;
- a record of the directions approached and refused in reaching it;
- connections to the positions around it;
- open faces that state what would have to arrive there to fit.
At the centre sits Coherence, the densest and most tested region. Beyond the edge lies the Dark Space. Settled structure gives off luminosity, so the frontier is lit from behind: a traverser can see where to look, but not what will be found. (Treatise 0.68, 2.74, 2.79, 2.118)
The same structure exists at two scales, built by the same mechanisms. Each person holds a personal Knowable: everything they have worked out or taken on. What they reach exists first there, real whether or not anyone else ever holds it, and most of it stays there. The collective structure is what civilisation holds. (Treatise 0.43, 2.15, 2.38)
The two meet at one point, submission. An INT enters the collective structure only by being submitted with its record. The Deductive Audit reads that record. It tests whether what was reached lies inside or outside what the collective structure already entailed, places what is new where its relations put it, and sets off the reassessment cascade by which the structure reorganises around the arrival. (Treatise 2.17, 2.75, 2.83 to 2.105, 2.143)
The traverser's own placement stands in their own structure, unchanged. The two records of one piece of work are both true. Where they differ, the difference is the only measure of self-location that does not depend on judging oneself. It has two sources that the comparison does not separate: the traverser's own reading of where they stood, and the movement of the collective structure between launch and submission. And it can be read only where the submission was confirmed, since one returned as assertion leaves no collective record to compare against. (Treatise 2.144, 2.146, 12.4)
3. The boundary
The confirmed structure at a given time entails a set of positions, its deductive closure. A crossing is an inductive event if, and only if, what it produces lies outside that closure. Everything inside it can be reached by retrieval along the highway, however much effort the reaching takes. (Treatise 2.28, 2.94)
Three features make the boundary usable:
- It is set by the collective structure, not by the traverser. So a person can genuinely work out something already known: a personal rediscovery, real for them and new for nobody else. (Treatise 2.29, 2.76)
- The audit reads the record. It does not search the closure. (Treatise 2.143)
- The record, not the endpoint, shows the difference. The outputs of retrieval and traversal can be identical while the processes differ. (Treatise 0.66)
4. Four concepts
The architecture rests on four ideas whose combination produces everything that follows. (Treatise 0.64)
- Traversal is not retrieval. (Treatise 2.28, 2.94)
- A crossing leaves a record; an assertion does not. (Treatise 2.70, 2.74)
- The boundary between the two is a fact about the collective structure, not about the traverser. (Treatise 2.29)
- Any two confirmed structures stand in a relation from the first confirmation by either. What one holds determines what the other can search, read, commit toward or cross. Power, strategy and governance are derived from that relation rather than added to the theory. (Treatise 10.7, 10.30)
5. One definition, five definitions
The definition of intelligence yields, by entailment, four further definitions. Each replaces a definition its field had worked with and could not complete. (Treatise 11.4, 12.23)
Intelligence. Prior definitions treat intelligence as a quantity a system has in some degree: a test score, expected performance across environments, skill-acquisition efficiency. On this account it is something a structure does or does not do on an occasion, and it can be read from the record. (Treatise 12.24)
Intelligence is not a property of a system. It is a property of a process. (Treatise 2.19)
Knowledge. Since Gettier, justified true belief has been known to admit cases that are not knowledge. The architecture does not amend the definition; it replaces the object. Every Gettier case is a position held with a record that does not belong to it. Justification, truth and belief become one object read three ways. (Treatise 12.25, 4.4, 8.150)
Knowledge is the temporal reading of confirmed geometry, the state, at a given moment, of everything the Deductive Audit has confirmed: the positions established, how much of each boundary has resolved, the record of the traversal that produced them, and the open surfaces at which the structure ends together with what each states would be required to extend it.
Power. Dahl, Lukes and Weber define power by acts, and Lukes's third dimension rests on a counterfactual that cannot be observed. Here power is derived as a relation between two structures. It runs by six routes and holds whether or not either party acts. Theorems IX and X state where it holds and in what extent. Authority follows as acceptance on a warrant. (Treatise 12.26, 10.8, 10.30, 10.31, 10.221)
Power to is the confirmed structure a traverser holds, together with the vantage that follows from holding it. Power over is the relation in which one such structure determines, to the extent that it does, what another traverser can search, read, commit toward or cross.
Strategy. Ends, ways and means is a frame with no object beneath it. The object that the commercial, military and assessment literatures shared without naming is adversarial traversal. The five adversarial routes all act on one thing, the structure's own reading, and the record is the diagnostic. (Treatise 12.27, 8.200 to 8.251, 10.98 to 10.104)
Strategy is adversarial traversal, which is contest conducted from a position: the selection, from the confirmed structure a traverser holds, of which contested absence to commit toward, and the conduct of that commitment where another structure acts on the traverser's reading of the terrain, by the routes at 10.10, to prevent the confirmation that would reach it.
Governance of knowledge. Governance has rested on the trustworthiness of institutions that vouch for what they admit. Here it rests on recomputation, not trust, and no discretionary act is left exposed to corruption. (Treatise 12.28, 11.54, 11.55)
Governance of knowledge is the keeping of the record and the integrity of the audit that reads it: a record kept with kind, receipt, precedence and the gate, so that every relation in it can be read rather than inferred, and an audit that is determinate, public and self-applied, held in the submitted disposition, so that nothing it determines rests on who computes it or on whose undertaking not to corrupt it.
The same object defines twenty further terms that the fields have held by measure or by trait: consciousness, complexity, wisdom, emotion, expertise, competence, learning, creativity, understanding, choice, persuasion, organisation, retrieval, truth, identity, character, risk, change, belief and value. None is an addition to the chain; each is the one definition read at a different place in the geometry, and each is stated where the architecture needs it. (Treatise 12.29)
6. Physical grounding
A crossing's irreversibility is a physical claim, and the treatise states it where physics can refute it.
Thermodynamics. The cost of a crossing is attached to the possibilities it forecloses. This is the Landauer Bridge, carried as an open formalisation target.
Quantum foundations. For facts that need a consciousness to become determinate, inductive traversal and measurement are one event seen from two sides.
The Container Inversion. This sets out the architecture's position on time and observer-independence, against Barbour's and Rovelli's programmes.
(Treatise 0.11, 8.23, 8.38, 5.22, Annex B.2)
7. The invariant and the theorems
One invariant follows from the mechanisms: intelligence preserves and increases option-generating capacity under geometric constraint. (Treatise 2.157)
Ten theorems follow as entailments. (Treatise 0.70)
- Theorem I: the outputs of intelligence cannot be reproduced by any process working within the prior deductive closure. It sets the structural conditions under which P ≠ NP is the expected result. (Treatise 7.4)
- Theorem II: a process is intelligence only if it both crosses the boundary and is integrated into permanent consequence. (Treatise 7.8)
- Theorem III, the Safety-Consciousness Mutual Exclusion: a system designed to preserve existing structure against disruption cannot also be intelligent. (Treatise 7.12 to 7.15)
- Theorem IV: peer review as the test of validity, citation as significance and credential as warrant each misidentify what they stand in for. (Treatise 7.16 to 7.21)
- Theorem V: zero-knowledge proof follows from the separability of a confirmed INT from the record that produced it. (Treatise 7.22)
- Theorem VI: organisations are not intelligent, but they can be structured to amplify or suppress the intelligence of those within them. (Treatise 7.24)
- Theorem VII, Integration Productivity: in a mature structure, crossing toward the gaps between separated regions returns more than deepening within one. (Treatise 7.35)
- Theorem VIII, the Alexandria Entailment: it states what follows if the definition is accepted, not whether it is correct. (Treatise 7.38)
- Theorem IX: power over another holds in extent only where the collective structure is incomplete or unevenly held. (Treatise 10.31)
- Theorem X: the relation of power holds between any two confirmed structures from the first confirmation by either, forward in time and beyond refusal. (Treatise 10.30)
8. The evidence
The architecture is read against thirty-three domains, from mathematics, thermodynamics, information theory and quantum foundations to law, medicine, ethics and AI safety. It is also read against thirty-eight paradoxes it was not built to address, which resolve to six primitives and four mechanisms.
In each domain, the field's own findings are stated first, in the field's own vocabulary. The architecture's reading follows, with where they converge and where they diverge. A domain that observed something the architecture does not explain is recorded as having done so. The recurring pattern is that each field approached the same boundary from its own direction and stopped at it. (Treatise 0.2, 0.71, 9.1)
9. What it provides now
Artificial intelligence. Current systems extend the highway: retrieval at a speed and scope no prior instrument approached, bringing traversers to a better vantage faster. No system has been shown to integrate what it reaches. The treatise specifies, in five components, what an artificial system would need in order to cross and integrate. It also supplies the basis for a coherent approach to alignment, though not a demonstrated method: an aligned system increases option-generating capacity under constraint, which Theorem III shows cannot be pursued in the same system as the current safety paradigm. What would demonstrate it is a system meeting the integration condition. (Treatise 11.30 to 11.38, 11.37, 12.9, 12.31)
Education. The collision between transmitting a canon, teaching skills and fostering critical consciousness is three traditions each holding one of three variables: confirmed structure, the capacity to cross, and vantage. All three are necessary and none substitutes for another. The instruments that follow are:
- building a learner's own structure, not placing someone else's in it;
- the curriculum as a sequence of absences;
- assessment as audit;
- self-location as the first task.
(Treatise 11.68 to 11.72, 12.7)
Science and authority. Standing is geometric, not institutional. A validity condition that reads the record and not the warrant frees traversal from institutional permission. (Treatise 7.16 to 7.21, 12.8, 12.33)
The information environment. Alexandria specifies how the difference between a position with a record and one without can be made permanently readable. The diagnosis holds before anything is built; the resolution holds wherever the record is kept. (Treatise 11.58 to 11.63, 11.22, 12.30)
A person. The personal audit is a way to tell what one worked out from what one took on, and to locate oneself honestly. A life of crossings is a permanent addition to what is known. (Treatise 2.164, 3.26, 12.34)
10. What can be built
Alexandria can be built, and the treatise specifies what would be built; nothing has yet been built to that specification. What it specifies is a record of the collective confirmed structure, held so that any traverser can interrogate it: every INT with its connections, colour, strand and record, and the frontier visible, with its gaps lit and shaped. (Treatise 11.22)
The record is:
- current, because each submission is audited and placed as it arrives;
- iterative, because every arrival runs the cascade through what was already held;
- auditable, because the audit returns the same result to anyone who computes it;
- open, because no position in it is beyond severance;
- language-free, because it holds positions and not the words they were reached in;
- permanent, because what a traverser contributed stands after them. (Treatise 11.22)
It has two layers. The instruments exist whether or not any software is written. The record, with its specification and display, is what is new. It cannot be owned, only governed, and its governance is the definition in section 5 applied to itself: derived from what the architecture requires to satisfy its own integrity condition, not aspirational. (Treatise 11.22 to 11.25, 11.52, 11.54, 12.32)
11. What the theory does not claim
The theory states its own boundary:
- It does not determine ends. Which of the available directions a traverser takes is given, not selected.
- It does not describe sensation and drive of bodily origin.
- It treats embodiment only where it has become confirmed structure.
- Its account of relation between people, built from shared traversal history, is a thin account of kinship and care, and claims no more than that.
(Treatise 1.11)
12. How it could be shown wrong
Theorem II. It fails in three cases:
- a confirmed INT arises through a process in which no integration occurs;
- an integrated crossing produces no lasting consequence;
- the deductive closure is shown to be uncomputable in principle.
The criterion for an artificial system crossing is fixed in advance and will not be moved. A system meeting it would confirm the theorem, not refute it. (Treatise 7.8, 7.10)
The physical derivation. It fails if facts of the second kind are shown to become determinate without a consciousness's integration, or if quantum states are shown to be observer-independent after all. (Treatise 5.22)
Theorem III. It predicts that the crossing signature should appear later and more weakly in heavily constrained systems than in less constrained ones. (Treatise 7.11)
The Free Energy Principle. Alexandria and the Free Energy Principle predict different things when dense adjacent knowledge and a model's expected information gain are varied independently. That experiment can be run with human subjects now. (Treatise 7.11, 8.72)
13. How it is offered
The treatise is itself a submission in its own sense. What it claims to be is a connecting INT between fields that stood apart. No announcement can establish that. The test is whether routes now exist between those fields that did not exist before. (Treatise 12.41)