KEDDEH SYSTEMS | RESEARCH AND FORMAL PUBLISHING | ENGINEER · AUTOMATE · SCALE

# KEX Algebra and Typed State Research Baseline
## Identity, definition, addressing and source-conformant evolution

Document ID: KEDDEH-WP-S02-001  
Editorial edition: Library R1 · 18 October 2026  
Source edition: Editorial E2 · 16 October 2026  
Author: Aboudy Keddeh — Systems Architect  
Publisher: Keddeh Systems  
Rights holder: Aboudy Keddeh  
Classification: Internal research · Owner review  
Status: Editorial research draft; approval and independent assessor not recorded  
Licence profile: KEDDEH-RESERVED-1.0

Architecture and research direction: Aboudy Keddeh. AI-assisted source synthesis, editorial development and document preparation: dot, powered by OpenAI. Authorship identifies contributions; copyright ownership, legal publishing capacity and approval remain separate records. S02 remains an independent research sector within a linked series.

## 1. Abstract

KEX Algebra and Typed State addresses a systems problem: the identity of a subject, its current value, the definition used to interpret that value and the storage location used to retrieve it can change independently. Merging these roles lets relocation resemble creation of a new object, permits a schema change to reinterpret data silently and allows a projected result to erase its constituents.

S02 specifies a typed assessment boundary around subject, state, definitions, addresses and transitions. The E2 native-algebra requirements retain participating identities and transformation history, exclude zero from native calculations and distinguish native coordinate semantics from host indices and compatibility arithmetic. Their assessment must use the author's actual definitions. [KEX, Scope; Architecture; Method]

This baseline belongs to an established series: July papers supply tagged states, conditional projection-loss proofs, frame contracts, provenance closure and a finite word/join algebra. These results apply under their stated domains and do not disappear because E2 is an assessment document. The task is to identify which definitions the particular implementation uses and map later origin and calculation requirements explicitly.

E2 reports an executed assessment of a one-based grid/run codec: 37 supplied unit tests, 13 typed exception cases and five benchmark corpora. This edition preserves those results, the source's conformance finding and the later literal nonnumeric **.1** clarification. It develops an evaluation path without replacing source implementation with an editorial model.

## 2. Scope and definition register

### 2.1 Sector and system roles

The canonical sector is **S02**, in the **KEX & Theorem** division. Its identity and standing come from the BRAINK v74 registry. Relations with other sectors do not merge its methods or proof obligations. KEX defines typed state, identity and addressing; BRAINK governs execution and evidence; IL-LLM carries semantic and knowledge processing and its separately specified research extensions. Host capabilities and external adapters must be identified where exercised. [KEX, Scope]

| Object | Role | Change to retain |
|---|---|---|
| Subject identity | Stable reference to the entity | Explicit creation, replacement or equivalence decision |
| Typed value | Current state under an applicable type | Admitted state transition |
| Definition revision | Meaning and compatibility rules | Versioned semantic/schema change |
| Storage address | Placement used to resolve the representation | Move or resolver update |
| Native coordinate address | Address admitted by the applicable ZCG grammar | Grammar-governed address transition |
| Host index/offset | Position in an array, file or display | Carrier operation, not automatic native meaning |
| Projection | Declared compatibility/observation mapping | Information preserved and discarded |
| Lineage | Participating identities, relations and history | Retention through admitted changes |

A subject can retain identity while changing placement or definition. A zero-valued host index does not thereby authorize zero as a constituent or intermediate value of a native calculation. These are typed boundaries, rather than instructions to relabel every implementation number. [KEX, Scope; Method]

### 2.2 Origin notation across editions

E2's 16 October text says “1 is the only addressable coordinate written as a single digit” and identifies one shared reference singularity across all axes and observables. It rejects independent per-axis origins and any identification with zero-point energy. That wording is retained as a historical source definition. [KEX, Scope; Revision history]

Current author directions, retained by 18 October, designate the shared referent with the exact nonnumeric text **.1**. This edition uses .1 in current explanatory language while retaining old formulas and implementation constants under their dates. It does not coerce .1 into 0, 0.1 or 1 or replace host tuples with that label. [CURRENT; STUDY §2.2]

Axis components, measured values, state-role labels, ranks, indices and locations must remain separately typed. Historical 2, 3, −2 and −3 labels do not become additional native origins simply because they appear in a model. Composite addresses require the applicable native grammar. An editorial tuple or per-axis baseline is not an alternative grammar. [KEX, Architecture]

## 3. Inherited mathematics and applicability

### 3.1 Tagged state and projection

Mathematical Foundations and Proof Obligations v1.0 §§2.1–2.4 gives primitive value, relation, frame, agent, source and event sets, with a disjoint tagged domain:

**KState = Present(V) ⊎ Absent(Reason) ⊎ Cancelled(R, Inputs) ⊎ Boundary(F, Tag) ⊎ PlaceValue(Symbol, Radix, Position) ⊎ ExternalLabel(Value, Unit, Scale, Source) ⊎ Invalid(Code, Cause).**

Its record is **s = (id, constructor, payload, observer, frame, parents, source_roots, version, checksum)**. A valid serialization resolves to exactly one constructor; observer, frame, parents and source roots are explicit. This supplies a defined earlier representation discipline for the fields assessed by S02. Actual runtime adoption and later semantic changes remain implementation questions. [MATH1 §2]

For a discrete random variable X over typed states and a deterministic projection φ, the same paper derives **L_φ(X) = H(X|φ(X)) > 0** when φ is noninjective on X's positive-probability support. H denotes Shannon conditional entropy under the discrete model defined in MATH1 §4.1, not differential entropy. Its seven-step proof identifies at least one positive-probability output with two positive conditional preimages, hence a positive entropy term in the nonnegative weighted sum. Injectivity on the support gives zero conditional entropy. This is a conditional theorem about a specified projection. It does not prove that the full observer transform T or every measurement loses information. [MATH1 §§4.1–4.4]

### 3.2 Algebra, frame and provenance definitions

Native Algebra and Artifact Verification v4.0 §§2–3 gives a disjoint constructor system and reports complete constructor-pair arithmetic surfaces, returning Invalid where numeric operations are inapplicable. Its §3.2 defines a sorted multiset of signed contributions and a canonical additive reduction. Since multiset union and numeric addition are commutative and associative, that declared reduction is independent of input ordering/grouping while retaining a cancellation ledger. The source reports 343 ordered triples over −3 through 3. This is source-reported evaluation of that numeric aggregation model, not a new rerun here. [NATIVE4]

The sorted multiset is an intentional projection. It must not replace ORM's ordered predecessor-dependent relations or a full observation record. ORM §5's long-form rule requires a projection to declare what it retains and discards. The two formulations can be linked at that explicit projection boundary; they are not identical operations. [ORM §§2, 5]

MATH1 §7.1 defines **O_E = (frame_id, origin q₀, basis/orientation B, units U, transform metadata)**. Section 7.3 requires a relation to be invariant, covariant by an explicit rule, or observer-indexed, writing **R_(O′)(gA,gB) = ρ_g(R_O(A,B))**. This supplies a linked frame-transition obligation, not a new native .1 definition.

For a finite acyclic provenance graph, MATH1 §11 defines a derived object's source roots as the union of its parents' roots, with explicit source leaves. Finiteness and acyclicity are premises. Cyclic dependencies need additional declared semantics, and merely retaining a hash does not establish all dependency edges. These are meaningful inherited definitions and conditions for the assessment, not absent assumptions.

### 3.3 Finite word/join algebra and a source worked example

NATIVE4 §6 defines, for hexadecimal nibbles a,b,j in Z16:

**F_j(a,b) = [ab + (a XOR b) + j] mod 16.**

For twelve-nibble words A,B,C in Z16^12, it extends this position-wise and solves the join vector:

**F_J(A,B)_r = [a_r b_r + (a_r XOR b_r) + j_r] mod 16.**  
**J(A,B→C)_r = c_r − [a_r b_r + (a_r XOR b_r)] mod 16.**

The solution is direct subtraction modulo 16 at each position; substitution recovers C. The source's worked triad is A=6761BC62566D, B=0843EB095784 and C=6761BC62566D, with J=00CCC105CB80. Its first three positions are:

- r=0: [6·0 + (6 XOR 0) + 0] mod16 = 6.
- r=1: [7·8 + (7 XOR 8) + 0] mod16 = [56+15] mod16 = 7.
- r=2: [6·4 + (6 XOR 4) + 12] mod16 = [24+2+12] mod16 = 6.

NATIVE4 reports this triad 31 times in the supplied sequence and exact target reproduction for the 100 most frequent triads. Those are retained source reports, not new execution. For fixed j, 256 input pairs map to 16 outputs, so the scalar fold cannot be injective. A deterministic fold is not lossless recovery without additional information. The definition applies to its declared artifact word domain; it is not automatically the codec for every later semantic address or current .1. [NATIVE4 §§6.1–6.3]

## 4. Architecture and implementation contract

### 4.1 Typed transitions

The proposed S02 assessment keeps stable subject, typed value, definition revision and storage address independently visible. Relocation changes placement; definition migration requires an explicit compatibility rule. Neither silently creates a new subject. Native operators retain their source meanings, and adapters must map them before scalar or host execution. [KEX, Architecture]

The reader path is: identify the subject, resolve the carrier, bind retained state to its definition, admit a transition, preserve predecessor and successor, then observe and retain the result. Each experiment must identify its admitted inputs, outputs, transitions, durable objects, rejection behaviour and source version.

### 4.2 Native calculation and compatibility projection

The E2 author-supplied contract excludes zero from native calculations, including intermediate native values, and prohibits zero as a replacement constituent. Participating identities, relations and history remain retained. Balanced or cancelling projections cannot authorize erasure of the contributing records. [KEX, Method]

That contract is later than the July typed Present(0) and finite-nibble models. The assessment must preserve their edition-specific definitions and state whether a given operation is native under the current contract, a retained earlier construction or a declared compatibility projection. It must not silently declare one to be the other. The existence of the earlier proofs is acknowledged; applicability to a changed contract must be established.

| Boundary | Required retained input | Observable result |
|---|---|---|
| Address resolution | Address and applicable grammar/version | Correct resolution or identified rejection |
| Type admission | State, type and definition revision | Admitted value or typed failure |
| Relocation | Prior subject, value and placement | New placement with retained subject/history |
| Definition migration | Old/new definitions and compatibility rule | Compatible successor or recorded incompatibility |
| Native relation | Source operator, domain and ordered constituents | Contract-conformant output/intermediate state or failure |
| Projection | Named mapping and retained source | Projected value and explicit preservation/loss account |

### 4.3 Editorial worked relocation example

Take a hypothetical specimen-A bound to definition D1 and resolved from carrier address A. Moving its bytes to carrier address B should update placement while retaining specimen-A, D1 and its prior record. A subsequent D2 definition requires a declared D1-to-D2 compatibility rule. Merely reading the same bytes at B does not establish compatibility. If the rule is missing or the value is inadmissible, retain an incompatible-definition result rather than silently relabelling it.

This is an editorial illustration of E2's contract, not an executed test or a native address syntax. It shows why identity, meaning and placement must be distinguishable.

The line-origin paper applies a related path: stable line, matrix, BRAINK retained state, IL-LLM traversal and KEX typed evolution. Later S29 membership and P15 quantum-state operations remain distinct implementations; the finite KRS counterpart filter is a third mechanism. Their names do not establish a shared native encoder or a demonstrated KRS-to-P15 adapter. [LINE §13; MECHANISM §§1–4]

## 5. Method and evaluation programme

Before accepting a graph or test as native conformance evidence, identify its actual source model and verify shared referent, address grammar, constituent retention, permitted intermediate values and projection mapping. A graph with independent axis origins, or an oracle that replaces constituents with zero, assesses a different model. Its result remains meaningful under that model's identity. [KEX, Method]

Test address round trips, type preservation and migration on the same frozen object corpus. Count collisions over distinct admitted addresses and migration failures over attempted migrations. Report resolution latency for a declared hit/miss workload. Include invalid addresses and stale definitions as separate rejection cases.

The source's five failure classes are address collision, stale type, incompatible schema, stride/alignment defect and invalid relation. Measure resolution latency, bytes per node, traversal throughput and migration loss, recording hardware/browser/interpreter context, data size, repetitions and uncertainty. Fix acceptance values before qualification; none are invented here.

Performance comparison follows matching of tasks, required outputs and retained context. A compact result that omits required lineage has changed the task. More retained structure does not establish an advantage either; compare resource cost under equivalent requirements.

E2 names the required documentation outputs as KEX type system, address grammar, algebra reference and benchmark protocol. The linked definitions above are part of that route. The executable package still needs the exact versions it imports, full failure traces and independent assessment. A mathematical definition, implementation conformance and deployment qualification remain distinct milestones.

## 6. Results and retained evidence

### 6.1 Archived grid/run codec, reported 16 October 2026

E2 records **37 supplied unit tests** and **13 typed exception-replay cases** passing on the recovered core-engine archive. An earlier exception-boundary archive passed **14 unit tests**, overlapping the later suite; they are not 14 additional independent properties. This editorial edition did not rerun those archives. [KEX, Results]

The inspected mapping.py reserves **ToroidalCoordinate(1, 1, 1)**, maps run ordinals using ordinary row-major arithmetic and exposes x/y/z integer components. E2 identifies this as a one-based grid adapter and packed run codec, rather than the native ZCG address grammar supplied by the author in that review. The tuple remains exact historical implementation evidence, neither rewritten to .1 nor used to infer a native encoder.

### 6.2 Five-corpus benchmark

Each corpus contains **4,096 bytes**. E2 supplies output sizes in order without individual corpus identities in the manuscript:

| Corpus order | Input bytes | Packed output | Conventional run-length baseline |
|---|---:|---:|---:|
| 1 | 4,096 | 15 | 3 |
| 2 | 4,096 | 8,206 | 8,192 |
| 3 | 4,096 | 8,206 | 8,192 |
| 4 | 4,096 | 8,168 | 8,154 |
| 5 | 4,096 | 7,890 | 7,876 |

These are measured archive results as reported by E2. They do not establish general compression advantage or evaluate an alternative native model absent from that archive. Hardware, timing, uncertainty and corpus identities must come from the associated execution package before a new performance qualification is made; this edition does not invent them.

### 6.3 E2's applicability finding and July results

E2's dated review identifies MATH1's per-record existence indicator (§6.3), Present(0) lifted numeric model (§§5–6), phase-tagged −1/+1 mapping (§8) and finite 0…15 fold (§9) as distinct constructions. Its conclusion is that those results do not automatically establish the shared-addressable-origin or zero-free native calculation contract supplied in the October review. [KEX, Results]

The earlier derivations remain valid claims within their stated models and conditions. For example, MATH1 distinguishes an instantiated record's existence value from its magnitude, while its phase-tagged coordinate inverse concerns the solid/liquid seam. Neither phase meaning is automatically complex quantum phase. The conformance question is which model and mapping the current implementation uses, not whether the earlier papers contain definitions or proofs at all.

### 6.4 Registry standing

The source-reported standing is **PACKAGE BUILT · COMPILE/REPLAY PASS · CORRECTION ROUTE PASS · EVOLUTION PASS**. It is a retained registry status, not new S02 replication. The v74 report's controlled local recovery and persistence-gate tests do not automatically qualify another deployment or prove scientific propositions. The proposed assessment contract adds method and evidence requirements; it is not itself a new sector experiment. [V74; KEX, Results]

## 7. Discussion and next result

S02's assessment architecture distinguishes software passing its own tests from software conforming to the intended source definition. That distinction preserves useful engineering evidence and the actual mathematical work already present in the linked series.

Current .1 remains a nonnumeric referent label; old one-based tuples remain carrier values; July type, projection and word-algebra definitions retain their conditions. A source-linked experiment can then ask precisely which operation crossed which boundary and what evidence connects its output to the intended definition.

Performance, security, physical and market claims need their relevant observables. A logical invariant, recovered record and deployed hardware guarantee require different evidence. The next comparison should bind its inherited definitions, source implementation, admitted corpus and required retained information before testing. New operators or bridges should be added through explicit definitions and results rather than a generic substitute algebra.

## 8. Conclusion and revision history

KEX S02 makes subject identity, typed value, definition and addressing independently inspectable. Its linked series provides actual formal objects and conditional theorems; its assessment route identifies the mapping needed for a concrete runtime and the current author contract. The archive evidence remains concrete and bounded.

- Version 1.0 established the sector identity, baseline and evaluation programme; assessor and approval were not recorded.
- 15 October 2026, E1 preserved the original in a separate copy and expanded mechanisms, denominators and inference limits; it asserted no new execution or independent approval.
- 16 October 2026, E2 applied the shared-reference clarification written then as 1, distinguished native addressing from projections/indices, and rejected independent axis origins and zero-point-energy interpretation.
- 18 October 2026, Library R1 retains E2's complete results, dates the later literal .1 clarification, incorporates scoped cross-paper definitions and a source worked join example, and adds an editorial relocation example. No original document or code was changed.

### References and source edition

- [KEX] KEX Algebra and Typed State Research Baseline, S02, KEDDEH-WP-S02-001, E2, 16 October 2026. [Source](https://docs.google.com/document/d/1WnuzNoWlUHthjWp27UpqEryS_nJcpFIg3eDfQ4b4Mx0/edit). Original reviewed capture: 10,805 bytes; SHA-256 f8c05a194e8935d42db972c1ab3cfdad1b639ff6fcba277fa125bde430d6413e. A separately hashed paragraph recapture was retained after workspace recovery; the Drive title still says E1 while the body identifies E2.
- [KEX-E1] [Earlier revision identified by E2](https://docs.google.com/document/d/1lkz4QnrrGS8S1AmF3Q5mgdeQVNnmnkCctD0xXo9UyHs/edit?usp=drivesdk).
- [CURRENT] Aboudy Keddeh. Current directions retained in Observation and Identity Review R3; current at 18 October 2026.
- [MATH1] Mathematical Foundations and Proof Obligations, v1.0, 13 July 2026, §§2, 4–9, 11. [Source](https://docs.google.com/document/d/1x3m7DwJbqSFMK4skvzf0pWUXauM-R_gX/edit).
- [NATIVE4] Native Algebra and Artifact Verification, v4.0, 14 July 2026, §§2–6. [Source](https://docs.google.com/document/d/1sz_cntJj0g1eqbSBrEGnC9gjjR1Jqp9p/edit).
- [ORM] Observer Relational Mathematics, R1, 10 October 2026, §§2, 5. [Source](https://docs.google.com/document/d/1ibTDjjWLeDcKFclPKWClLs1dElMsjKDPfXal9UmJWhc/edit).
- [LINE] KEDDEH Recursive Line-Origin Matrix Formalism, R3, 17 August 2026, §13. [Source](https://drive.google.com/file/d/1XCUFOmHxYHYtDiNrdO9vTZGQnU11toys/view).
- [MECHANISM] Relational and quantum software: inspected mechanisms and evidence, 18 October 2026. Retained inspection report.
- [STUDY] From Observation to Governed Computation, R1, 18 October 2026; retained study and source register.
- [V74] KEDDEH-SRC-V74. BRAINK v74 Engineering Research and Stress Report, sector registry and claim boundary. Bibliographic identity retained from E2; no new deployment status inferred.
- [DID] W3C. [Decentralized Identifiers v1.0](https://www.w3.org/TR/did-core/), identity comparator retained from E2. KEX addresses are not automatically DIDs.
- [PUBLISHING] Keddeh Research and Formal Publishing Standard, KEDDEH-STD-PUBLISHING-001, v1.0.

© 2026 Aboudy Keddeh | Keddeh Systems | All rights reserved  
KEDDEH-WP-S02-001 | Library R1 | Internal research · Owner review | Draft  
KEDDEH-RESERVED-1.0. No licence is granted. Separate written permission governs reuse, redistribution and implementation, subject to applicable law. Third-party material retains its rights. No copyright assignment, patent or trademark licence, or independent approval is asserted.