CHANGELOG.md

Changelog

Unreleased

• audits the first-principles picture after notes 42--63: the horizontal distinguishability and vertical objectification axes form a clear theory spine, while transported history payloads, task resources, holonomy, Bellman/coarea, and partition semantics remain an emerging T0/T1 transversal rather than a new Core node;

• moves ConstraintCanonicalization, ObserverConnection, and CanonicalDecomposition to canonical ownership under process_geometry.experimental, removes their modules from the declared Presentation/Analysis namespace surfaces, and retains the historical paths as identity-preserving 0.0.x compatibility shims;

• makes the exact finite task-quotient certificate read-only and adds direct quotient execution/observation helpers without broadening its finite, deterministic theory contract;

• updates package license metadata to the SPDX Unlicense expression while retaining LICENSE in distributions, removing the setuptools metadata deprecation found during the audit build;

• opens the (S^6) complex-structure Sonnet as a two-axis research contract: first use the proposed torus-fibred threefold as a lossless external calibration of lifted history, task-visible holonomy, non-normal canonicalization, and singular completion; independently test whether the complex Addition/Multiplication/Power closure admits a task-sufficient rank-four integral shadow; source reproduction, strict equivariance, oracle firewalls, stage gates, and negative-result conditions are frozen before implementation, with no Theory Map or API promotion;

• adds a research-local thermodynamic-objectification record and two exact finite executable essays: same-scale weighted pushforward/flattening, different-scale power sums, unit-cell carry arithmetic, the partition/coarea cumulant bridge, plethystic free assembly, and a Gamma(2)-twisted finite trace/determinant red team; the V2/V3 boundary is refined without a new API or theory-maturity promotion;

• freezes PCR3BP Phase 2 as a return--partition--holonomy experiment: reconcile the Phase-0 and Phase-1 gate presentations first, then build a converged sub-Markov return/absorbing ensemble and compare scalar with task-visible twisted partition data; Bellman/Huffman and continuous zeta claims remain explicitly downstream;

• standardizes the repository on the canonical Huffman spelling across research notes, filenames, workflows, internal Python identifiers, and opt-in environment variables; numerical results and research claims are unchanged, and a repository-wide hygiene test prevents regression;

• opens the stochastic feedback-trap first-passage Sonnet with a frozen dimension/task/oracle contract; the first executable gate proves exact Ito generator covariance for w=u+u^3, exposes the nonzero residual left by a naive deterministic chain rule, and recovers that rule at zero noise;

• executes the depth-two A/M presentation census for the stochastic Sonnet, retaining literal syntax before exact polynomial quotienting and certifying strict orientation on the full stopping interval rather than by samples; the exact 604 -> 60 -> 16 result has only affine survivors and gives a bounded negative certificate for nonlinear presentation discovery;

• freezes the depth-three stochastic-presentation enlargement before execution: 365,424 literal trees are counted combinatorially while a complete semantic closure over the 60 depth-two values avoids redundant materialization;

• executes that closure: 1,519 semantic polynomials leave 242 exact monotone presentations and 155 nonlinear survivors; u+u^3 is included but not unique, so grammar coverage succeeds while canonical selection fails;

• executes the full stopped-process quotient over all 242 monotone charts: exact Itô generator transport, labelled sections, initial point, and clock produce one task class, while the no-Itô-correction red team splits exactly into 87 affine passes and 155 nonlinear failures;

• freezes and executes an independent stochastic first-passage numerical gate: transported-node backward BVPs in one affine and two nonlinear charts converge to a common value near 0.572194, direct Euler-Maruyama estimates agree within sampling/time-step error, and uniform target-coordinate meshes expose the predicted finite-resolution coordinate artifact;

• closes the stochastic calibration with a dimensionful regenerative Bellman task: three charts independently select left-reset, optimal values agree within 5e-6 and scale as L/V, while swapped labels and coordinate-distance charging expose the required task and resource failures;

• calibrates the S2-prime moving-observer contract on a known Riccati control: differentiated normalization induces observer motion, shape plus transport reconstructs exactly, and an out-of-grammar cubic survives as residual payload; this remains research-local and is not AM-native discovery;

• executes the first blind S2-prime gate: bounded A/M expression histories derive coefficient jets compositionally, and an exact frozen normalization grammar uniquely recovers the oriented moving Riccati observer while static candidates fail and cubic completion payload survives;

• freezes three normalization families and runs a held-out selector: affine and asymmetric-quadratic controls select exact moving observers, a cubic carrier is rejected without truncation, and a centered quadratic exposes two equal-cost canonical lifts, proving flatness alone does not ensure canonicity;

• quotients the centered-quadratic minimum slice by its exact clock-preserving affine chart morphism: both lifts carry the same physical stopping sections and become one task-relative canonical class, while a coordinate-squared penalty is exposed as a non-invariant false selector;

• blindly discovers that task-preserving morphism from a bounded A/M grammar using section, reconstruction, and dynamics certificates together; a corrupted-dynamics control fails, and the quotient remains unchanged when the observer grammar grows from depth one to depth two;

• restores dimensions to the moving centered process and executes the complete clock/Bellman covariance check in u and nonlinear w=u+u^3: equal-clock values and policies agree, value scales as L/V, while equal-coordinate grids change both optimized value and tree;

• archives the calibration's physical realization as an overdamped particle in a feedback-synthesized translating nonlinear potential, distinguishes local saddle-node relevance from global exactness and equivariance from Noether conservation, and closes the affine deterministic phase before a separate stochastic first-passage continuation;

• records PR #81's pendulum/cross-problem canonical-history bridge and the separation of task quotienting from costed Huffman/Bellman optimization;

• records PR #82's AM universal-history correction, dimensional/covariant Bellman and Fermat/Noether calibrations, and magnetic/Berry history payloads;

• records PR #83's separation of the classical AD control from the exact supported A/M process-frame symmetry slice and its AMJet/Noether boundary;

• executes the predicted equal-clock versus equal-coordinate pendulum section red team: (X=U^3) preserves clock-measured stopping policy and Bellman value, while coordinate spacing changes both;

• adds a recent-research index for docs 53--61; docs 55 retains its explicit correction banner pointing to docs 56;

• adds docs/55-pendulum-lifted-clock-global-quotient.md and a research executable essay globalizing the lifted-clock/geometric-phase chain of docs/54 on the lemniscatic leaf;

• certifies the exact sigma symmetry sigma(U,Y)=(-U,iY) with sigma*omega = i omega (hence tau = i), the complex-flow carrier ODE dU/dz = Y, and — from the invoked Jacobi half-period relations with K(i) = varpi/2 — the primitive square period lattice Lambda = Z omega_A + Z i omega_A with omega_A = sqrt(2) varpi;

• certifies the decoder-boundary analysis: the local chart becomes 0/0 at U = +/-1, but on the real E = 0 loop the reached point U = -1 has two distinct Cartesian lifts v_x = +/- sqrt(2); the signed Jacobi continuation q_x = cn dn proves sheet transport +1 -> -1 -> +1, so the physical cover is the nontrivial unramified double cover of the real carrier loop and closes after 2 omega_A;

• reads off the clock-chain information contract: R -> R/omega_A Z forgets winding, and the real phase embeds in C/Lambda because R cap Lambda = omega_A Z — the third arrow is a complexification, not a further quotient; corrects the merged P10 period naming (T_p = 2 omega_A is the physical pendulum period, not the primitive carrier period);

• keeps the result research-local: no Theory Map node/edge change, no public or experimental API pressure, and no promotion of TR-0001; the E = 0 leaf is the declared scope, other energy leaves remain open;

• adds docs/54-pendulum-elliptic-group-rank-lowering.md and a research executable essay providing a second, geometric (non-arithmetic) calibration of V3–V4: the elliptic group law of the pendulum observable carrier as compositional rank lowering of the flow-translation schema;

• certifies the exact Weierstrass reduction of the carrier cubic (lemniscatic leaf j = 1728), the chord-tangent group law with exact rational associativity/inverse spot checks, and Euler's addition theorem exactly as an invariant-differential identity modulo the curve ideal in both chord and tangent form, yielding Abel-Jacobi additivity modulo periods;

• realizes the E = 0 flow in closed form U(t) = -sn^2(t/sqrt(2), i) and numerically samples the 2-torsion base-point correction P(t1) (+) P(t2) = P(t1+t2) (+) P(0) and the expected corrected one-parameter subgroup law;

• separates lifted translations carrying t in R from geometric actions carrying [t] in R / T_p Z, with the period group as the kernel of projection;

• records three red teams parallel to the AEG ones: an unmarked endpoint merges distinct Cartesian continuations, a fixed curve point does not identify the flow schema (period ambiguity and torsion twist), and coordinatewise addition leaves the carrier;

• keeps the result research-local: no Theory Map node/edge change, no public or experimental API pressure, and no promotion of TR-0001; the group layer of the marked-carrier chain becomes executable for one carrier without claiming canonicity;

• adds docs/51-aeg-addition-multiplication-rank-transition.md and a research executable essay for the second arithmetic rank transition: Multiplication objectifies a uniform repeated-Addition endomorphism on Translation objects rather than one fixed additive composite;

• lowers pure multiplicative words to Translation endomorphisms and arbitrary mixed Translation/Dilation words to the positive affine monoid, with exact relation soundness for T_a T_b = T_(a+b), D_k D_l = D_(kl), and the AEG cross relation D_k T_a = T_(ka) D_k;

• records two second-rank red teams: one fixed additive result cannot identify the multiplicative process schema, and a nontrivial dilation cannot lower to any single Translation object; connects the finite cross relation to the existing A/M finite law and [A, M] = A without claiming V5 analytic closure;

• adds docs/50-aeg-translation-objectification-rank-lowering.md and a research executable essay providing the first complete concrete V1→V4 calibration: continuation-stable semantic compression, objectification into reusable Addition/translation primitives, free higher-rank composition, compositional lowering, and relation soundness;

• adds an adversarial absolute-displacement compression showing that terminal equality is insufficient when continuation can recover a discarded distinction;

• refines docs/THEORY_MAP.md so V2–V4 now record one complete AEG/Addition research calibration while generic objectification, rank, and lowering abstractions remain intentionally unpromoted;

• adds docs/49-theory-implementation-structural-alignment.md, mapping H0–H4 and V0–V5 in the living Theory Map to current executable implementations, calibrated shadows, and missing semantics;

• adds an explicit process_geometry.experimental incubation namespace without changing the stable root API;

• adds FiniteTaskQuotient / minimize_finite_task_process, an exact finite deterministic H1/V1 calibration using stable partition refinement and future distinguishing continuations, deliberately narrower than a generic Process Geometry task quotient;

• refines docs/THEORY_MAP.md so H1/V1 distinguish bounded public task-continuation evidence from the exact finite Experimental quotient slice;

• aligns current program vocabulary with the Process Geometry foundation in docs/42–45 and records the mapping in docs/48-foundation-naming-audit.md;

• makes TaskContinuationSignature, task_continuation_signature, and history_task_continuation_signature the canonical names for the bounded future-task signature, retaining the historical ProcessJet* names as 0.0.x aliases so jet remains available for genuinely differential/local structure;

• exposes qualified discovery names ObservableAlgebraicQuotient, discover_first_order_observable_quotient, search_first_order_observer_presentations, and structural_first_order_observer_presentation_cost, while retaining historical backend aliases;

• reserves strong theory words such as task/process quotient, objectification, process rank, rank lowering, and analytic closure until their full semantics are implemented rather than promoting new classes by nomenclature alone;

• further migration of repository-owned historical tests/docs from aeg_shakespeare remains follow-up work; no new implementation may depend on the deprecated namespace.

0.0.4 — canonical process_geometry namespace

Second identity-migration release of Process Geometry.

Highlights:

• makes process_geometry the canonical Python import namespace;
• moves the complete implementation tree under src/process_geometry;
• reduces src/aeg_shakespeare to a temporary compatibility alias rather than keeping a second implementation copy;
• preserves deep Python object identity between representative legacy and canonical imports;
• changes public quickstarts, wheel smoke tests, and publish smoke tests to use process_geometry as the primary namespace;
• retains aeg_shakespeare only as a deprecated compatibility path for code written against 0.0.1–0.0.3;
• establishes the dependency direction process_geometry <- aeg_shakespeare alias and adds release/hygiene requirements preventing canonical implementation code from depending on the historical namespace;
• leaves the semantic public pipeline Process -> Presentation -> Discovery -> Analysis unchanged and promotes no new Process Geometry research abstractions solely because of the namespace migration.

Historical PyPI distributions named aeg-shakespeare should not be installed alongside current process-geometry releases because they ship their own aeg_shakespeare package and may shadow the compatibility alias.

0.0.x APIs remain experimental and may change without compatibility guarantees.

0.0.3 — Process Geometry release identity

First pre-alpha release under the Process Geometry distribution identity.

Highlights:

• renames the PyPI distribution from aeg-shakespeare to process-geometry while preserving the existing aeg_shakespeare import namespace for this transition release;
• moves the repository identity to mountain/process-geometry and updates package, citation, CI, and publish metadata accordingly;
• retains the semantic public pipeline Process -> Presentation -> Discovery -> Analysis;
• adds the minimal public presentation.morphism.PresentationMorphism record after independent KdV, resistor-network, and braid/Markov calibrations;
• keeps morphism validity explicitly task-relative through caller-defined semantics and certificates, allows heterogeneous source/target presentation types, and deliberately postpones universal verification, composition, inverses, normal forms, and category/groupoid structure;
• adds the first-principles Process Geometry foundation: exact distinguishability quotients, the Myhill–Nerode minimal-presentation calibration, and the topological threshold;
• adds the vertical objectification program: semantic compression, new primitives, free higher-rank composition, and compositional rank lowering;
• records the lineage against algebraic theories, operads, Baez–Dolan slicing, polygraphs/computads, definitional extension, abstract interpretation, and sheaf-style locality;
• distinguishes semantic, topological, and analytic cross-rank closure, with AEG retained as the first model organism rather than as a package-wide arithmetic restriction;
• keeps Shakespeare/Sonnet as the problem-driven research program rather than the software distribution identity.

Historical releases 0.0.1 and 0.0.2 were published under the PyPI distribution name aeg-shakespeare. The new process-geometry distribution should not be installed side-by-side with the historical distribution because both currently provide the same transitional aeg_shakespeare import package.

0.0.x APIs remain experimental and may change without compatibility guarantees.

0.0.2 — semantic API and discovery expansion

Second pre-alpha research release of AEG Shakespeare.

Highlights:

• reorganizes the public API into the semantic pipeline Process -> Presentation -> Discovery -> Analysis;
• contracts the package root to namespace navigation while retaining lazy 0.0.x compatibility shims;
• physically consolidates finite process families, characters, actions, and process cocycles under process.finite;
• moves literal histories and local process realizations to canonical process.history / process.local ownership and moves search budgets to presentation;
• adds bounded polynomial invariant and observable-quotient discovery;
• adds structured pairing-based observer proposals and Pareto selection of first-order algebraic quotients;
• adds explicit coefficient-language extension for discovered process relations, with oscillator red-team tests showing that finer splitting is not universally cheaper;
• adds finite ProcessFamily, ProcessCharacter, FamilyAction, and ProcessCocycle calibrations through translation, dilation, A/M, Galilean mechanics, and magnetic translations;
• extends the global Abelian-history line through lifted cycles, period matrices, intersection structure, and normalized Abel–Jacobi history quotients;
• adds repository-level AST import-hygiene and physical-ownership checks so compatibility modules cannot silently become architectural dependencies;
• supports and tests CPython 3.10, 3.11, 3.12, 3.13, and 3.14.

0.0.x APIs remain experimental and may change without compatibility guarantees.

0.0.1 — research preview

First installable pre-alpha release of AEG Shakespeare.

Highlights:

• ordered ProcessWord histories and explicit noncommutative rewriting;
• bounded task-sufficient process signatures;
• history depth/boundary observables and Huffman prefix strategy;
• construction-history-preserving primitive proposals;
• generated finite process grammars and exact return-relation decomposition;
• algebraic constraint quotients and prolongation;
• multi-axis presentation cost and Pareto search;
• generic ProcessFrame and ProcessFunctionModule abstractions;
• Addition/Multiplication (A/M) function-theory prototype with resonance and path-flow structure;
• classical/research calibration program including constrained pendulum and genus-hierarchy probes;
• public-domain dedication and literate-programming/citation discipline.

0.0.x APIs are experimental and may change without compatibility guarantees.