Experiment atlas

Follow the experiment
all the way to the code.

The books necessarily compress implementation detail. This atlas preserves the missing trail: chapter, experiment family, code package, executable entry point, frozen result artifact, and claim boundary.

38 experiment families79 displayed Experiment cards77 curated studies229 resolved run records625 resolved source files15 public code packages148 verified starting files55 pinned GitHub file links38 traceability-complete families

Traceability is not replication

This registry links the experiments reported by the three books to their executable or archived code surfaces. A code link establishes traceability, not independent reproduction, correctness, or support for a claim beyond the experiment's stated boundary.

Download the experiment-to-code crosswalk ↓ · Download the machine-readable catalog ↓ · Download companion release specifications ↓ · Download all family documentation ↓ · Download the card-to-code crosswalk ↓ · Download the documentation closure register ↓ · Download the machine-readable closure register ↓ · Download the entry-point audit ↓ · Download the live-link rehearsal ↓ · Download the software catalog ↓ · Download the machine-readable software catalog ↓

The release audit resolves all 148 starting files: 55 immutable public GitHub destinations and 93 byte-present files in curated, nonpublic release packets. It reports 0 unresolved mappings. Public destinations are pinned to immutable repository commits. The manifest-facing audit remains deterministic and offline. A separate bounded live-link rehearsal reached 55 of 55 public starting files; network availability is recorded without being treated as scientific reproduction.

The card-level audit resolves all 79 highlighted Experiment cards: 34 have at least one public code destination, while 45 retain an explicit private-code or proposed-protocol boundary. Overrides resolve experiment cards whose chapter contains several experiment families, whose implementation lives in a deliberately unreleased workspace, or whose card specifies a proposed benchmark rather than a completed run. An association is provenance, not evidence that the displayed protocol was executed by every linked package.

Documentation coverage

What is traceable today—and at what depth.

All 38 registered families have a durable record and at least one machine-resolved code surface. Only 12currently have fully public core code; 26depend on one of the five curated local companions. Public code and frozen run packets are reported separately because they establish different levels of access and evidence.

Book 01

Categories for AGI

Documented families
11 / 11
Public core implementations
4 / 11
Families with any public package
9 / 11
Frozen-run families
0 / 11
Resolved code surfaces
133
Curated studies
42
Frozen run records
0

Code and result provenance are documented; run-level public packets remain a release task.

Book 02

Machine Learning from Enforcing Compositionality

Documented families
8 / 8
Public core implementations
8 / 8
Families with any public package
8 / 8
Frozen-run families
0 / 8
Resolved code surfaces
134
Curated studies
35
Frozen run records
0

Code and result provenance are documented; run-level public packets remain a release task.

Book 03

Infinitesimal Creativity

Documented families
19 / 19
Public core implementations
0 / 19
Families with any public package
7 / 19
Frozen-run families
19 / 19
Resolved code surfaces
358
Curated studies
0
Frozen run records
229

Book 04

Discovering Compositional Worlds from Interaction

Documented families
0 / 0
Public core implementations
0 / 0
Families with any public package
0 / 0
Frozen-run families
0 / 0
Resolved code surfaces
0
Curated studies
0
Frozen run records
0

Code and result provenance are documented; run-level public packets remain a release task.

Public code shelf

The inspectable GitHub packages in one place.

Each destination is pinned to the revision audited for this edition. The family links return to the corresponding experimental contract, evidence record, and chapter locations.

multi-package software companion

LINCS software companion ↗

Public companion containing formal, BRIDGE/SKFM, ALLORA, LASKO, GIRL, LINCS-RLHF, RADAR, and SID packages.

Edition revision
c78bfe50ded8d40db90c614e40dd2e0e146a9875
License
Apache-2.0
Used by
Machine Learning from Enforcing Compositionality

Code companion release queue

The missing repositories are now bounded.

A companion specification identifies the smallest audited source surface that supports navigation from a book experiment to code. It is not a claim that the package reproduces every reported number, that dependencies are frozen, or that the underlying data and media may be redistributed. Across the five local archives, the audited book-facing surface contains 352 files rather than the full research workspaces.

private-clean-environment-verified

Categories for AGI learning experiments

Registry ID
categories-gt-archive
Disposition
split-release
Audited surface
35 files · 476,306 bytes
Experiment families
4
Result packet
15 claimed artifacts
License
Decision required
Canonical result provenance

Categories for AGI, Chapters 6--8 and 10--11: the executable DB energy illustration, two archived commutator-proxy figures, and the strict-causal, future-carrier, and archived block-4 language-model tables

Artifacts
15 archived CSV/JSON files
Packet size
157,873 bytes
Verification
python verify_chapters_6_8.py && python verify_book_tables.py

Evidence boundary. The packet executes the Chapter 6 numerical notebook cells and verifies the exact archived Chapter 8 figure bytes, dimensions, producing-script hashes, and manuscript captions. The figures have no surviving per-step raw logs and are therefore plot-only evidence, not independently reproduced runs. Separately, the packet verifies the rounded Chapter 10 and 11 table cells against twelve archived CSV/JSON artifacts and preserves the exact run configurations. It does not rerun training; the Chapter 10 values are single runs, the Chapter 11 PTB rows are five-seed aggregates, and the remaining Chapter 11 slices are unmatched single-run snapshots.

Claimed artifacts

  • Category-Theory-for-AGI-UMass-CMPSCI-692CT/notebooks/week03_db_colimit_energy.ipynb
  • NeurIPS 2026 Arxiv Papers/Category Theory for AGI Revised/two_moons_commutator_energy.png
  • NeurIPS 2026 Arxiv Papers/Category Theory for AGI Revised/wiki103_prefix_suffix_cech_obs.png
  • outputs/causal_ablation/lm_causal_comparison_ptb_L2_D256_20260305_010640/summary.csv
  • outputs/causal_ablation/lm_causal_comparison_ptb_L2_D256_20260305_010640/summary.json
  • outputs/causal_ablation/lm_causal_comparison_wikitext2_L2_D256_20260305_011350/summary.csv
  • outputs/causal_ablation/lm_causal_comparison_wikitext2_L2_D256_20260305_011350/summary.json
  • outputs/causal_ablation/lm_causal_comparison_wikitext103_L2_D256_20260305_020457/summary.csv
  • outputs/causal_ablation/lm_causal_comparison_wikitext103_L2_D256_20260305_020457/summary.json
  • KET/ket_experiments/results/logs/lm_unified_ptb_L16_D64_20260425_102233/aggregate.csv
  • KET/ket_experiments/results/logs/lm_unified_ptb_L16_D64_20260425_102233/aggregate.json
  • KET/ket_experiments/results/logs/lm_unified_wikitext2_L2_D256_20260307_184700/aggregate.csv
  • KET/ket_experiments/results/logs/lm_unified_wikitext2_L2_D256_20260307_184700/aggregate.json
  • KET/ket_experiments/results/logs/lm_unified_wikitext103_L2_D256_20260307_192034/aggregate.csv
  • KET/ket_experiments/results/logs/lm_unified_wikitext103_L2_D256_20260307_192034/aggregate.json
Release boundary and outstanding work

Phase one includes

  • The audited implementation, runner, and dependency files linked from Chapters 6–14
  • An executable Chapter 6 DB energy notebook and two exact archived Chapter 8 commutator-proxy figures with explicit plot-only provenance
  • Twelve bounded CSV/JSON artifacts that support the revised Chapter 10 and 11 language-model tables
  • Deterministic verifiers for the Chapter 6 notebook, Chapter 8 artifact provenance, and 87 Chapter 10–11 run-metadata and rounded table-cell assertions
  • A dependency specification and one CPU smoke test per experiment family
  • The public RADAR code and result records supporting the Chapter 12 manifold studies

Excluded pending review

  • The 651 GB parent research workspace
  • Model checkpoints, downloaded corpora, caches, build products, and unrelated projects
  • Result directories not tied to a named book claim

Outstanding

  • Choose the companion repository boundary and license
  • Recover per-step raw logs for the Chapter 8 plots if they still exist; otherwise retain their explicit archived plot-only status
  • Preserve Chapters 13–14 as archived-only evidence unless complete raw run artifacts are found

Diagrammatic Backpropagation and Geometric Transformers · GT information regimes, KET, and TopoCoend · Mean-field diagnostics for Geometric Transformers · Depth sweeps for Geometric Transformers

private-clean-artifact-verified

Judo Calculus supplementary experiments

Registry ID
kan-do-dcdi-archive
Disposition
merge-or-snapshot
Audited surface
21 files · 73,720 bytes
Experiment families
1
Result packet
7 claimed artifacts
License
Decision required
Canonical result provenance

Categories for AGI, Judo Calculus chapter: four directed-recovery rows and three Radon--Nikodym calibration rows

Artifacts
7 archived CSV/JSON files
Packet size
825 bytes
Verification
python verify_book_tables.py

Evidence boundary. The packet verifies the chapter's rounded table cells against seven archived CSV artifacts. It does not independently reproduce the Sachs, LINCS, or PISA runs; those require dataset provenance review, and the archived PISA calibration runner has a formatting defect that must be corrected before a clean rerun.

Claimed artifacts

  • runs/synth10_kando_metrics.csv
  • runs/j_sheaf_fast/results_metrics.csv
  • runs/s9_kando_metrics.csv
  • runs/s9_kando_multi_metrics.csv
  • calibration_synth.csv
  • runs/lincs_do_hspa8/lincs_rn_calibration_HSPA8.csv
  • runs/pisa2022_do_hisei_escs/pisa_rn_calibration_hisei_escs.csv
Release boundary and outstanding work

Phase one includes

  • The audited Python runners, core operators, and transitive local helper modules linked from Chapter 20
  • Seven bounded CSV artifacts that exactly support the chapter's four directed-recovery rows and three calibration rows
  • A deterministic table-cell verifier that retains the weak Sachs results as negative evidence
  • A map to the overlapping public Causal Discovery on Sheaves and LINCS BRIDGE/SKFM code

Excluded pending review

  • Virtual environments, Python caches, duplicate ZIP bundles, and generated figures
  • Large run directories that are not needed by a named table or figure
  • Third-party datasets without an explicit redistribution record

Outstanding

  • Decide whether the packet belongs in Causal Discovery on Sheaves or a dated snapshot repository
  • Choose a license compatible with the public companion
  • Document redistribution rights and provenance for the Sachs, LINCS, and PISA inputs
  • Repair the PISA runner's diagnostic formatting defect and repeat the retained experiments in a clean environment

Judo Calculus causal-discovery experiments

private-clean-environment-verified

ATLAS/CSQL worked examples

Registry ID
atlas-csql-archive
Disposition
curate-separately
Audited surface
3 files · 19,008 bytes
Experiment families
1
Result packet
Not yet attached
License
Decision required
Release boundary and outstanding work

Phase one includes

  • The two audited SQL query files and one starter notebook linked from Chapter 21
  • A generated, rights-clear synthetic categorical database instance
  • Schema, query, expected-result documentation, and a deterministic DuckDB smoke test

Excluded pending review

  • Article text or other source material not licensed for redistribution
  • Derived database instances whose provenance has not been documented
  • Notebook outputs that contain embedded source text or credentials

Outstanding

  • Choose a license for the public companion
  • Decide whether the corpus-specific notebook belongs in the public packet and pin its separate environment

CSQL and ATLAS database construction

private-clean-environment-verified

Deep URL reproducibility packet

Registry ID
deep-url-archive
Disposition
curate-separately
Audited surface
5 files · 35,994 bytes
Experiment families
1
Result packet
30 runs · 3 reports
License
Decision required
Canonical result provenance

Categories for AGI, Deep URL chapter: the thirty-run synthetic line-world ablation table

Run family
paper_short/GT_RL_Coalgebra/outputs/ablation/20260223_105353/runs
Model grid
gt · gtdb · mlp
Seeds
0, 1, 2, 3, 4, 5, 6, 7, 8, 9
Packet size
33 files · 19,503,739 bytes

Evidence boundary. Result packets contain only artifacts needed to audit a named book claim. Inclusion establishes provenance and re-aggregation, not independent replication.

Training command

python -m GT_RL_Coalgebra.run_synthetic_mdp_ablation --episodes 300 --max-steps 36 --seeds 10 --seed-start 0 --models gtdb,gt,mlp --n-states 32 --random-start --slip-prob 0.10 --reward-noise-std 0.05 --db-coef 0.1 --device cpu

Report command

python -m GT_RL_Coalgebra.report_synthetic_mdp_ablation --inputs-glob 'results/canonical_30_run/raw/*.csv' --out-dir reproduced_report --success-window 20 --success-target 0.8 --bootstrap-samples 2000
Release boundary and outstanding work

Phase one includes

  • The five audited Python bridge, runner, and reporting files linked from Chapter 28
  • The canonical thirty raw runs and three report artifacts referenced by the chapter
  • A deterministic CPU smoke configuration, result manifest, and byte-for-byte report reproduction check

Excluded pending review

  • Python caches, duplicate manuscript sources, and generated plots reproducible from CSV
  • Opaque binary artifacts without a documented reader
  • Exploratory outputs not used by the chapter

Outstanding

  • Choose a license for the public companion

Deep Universal Reinforcement Learning

private-cpu-smoke-and-result-packet-verified

Infinitesimal Creativity code companion

Registry ID
synthetic-creativity-archive
Disposition
dedicated-repository
Audited surface
288 files · 3,723,688 bytes
Experiment families
19
Result packet
629 bounded records · 229 runs
License
Decision required
Canonical result provenance

Infinitesimal Creativity, Chapters 10--19: the registered protocols, target cards, concise result summaries, completion records, and bounded numerical metrics for every documented DIAL-X experiment family

Families
19
Catalogued runs
229
Packet size
629 bounded records · 1,726,523 bytes
Excluded records
29 path or credential-pattern hits
Verification
python verify_result_packet.py

Evidence boundary. This is a text-only provenance packet, not a media release or an independent rerun. Selection is restricted to top-level registration, preregistration, target-card, result-summary, completion, freeze, status, and bounded metric files no larger than 250,000 bytes. It excludes images, video, prompts, raw model outputs, datasets, caches, nested environments, and large generated ledgers. Every documented experiment family must retain at least one selected record.

Coverage

The bounded packet covers every documented experiment family and catalogued run directory for this book.

Release boundary and outstanding work

Phase one includes

  • The 288 audited Python implementation and runner files linked from Chapters 10–19
  • Registration, configuration, and textual result records selected by an explicit allowlist
  • CPU-scale smoke tests for the DIAL core and one representative DIAL-X family
  • Pinned, portable CLIP model resolution with validated offline-path overrides

Excluded pending review

  • The full 22 GB experiment workspace
  • Private directories, prompts containing confidential material, model weights, caches, and downloaded corpora
  • Raw or generated imagery and video until rights, privacy, and provenance are approved
  • Developmental runs that lack a registration or declared evidence boundary

Outstanding

  • Separate reusable library code from one-off experiment drivers
  • Choose a license, dependency policy, model-access policy, and media provenance format
  • Complete manual security, privacy, and publication review beyond the passing narrow path, credential-pattern, media-type, and file-size checks
  • Publish or vendor the six-file alfworld_bridge dependency

DIAL core ablation ladder · DIAL-ALLORA prototype ladder · CLIC / DIAL-SKFM · OPTIC / DIAL-SkillOpt · RELIC and DIAL-Schema-GIRL · AGENTIC workflow composition · AGENTIC-Lea theorem-construction ladder · Infinitesimal pattern-construction ladder · Categorical AI-Feynman ladder · DIAL-URL coalgebraic theory construction · Simulator-grounded theory extension · Corpus-to-testable-theory construction · GLP-1 theory and grant construction · ARTISTIC textual-inversion and natural-image ladder · ARTISTIC Bongard problems · ARTISTIC sports, pose, and scene enforcement · DIAL-CAN structural-obstruction calibration · DILATE paintbrush geometry · DILATE video, codec, and cross-video transport

Book 01

Categories for AGI

11 documented experiment families

Local core code · curated results

Diagrammatic Backpropagation and Geometric Transformers

A small executable pullback/pushout energy demonstration and two archived commutator-proxy trajectories for Geometric Transformer variants.

Book location
Chapter 6: Diagrammatic Backpropagation · Chapter 7: Geometric Transformers · Chapter 8: Dynamic Compositionality
Code packages
Entry points
scripts/gt/ · scripts/resmlp_gt_commutator_demo.py · scripts/seq2seq_wiki103_prefix_suffix.py · Category-Theory-for-AGI-UMass-CMPSCI-692CT/notebooks/week03_db_colimit_energy.ipynb
Result artifacts
week03_db_colimit_energy.ipynb · two_moons_commutator_energy.png · wiki103_prefix_suffix_cech_obs.png · CHAPTERS_6_8_PROVENANCE.md

Evidence boundary. The Chapter 6 notebook is an executable numerical illustration, not a proof that optimization computes categorical limits in general. The Chapter 8 PNGs have exact code and byte provenance but no surviving per-step raw logs, so they are archived plot-only evidence. Their residuals are task-relative order-sensitivity diagnostics, not categorical limit certificates, causal demonstrations, or general proofs of improved learning.

Open the durable experiment record →

8 machine-resolved code surfaces
  • categories-gt-archive:scripts/datasets_linked_wt2.pydependency
  • categories-gt-archive:scripts/datasets_ptb_wt.pydependency
  • categories-gt-archive:scripts/gt/__init__.pyimplementation
  • categories-gt-archive:scripts/gt/geometric_transformer.pyimplementation
  • categories-gt-archive:scripts/gt/simplicial_mp.pyimplementation
  • categories-gt-archive:scripts/resmlp_gt_commutator_demo.pyimplementation
  • categories-gt-archive:scripts/seq2seq_gt_compare.pyimplementation
  • categories-gt-archive:scripts/seq2seq_wiki103_prefix_suffix.pyimplementation

Documented studies

  • Diagrammatic pullback and pushout energy demonstration
  • Two-moons commutator-proxy trajectory
  • WikiText-103 early commutator-proxy trajectory
  • Geometric Transformer architecture comparison

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

GT information regimes, KET, and TopoCoend

Strict-causal and future-carrier language-modeling comparisons on PTB and WikiText corpora.

Book location
Chapter 9: Information Regimes in Geometric Transformers · Chapter 10: Kan Extension and Topological Coend Transformers · Chapter 11: Structured Language Modeling
Code packages
Entry points
KET/ket_experiments/ · scripts/ket_experiments/
Result artifacts
loss/perplexity plots and recorded comparison tables

Evidence boundary. Future-carrier runs are leakage controls, not valid autoregressive improvements; point estimates are not population effects.

Open the durable experiment record →

42 machine-resolved code surfaces

Documented studies

  • Strict-causal versus future-carrier leakage control
  • PTB blockwise KET comparison
  • WikiText-2 KET diagnostics
  • WikiText-103 KET and TopoCoend comparison

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

Relational manifold learning with Geometric Transformers

Triangle-aware relational recovery, dimension diagnostics, Hodge controls, and the registered MovieLens negative result.

Book location
Chapter 12: Manifold Learning with Geometric Transformers
Code packages
Entry points
relational_descent.py · run_relational_dimension_diagnostic.py · run_hodge_baselines.py · run_movielens_experiment.py
Result artifacts
RELATIONAL_CONFIRMATORY_RESULTS.md · RELATIONAL_DIMENSION_DIAGNOSTIC.md · HODGE_BASELINE_RESULTS.md · MOVIELENS_RESULTS.md

Evidence boundary. The synthetic advantage is specific to the declared two-dimensional bottleneck, and the MovieLens study is a retained negative result rather than evidence of universal relational benefit.

Open the durable experiment record →

12 machine-resolved code surfaces

Documented studies

  • Controlled relational recovery
  • Post-confirmatory dimension diagnostic
  • Pooled Hodge baseline
  • MovieLens registered negative calibration

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

Mean-field diagnostics for Geometric Transformers

Local commutator-energy, normalization, transport, and task diagnostics for a data-limited prefix-to-suffix pilot.

Book location
Chapter 13: Mean-Field Theory of Geometric Transformers
Code packages
  • Geometric and Kan Extension Transformer archivelocal-not-public
Entry points
scripts/seq2seq_scaling_preln_transformer.py · scripts/seq2seq_scaling_GTFull.py · scripts/seq2seq_wiki103_scaling-GT-Lite.py · scripts/seq2seq_wiki103_scaling-GT-Full.py · scripts/seq2seq_wiki103_scaling-GT-MoE.py
Result artifacts
archived manuscript plots and final-value table; raw run logs unavailable

Evidence boundary. The surviving plots and summary values do not include a complete executable snapshot or raw per-example outputs; the chapter treats them as archived pilot evidence only.

Open the durable experiment record →

8 machine-resolved code surfaces
  • categories-gt-archive:scripts/datasets_linked_wt2.pydependency
  • categories-gt-archive:scripts/datasets_ptb_wt.pydependency
  • categories-gt-archive:scripts/seq2seq_scaling_GTFull.pyrunner
  • categories-gt-archive:scripts/seq2seq_scaling_preln_transformer.pyrunner
  • categories-gt-archive:scripts/seq2seq_wiki103_scaling-GT-Full.pyrunner
  • categories-gt-archive:scripts/seq2seq_wiki103_scaling-GT-Lite.pyrunner
  • categories-gt-archive:scripts/seq2seq_wiki103_scaling-GT-MoE.pyrunner
  • categories-gt-archive:scripts/topo_probe.pydependency

Documented studies

  • Residual commutator diagnostic
  • Pre-LN and GT prefix-to-suffix pilot
  • Topological and Dirichlet diagnostic comparison

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

Depth sweeps for Geometric Transformers

GT-Full and Pre-LN depth comparisons with commutator, task, and topological diagnostics.

Book location
Chapter 14: Depth Sweeps for Geometric Transformers
Code packages
  • Geometric and Kan Extension Transformer archivelocal-not-public
Entry points
scripts/seq2seq_scaling_depth.py · scripts/seq2seq_wiki103_scaling-GT-Full.py
Result artifacts
plots/GTFull-* · plots/GTLite-*

Evidence boundary. The archived figures establish finite descriptive trends, not a fitted exponent, asymptotic scaling law, or GT-specific mechanism.

Open the durable experiment record →

6 machine-resolved code surfaces
  • categories-gt-archive:scripts/datasets_linked_wt2.pydependency
  • categories-gt-archive:scripts/datasets_ptb_wt.pydependency
  • categories-gt-archive:scripts/seq2seq_scaling_GTFull.pydependency
  • categories-gt-archive:scripts/seq2seq_scaling_depth.pyrunner
  • categories-gt-archive:scripts/seq2seq_wiki103_scaling-GT-Full.pyrunner
  • categories-gt-archive:scripts/topo_probe.pydependency

Documented studies

  • GT-Full depth sweep
  • Pre-LN depth sweep
  • Average-versus-accumulated obstruction audit

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

Democritus causal claims from language

Causal-claim extraction, Geometric Transformer manifold rendering, and UniCausal comparison.

Book location
Chapter 16: Causal Claims from Language · Chapter 22: Homotopy in Language and Causal Inference
Entry points
pipeline_llmdriver.py
Result artifacts
saved extraction outputs and GT/UMAP ablations

Evidence boundary. Extracted claims and rendered manifolds do not certify causal truth or identification.

Open the durable experiment record →

3 machine-resolved code surfaces

Documented studies

  • Four-domain causal-claim extraction audit
  • Geometric Transformer versus UMAP manifold ablation
  • UniCausal extractor comparison
  • Homotopy-aware claim normalization audit

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

Temporal diffusion over company causal trajectories

Company-year trajectory construction, block denoising, KET comparisons, and animated risk/brand profiles.

Book location
Chapter 17: Temporal Diffusion over Causal Trajectories
Entry points
repository scripts and notebooks
Result artifacts
26-company outputs and movie artifacts

Evidence boundary. The evidence supports a reproducible representation and diagnostics, not validated corporate-risk forecasts.

Open the durable experiment record →

4 machine-resolved code surfaces

Documented studies

  • Twenty-six-company trajectory panel
  • Block-denoising holdout study
  • KET trajectory comparison
  • Animated brand and risk-profile rendering

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

BASKET and ROCKET workflow construction

Partial-order plan construction and section-aware financial workflow reranking.

Book location
Chapter 18: Building Agentic Systems using Kan Extension Transformers · Chapter 24: Predictive State Representations in a Topos
Entry points
BASKET package and output manifests
Result artifacts
summary.json · reranked_summary.json · action_vocab.csv

Evidence boundary. Validation loss and constructed reranking scores are not workflow accuracy, financial forecasts, or causal effects.

Open the durable experiment record →

11 machine-resolved code surfaces

Documented studies

  • BASKET partial-order workflow construction
  • ROCKET section-aware reranking
  • Action-vocabulary and plan-manifest audit

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

Judo Calculus causal-discovery experiments

Sheaf/Kan causal-discovery experiments with DCDI and GES comparisons.

Book location
Chapter 20: Judo Calculus
Code packages
Entry points
artifacts/ · kan_do_dcdi experiments
Result artifacts
synthetic and Sachs result tables

Evidence boundary. The mixed and negative Sachs results remain part of the evidence; runtime plots are descriptive implementation measurements.

Open the durable experiment record →

31 machine-resolved code surfaces

Documented studies

  • Synthetic j-stable DCDI comparison
  • Descriptive runtime measurement
  • Sachs pooled versus regime-stable study
  • GES, CGES, and TC-GES comparison
  • OECD PISA regime-stability study

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

CSQL and ATLAS database construction

Document-to-schema mapping and checked categorical SQL node/edge instances.

Book location
Chapter 21: Csql: Mapping Documents into Topos Causal Model Databases
Code packages
Entry points
ATLAS database builders
Result artifacts
atlas_summary.md · checked Parquet relations

Evidence boundary. Source-edge deletion is an ablation, not a Pearl intervention or counterfactual.

Open the durable experiment record →

3 machine-resolved code surfaces
  • atlas-csql-archive:atlas_WaPoHumanOrigins/atlas_queries.sqlimplementation
  • atlas-csql-archive:atlas_WaPoHumanOrigins/atlas_queries_quant.sqlimplementation
  • atlas-csql-archive:atlas_WaPoHumanOrigins/ncsql_starter_colab.ipynbimplementation

Documented studies

  • Document-to-CSQL schema construction
  • Merged ATLAS provenance audit
  • Causal hub and path queries
  • Source-edge deletion sensitivity analysis

Family-level code link verified; run-level public packet still requires curation.

Local core code · curated results

Deep Universal Reinforcement Learning

GT and diagrammatic-backpropagation ablations for coalgebraic RL losses.

Book location
Chapter 28: Deep URL with Geometric Transformers
Code packages
Entry points
GT_RL_Coalgebra experiment scripts
Result artifacts
thirty-run return and residual summaries

Evidence boundary. Lower structural residual or graph smoothness is not proof that a coalgebra morphism was learned.

Open the durable experiment record →

5 machine-resolved code surfaces
  • deep-url-archive:report_synthetic_mdp_ablation.pyrunner
  • deep-url-archive:report_synthetic_mdp_metrics.pyrunner
  • deep-url-archive:run_synthetic_mdp_ablation.pyrunner
  • deep-url-archive:run_synthetic_mdp_loop.pyrunner
  • deep-url-archive:synthetic_mdp_bridge.pyrunner

Documented studies

  • Thirty-run base learner comparison
  • Geometric Transformer structural-loss ablation
  • Diagrammatic-backpropagation residual ablation
  • Return, residual, and graph-smoothness comparison

Family-level code link verified; run-level public packet still requires curation.

Book 02

Machine Learning from Enforcing Compositionality

8 documented experiment families

Public core code · curated results

BRIDGE/SKFM geometric causal discovery

Random-DAG, injected-confounding, Sachs, screening, graph extraction, and runtime experiments.

Book location
Chapter 11: Geometric Causal Discovery
Entry points
kan_do_dcdi/experiments/
Result artifacts
package reports and benchmark summaries

Evidence boundary. Lie-bracket nonclosure is a diagnostic under registered semantics, not unique identification of a latent confounder.

Open the durable experiment record →

30 machine-resolved code surfaces

Documented studies

  • Random-DAG closure calibration
  • Injected latent-confounding recovery
  • Sachs benchmark study
  • Lie-bracket screening and graph extraction
  • Runtime and robustness comparison

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

LINCS-KET blockwise repair

Five-seed synthetic architecture test and frozen-module PTB block-repair experiments.

Book location
Chapter 12: Repairing Kan-Extension Structure
Entry points
KET and theory-diagnostic scripts
Result artifacts
repair traces and PTB summaries

Evidence boundary. The failed joint gate and all admission conditions remain visible; accepted local steps do not establish general KET superiority.

Open the durable experiment record →

37 machine-resolved code surfaces

Documented studies

  • Five-seed synthetic architecture test
  • Frozen-module PTB block repair
  • Local admission-gate ablation
  • Joint-gate negative result

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

ALLORA adapter composition

Pair/triple composition, scaling, safety-aware repair, and generative adapter experiments.

Book location
Chapter 13: Composable Neural Adapters
Entry points
scripts/run_smoke.sh · scripts/
Result artifacts
export manifest and experiment summaries

Evidence boundary. Reduced noncommutativity does not by itself imply improved generation quality or safety.

Open the durable experiment record →

16 machine-resolved code surfaces

Documented studies

  • Pairwise adapter composition
  • Triple adapter composition
  • Scale and noncommutativity study
  • Safety-aware repair
  • Generative adapter composition

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

LASKO skill optimization

Cheap bracket probes, validation calls, scale experiments, and anchor-chain benchmarks.

Book location
Chapter 14: Skills over Lie Algebroids
Code packages
Entry points
scripts/run_anchor_chain_benchmark.py
Result artifacts
benchmark summaries and workflow traces

Evidence boundary. Compute savings are relative to the supplied skill language and validation protocol; the system does not invent that language.

Open the durable experiment record →

3 machine-resolved code surfaces

Documented studies

  • Cheap Lie-bracket probe calibration
  • Validation-call economy study
  • Skill-library scale experiment
  • Anchor-chain benchmark

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

GIRL infinitesimal reinforcement learning

Bellman-obstruction, regret, ranking, harmful-admission, transport, and post-shift tests.

Book location
Chapter 15: Infinitesimal Reinforcement Learning
Code packages
Entry points
scripts/run_smoke.sh · scripts/run_full.sh
Result artifacts
full-suite reports

Evidence boundary. Evaluation-only exact utility is not available to the controller; missed thresholds and abstentions remain negative evidence.

Open the durable experiment record →

24 machine-resolved code surfaces

Documented studies

  • Bellman-obstruction calibration
  • Regret and ranking comparison
  • Harmful-admission control
  • Policy-transport study
  • Post-shift recovery and abstention test

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

LINCS-RLHF preference experiments

Calibration/power, opposite-stratum localization, relational games, and incomplete-support studies.

Book location
Chapter 16: Learning from Structured Preferences
Code packages
Entry points
scripts/run_smoke.sh · scripts/run_full.sh
Result artifacts
E1–E4 registered result grids

Evidence boundary. Oracle population quantities are evaluation-only and do not characterize deployed preference learning.

Open the durable experiment record →

3 machine-resolved code surfaces

Documented studies

  • E1 calibration and power
  • E2 opposite-stratum localization
  • E3 relational preference game
  • E4 incomplete-support and abstention

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

RADAR relational manifold recovery

Relational descent, chart construction and repair, solver comparisons, and MovieLens evaluation.

Book location
Chapter 17: Relational Manifold Recovery
Code packages
Entry points
scripts/run_smoke.sh · scripts/prepare_movielens_data.py
Result artifacts
registered plans and result reports

Evidence boundary. The MovieLens calibration failure remains visible; task accuracy and structural admission are separate outcomes.

Open the durable experiment record →

14 machine-resolved code surfaces

Documented studies

  • Relational-descent calibration
  • Chart construction and repair
  • Solver comparison
  • MovieLens evaluation and negative admission result

Family-level code link verified; run-level public packet still requires curation.

Public core code · curated results

SID sheaf-based distributed learning

Thirty-seed typed compatibility, prediction, and promotion replays.

Book location
Chapter 18: Sheaf-Based Distributed Learning
Code packages
Entry points
scripts/run_smoke.sh · scripts/run_full.sh
Result artifacts
reconstructed full-suite reports

Evidence boundary. Reference promotion labels are evaluation-only; replay success does not certify external truth or safety.

Open the durable experiment record →

7 machine-resolved code surfaces

Documented studies

  • Thirty-seed compatibility replay
  • Typed prediction replay
  • Promotion-gate replay
  • Reference-label and abstention audit

Family-level code link verified; run-level public packet still requires curation.

Book 03

Infinitesimal Creativity

19 documented experiment families

Local core code · frozen run records

DIAL core ablation ladder

Exact, differentiated, mixed-witness, operator-estimation, static-causal, active, sequential, and nonstationary DIAL tests.

Book location
Chapter 10: DIAL as an Algorithmic Family
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py or named family script in each packet
Result artifacts
RESULTS.md · completion.json · registration.json

Evidence boundary. These studies calibrate component semantics and do not establish autonomous creativity.

Open the durable experiment record →

11 machine-resolved code surfaces
  • synthetic-creativity-archive:dial0a_linear_exact.pyimplementation
  • synthetic-creativity-archive:dial0b_differentiation.pyimplementation
  • synthetic-creativity-archive:dial0c_mixed_witness_ablation.pyimplementation
  • synthetic-creativity-archive:dial0d_dual_operator_estimation.pyimplementation
  • synthetic-creativity-archive:dial1_static_causal_identification.pyimplementation
  • synthetic-creativity-archive:dial1b_equivalence_intervention_design.pyimplementation
  • synthetic-creativity-archive:dial1c_active_intervention_selection.pyimplementation
  • synthetic-creativity-archive:dial1d_sequential_intervention_escalation.pyimplementation
  • synthetic-creativity-archive:dial2a_nonstationary_decisions.pyimplementation
  • synthetic-creativity-archive:dial2n1_girl_state_admission.pyimplementation
  • synthetic-creativity-archive:dial2n_girl_recurrence.pyimplementation
13 resolved run records
  • 2026-08-09_dial0_ablation_design
  • 2026-08-09_dial0a_linear_exact_calibration
  • 2026-08-09_dial0a_linear_exact_v2_calibration
  • 2026-08-09_dial0b_differentiation_calibration
  • 2026-08-09_dial0c_mixed_witness_registered
  • 2026-08-09_dial0d_dual_operator_estimation_registered
  • 2026-08-09_dial1_static_causal_registered
  • 2026-08-09_dial1b_equivalence_design_registered
  • 2026-08-09_dial1c_active_intervention_registered
  • 2026-08-09_dial1d_sequential_escalation_registered
  • 2026-08-09_dial2a_nonstationary_decisions_registered
  • 2026-08-09_dial2n1_girl_state_admission_registered
  • 2026-08-09_dial2n_girl_recurrence_registered

Local core code · frozen run records

DIAL-ALLORA prototype ladder

Adapter-subspace estimation, generative composition, oracle-language, selective-learning, and active-acquisition tests.

Book location
Chapter 10: DIAL as an Algorithmic Family
Code packages
Entry points
family scripts and registered packet run.py files
Result artifacts
RESULTS.md · registration.json

Evidence boundary. The ladder studies whether missing adapter structure can be proposed inside bounded languages; it is not a general creative adapter system.

Open the durable experiment record →

39 machine-resolved code surfaces
23 resolved run records
  • 2026-08-12_dial_allora_a0b0_registered
  • 2026-08-12_dial_allora_b1_registered
  • 2026-08-12_dial_allora_b2_registered
  • 2026-08-12_dial_allora_b3_registered
  • 2026-08-12_dial_allora_b4_registered
  • 2026-08-12_dial_allora_b51_registered
  • 2026-08-12_dial_allora_b5_registered
  • 2026-08-12_dial_allora_b61_registered
  • 2026-08-12_dial_allora_b62_registered
  • 2026-08-12_dial_allora_b63_registered
  • 2026-08-12_dial_allora_b64_registered
  • 2026-08-12_dial_allora_b65_registered
  • 2026-08-12_dial_allora_b66_registered
  • 2026-08-12_dial_allora_b67_registered
  • 2026-08-12_dial_allora_b68_registered
  • 2026-08-12_dial_allora_b6_registered
  • 2026-08-13_dial_allora_active_acquisition_registered
  • 2026-08-13_dial_allora_affine_learner_registered
  • 2026-08-13_dial_allora_b69_registered
  • 2026-08-13_dial_allora_boundary_enrollment_registered
  • 2026-08-13_dial_allora_oracle_expanded_registered
  • 2026-08-13_dial_allora_oracle_registered
  • 2026-08-13_dial_allora_selective_learner_registered

Local core code · frozen run records

CLIC / DIAL-SKFM

Exact closure repair, finite operator uncertainty, active counter-witness acquisition, and unsupported-alternative posterior studies.

Book location
Chapter 11: Causal Learning with Infinitesimal Creativity
Code packages
Entry points
run.py in registered packets
Result artifacts
RESULTS.md · completion.json · STATUS.md

Evidence boundary. Only the exact rung meets its registered admission target; later uncertainty studies retain their negative decisions.

Open the durable experiment record →

32 machine-resolved code surfaces
7 resolved run records
  • 2026-08-11_dial_skfm_factored_registered
  • 2026-08-11_dial_skfm_typed_transducer_registered
  • 2026-08-13_dial_skfm_1a_exact_closure_repair_v1_registered
  • 2026-08-14_dial_skfm_1b_finite_operator_uncertainty_v1_registered
  • 2026-08-14_dial_skfm_1b_finite_operator_uncertainty_v2_registered
  • 2026-08-14_dial_skfm_1c_active_counterwitness_v1_registered
  • 2026-08-14_dial_skfm_1d_explicit_alternative_posterior_v1_registered

Local core code · frozen run records

OPTIC / DIAL-SkillOpt

Held-out calibration, sealed compound-skill failure, and factored-interface repair.

Book location
Chapter 12: Skill Optimization with Infinitesimal Creativity
Code packages
Entry points
packet scripts and DIAL-SkillOpt driver
Result artifacts
GPT_OSS_20B_REPORT.md · TERMINATION_REPORT.md · DIAGNOSTIC_REPORT.md

Evidence boundary. The v2 interface repair is a separately registered schema correction, not a post-hoc relabeling of the failed run.

Open the durable experiment record →

5 machine-resolved code surfaces
  • lincs-lasko:scripts/run_anchor_chain_benchmark.py ↗base-system
  • synthetic-creativity-archive:dial_skillopt1_preregistered.pyimplementation
  • synthetic-creativity-archive:dial_skillopt1d_factored.pyimplementation
  • synthetic-creativity-archive:dial_skillopt1d_factored_v2.pyimplementation
  • synthetic-creativity-archive:dial_skillopt_registered.pyimplementation
4 resolved run records
  • 2026-08-11_dial_skillopt1_preregistered
  • 2026-08-11_dial_skillopt1d_factored_registered
  • 2026-08-11_dial_skillopt1d_factored_v2_registered
  • 2026-08-11_dial_skillopt_registered

Local core code · frozen run records

RELIC and DIAL-Schema-GIRL

Switching-world filters, schema acquisition, TextWorld/ALFWorld planning, active counter-witnesses, composition, and behavioral quotients.

Book location
Chapter 13: Reinforcement Learning with Infinitesimal Creativity
Code packages
Entry points
family scripts and registered packet run.py files
Result artifacts
RESULTS.md · completion.json · recovery/status artifacts

Evidence boundary. Recovery runs repair documented execution defects and are not independent replications; failed and abstinent rungs remain in the record.

Open the durable experiment record →

64 machine-resolved code surfaces
61 resolved run records
  • 2026-08-11_dial_girl_anytime_mixture_development
  • 2026-08-11_dial_girl_anytime_mixture_registered
  • 2026-08-11_dial_girl_composite_markov_development
  • 2026-08-11_dial_girl_composite_markov_registered
  • 2026-08-11_dial_girl_composite_null_development
  • 2026-08-11_dial_girl_composite_null_registered
  • 2026-08-11_dial_girl_decision_value_registered
  • 2026-08-11_dial_girl_learned_filter_development
  • 2026-08-11_dial_girl_learned_filter_registered
  • 2026-08-11_dial_girl_learned_prototypes_development
  • 2026-08-11_dial_girl_learned_prototypes_registered
  • 2026-08-11_dial_girl_long_horizon_markov_registered
  • 2026-08-11_dial_girl_pilot_priors_development
  • 2026-08-11_dial_girl_pilot_priors_registered
  • 2026-08-11_dial_girl_rate040_equivalence_development
  • 2026-08-11_dial_girl_rate040_equivalence_registered
  • 2026-08-11_dial_girl_sequential_evidence_registered
  • 2026-08-11_dial_girl_skillopt_development
  • 2026-08-11_dial_girl_skillopt_registered
  • 2026-08-11_dial_girl_switch_transport_development
  • 2026-08-11_dial_girl_switch_transport_registered
  • 2026-08-11_dial_girl_trajectory_development
  • 2026-08-11_dial_girl_trajectory_registered
  • 2026-08-11_dial_schema_girl_0_registered
  • 2026-08-11_dial_schema_girl_1a_textworld_registered
  • 2026-08-11_dial_schema_girl_1a_textworld_v2_registered
  • 2026-08-11_dial_schema_girl_1b_balanced_textworld_registered
  • 2026-08-11_dial_schema_girl_1c_planning_registered
  • 2026-08-11_dial_schema_girl_1d1_balanced_relational_fit_registered
  • 2026-08-11_dial_schema_girl_1d_relational_counterwitness_registered
  • 2026-08-11_dial_schema_girl_1e_relational_planning_replay_registered
  • 2026-08-11_dial_schema_girl_1f1_total_location_state_registered
  • 2026-08-11_dial_schema_girl_1f_containment_discovery_registered
  • 2026-08-11_dial_schema_girl_1g1_colocated_take_planning_registered
  • 2026-08-11_dial_schema_girl_1g_fresh_containment_planning_registered
  • 2026-08-12_dial_schema_girl_1h_persistent_tool_planning_registered
  • 2026-08-12_dial_schema_girl_1h_recovery_analysis_registered
  • 2026-08-12_relic_alfworld_a0_full_validation_registered
  • 2026-08-12_relic_alfworld_a0_manifest_launcher_registered
  • 2026-08-12_relic_alfworld_a1_controls_sensitivity_registered
  • 2026-08-12_relic_alfworld_b0_1_textdqn_h5py_recovery_registered
  • 2026-08-12_relic_alfworld_b0_2_textdqn_torchvision_recovery_registered
  • 2026-08-12_relic_alfworld_b0_3_reporting_recovery_registered
  • 2026-08-12_relic_alfworld_b0_3_textdqn_distilbert_recovery_registered
  • 2026-08-12_relic_alfworld_b0_textdqn_smoke_registered
  • 2026-08-12_relic_alfworld_b1_intrinsic_calibration_registered
  • 2026-08-12_relic_alfworld_b2_learning_onset
  • 2026-08-13_relic_alfworld_b3_horizon_recovery_launcher_registered
  • 2026-08-13_relic_alfworld_b3_recovery_launcher_registered
  • 2026-08-13_relic_alfworld_b3_schema_accommodation_horizon_recovery
  • 2026-08-13_relic_alfworld_b3_schema_accommodation_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_v1_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_v2_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_v3_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_v4_registered
  • 2026-08-13_relic_alfworld_b4_composition_construction_v5_registered
  • 2026-08-13_relic_alfworld_b51_active_counterwitness_v1_registered
  • 2026-08-13_relic_alfworld_b52_behavioral_quotient_v1_registered
  • 2026-08-13_relic_alfworld_b52_behavioral_quotient_v2_registered
  • 2026-08-13_relic_alfworld_b5_stage_construction_v1_registered

Local core code · frozen run records

AGENTIC workflow composition

Exact and estimated CLIC/OPTIC/RELIC composition with active and hierarchical acquisition.

Book location
Chapter 14: Composing Creative Workflows
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in registered packets
Result artifacts
RESULTS.md · completion.json

Evidence boundary. The authoritative 0D.2 result is conditional and small-scale; earlier rungs are calibrations and failures, not pooled trials.

Open the durable experiment record →

8 machine-resolved code surfaces
  • synthetic-creativity-archive:agentic_0a_exact_workflow_composition.pyrunner
  • synthetic-creativity-archive:agentic_0b1_active_boundary_acquisition.pyrunner
  • synthetic-creativity-archive:agentic_0b_estimated_clic_composition.pyrunner
  • synthetic-creativity-archive:agentic_0c1_active_optic_acquisition.pyrunner
  • synthetic-creativity-archive:agentic_0c_estimated_optic_composition.pyrunner
  • synthetic-creativity-archive:agentic_0d1_active_relic_acquisition.pyrunner
  • synthetic-creativity-archive:agentic_0d2_hierarchical_active_relic.pyrunner
  • synthetic-creativity-archive:agentic_0d_estimated_relic_composition.pyrunner
8 resolved run records
  • 2026-08-14_agentic_0a_exact_workflow_composition_v1_registered
  • 2026-08-14_agentic_0b1_active_boundary_acquisition_v1_registered
  • 2026-08-14_agentic_0b_estimated_clic_composition_v1_registered
  • 2026-08-14_agentic_0c1_active_optic_acquisition_v1_registered
  • 2026-08-14_agentic_0c_estimated_optic_composition_v1_registered
  • 2026-08-14_agentic_0d1_active_relic_acquisition_v1_registered
  • 2026-08-14_agentic_0d2_hierarchical_active_relic_v1_registered
  • 2026-08-14_agentic_0d_estimated_relic_composition_v1_registered

Local core code · frozen run records

AGENTIC-Lea theorem-construction ladder

Tool capability, scaffolded abstraction, fixed-vocabulary proof construction, structural repair, and early-stop tests.

Book location
Chapter 15: Mathematical Theory Construction
Code packages
Entry points
packet scripts invoking Lea
Result artifacts
RESULTS.md · RESULTS_QWEN.md

Evidence boundary. Formal validity in a supplied or expanded language does not by itself establish mathematical creativity.

Open the durable experiment record →

21 machine-resolved code surfaces
  • lea-upstream:apps/lea-standalone/prover/examples/Ackermann.lean ↗external-dependency
  • lea-upstream:apps/lea-standalone/prover/examples/Logic.lean ↗external-dependency
  • lea-upstream:apps/lea-standalone/prover/examples/StackMachine.lean ↗external-dependency
  • lea-upstream:apps/lea-standalone/prover/examples/test_proof.lean ↗external-dependency
  • synthetic-creativity-archive:agentic_lea_0/extract_proposal.pyrunner
  • synthetic-creativity-archive:agentic_lea_0/run_audit.pyrunner
  • synthetic-creativity-archive:agentic_lea_0/run_lea.pyrunner
  • synthetic-creativity-archive:agentic_lea_1/run_audit.pyrunner
  • synthetic-creativity-archive:agentic_lea_1/run_followup.pyrunner
  • synthetic-creativity-archive:agentic_lea_1/run_followup2.pyrunner
  • synthetic-creativity-archive:agentic_lea_1/run_lea.pyrunner
  • synthetic-creativity-archive:agentic_lea_2/run_audit.pyrunner
  • synthetic-creativity-archive:agentic_lea_2/run_lea.pyrunner
  • synthetic-creativity-archive:agentic_lea_2b/run_audit.pyrunner
  • synthetic-creativity-archive:agentic_lea_2b/run_lea.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/freeze_stage_a.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/run_audit.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/run_stage_a.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/run_stage_a_followup.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/run_stage_a_followup2.pyrunner
  • synthetic-creativity-archive:agentic_lea_2c/run_stage_b.pyrunner
5 resolved run records
  • agentic_lea_0
  • agentic_lea_1
  • agentic_lea_2
  • agentic_lea_2b
  • agentic_lea_2c

Local core code · frozen run records

Infinitesimal pattern-construction ladder

Symmetry, latent factors, entangled worlds, counter-witnesses, predictive-state sheaves, holonomy, power/coverage, and active intervention studies.

Book location
Chapter 15: Mathematical Theory Construction
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
named IPC scripts and packet drivers
Result artifacts
GPT_OSS_20B_REPORT.md · RESULTS.md · registration.json

Evidence boundary. Version-matched reported rungs are separated from superseded pilots; synthetic recovery is not open-ended field construction.

Open the durable experiment record →

21 machine-resolved code surfaces
  • synthetic-creativity-archive:ipc0_beta.pyrunner
  • synthetic-creativity-archive:ipc0_beta_v2.pyrunner
  • synthetic-creativity-archive:ipc0_gamma.pyrunner
  • synthetic-creativity-archive:ipc0_smoke.pyrunner
  • synthetic-creativity-archive:ipc10_predictive_state_sheaf.pyrunner
  • synthetic-creativity-archive:ipc11_hidden_overlap_registration.pyrunner
  • synthetic-creativity-archive:ipc12_cech_cycle_consistency.pyrunner
  • synthetic-creativity-archive:ipc13_approximate_holonomy.pyrunner
  • synthetic-creativity-archive:ipc14_trajectory_holonomy.pyrunner
  • synthetic-creativity-archive:ipc15_power_coverage_ladder.pyrunner
  • synthetic-creativity-archive:ipc16_active_interventions.pyrunner
  • synthetic-creativity-archive:ipc1_symmetry.pyrunner
  • synthetic-creativity-archive:ipc2_diagnostic.pyrunner
  • synthetic-creativity-archive:ipc2_mixed_symmetry.pyrunner
  • synthetic-creativity-archive:ipc3_latent_property.pyrunner
  • synthetic-creativity-archive:ipc4_latent_factors.pyrunner
  • synthetic-creativity-archive:ipc5_noisy_latent_factors.pyrunner
  • synthetic-creativity-archive:ipc6_doctrine_obstruction.pyrunner
  • synthetic-creativity-archive:ipc7_entangled_latent_graph.pyrunner
  • synthetic-creativity-archive:ipc8_edge_counter_witness.pyrunner
  • synthetic-creativity-archive:ipc9_anchor_free_subspace.pyrunner
22 resolved run records
  • 2026-08-06_ipc0_alpha
  • 2026-08-06_ipc0_beta_registered
  • 2026-08-06_ipc0_beta_setup
  • 2026-08-06_ipc0_beta_v2_registered
  • 2026-08-06_ipc0_gamma_registered
  • 2026-08-06_ipc10_predictive_state_sheaf_registered
  • 2026-08-06_ipc11_hidden_overlap_registration_registered
  • 2026-08-06_ipc1_symmetry_registered
  • 2026-08-06_ipc2_delta_diagnostic_registered
  • 2026-08-06_ipc2_mixed_symmetry_registered
  • 2026-08-06_ipc3_latent_property_registered
  • 2026-08-06_ipc4_latent_factors_registered
  • 2026-08-06_ipc5_noisy_latent_factors_registered
  • 2026-08-06_ipc6_doctrine_obstruction_registered
  • 2026-08-06_ipc7_entangled_latent_graph_registered
  • 2026-08-06_ipc8_edge_counter_witness_registered
  • 2026-08-06_ipc9_anchor_free_subspace_registered
  • 2026-08-07_ipc12_cech_cycle_consistency_registered
  • 2026-08-07_ipc13_approximate_holonomy_registered
  • 2026-08-07_ipc14_trajectory_estimated_holonomy_registered
  • 2026-08-07_ipc15_unconditional_power_coverage_registered
  • 2026-08-07_ipc16_active_interventions_registered

Local core code · frozen run records

Categorical AI-Feynman ladder

Categorical symbolic regression, ambiguity, unnamed generators, sequential probes, open generator search, and proposal-aware acquisition.

Book location
Chapter 15: Mathematical Theory Construction
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
named AF scripts and packet drivers
Result artifacts
GPT_OSS_20B_REPORT.md · RESULTS.md

Evidence boundary. Superseded runs remain diagnostic records and are excluded from the reported endpoints for stated protocol defects.

Open the durable experiment record →

6 machine-resolved code surfaces
  • synthetic-creativity-archive:af0_categorical_feynman.pyrunner
  • synthetic-creativity-archive:af1_ambiguous_active_feynman.pyrunner
  • synthetic-creativity-archive:af2_unnamed_generator.pyrunner
  • synthetic-creativity-archive:af3_sequential_generator_probes.pyrunner
  • synthetic-creativity-archive:af4_open_generator_search.pyrunner
  • synthetic-creativity-archive:af5_proposal_aware_active.pyrunner
10 resolved run records
  • 2026-08-07_af0_categorical_feynman_registered
  • 2026-08-07_af0_categorical_feynman_v2_registered
  • 2026-08-07_af1_ambiguous_active_feynman_registered
  • 2026-08-07_af2_unnamed_generator_registered
  • 2026-08-07_af2_unnamed_generator_v2_registered
  • 2026-08-07_af3_sequential_generator_probes_registered
  • 2026-08-07_af3_sequential_generator_probes_registered_v2
  • 2026-08-07_af4_open_generator_search_registered
  • 2026-08-07_af4_open_generator_search_registered_v2
  • 2026-08-07_af5_proposal_aware_active_registered

Local core code · frozen run records

DIAL-URL coalgebraic theory construction

Exact type recovery, finite trajectories, active discrimination, grammar synthesis, and a withheld composition law.

Book location
Chapter 15: Mathematical Theory Construction
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · PREREGISTRATION.md

Evidence boundary. The final rung misses its hardest prespecified endpoint; no reported run constructs the full reusable coalgebraic learning theory.

Open the durable experiment record →

5 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-19_dial_url_0_exact_type_recovery_registered/run_dial_url_0.pyrunner
  • synthetic-creativity-archive:2026-08-19_dial_url_1_finite_trajectory_registered/run_dial_url_1.pyrunner
  • synthetic-creativity-archive:2026-08-19_dial_url_2_active_discrimination/run_dial_url_2.pyrunner
  • synthetic-creativity-archive:2026-08-19_dial_url_3_grammar_synthesis/run_dial_url_3.pyrunner
  • synthetic-creativity-archive:2026-08-19_dial_url_4_withheld_composition_law/run_dial_url_4.pyrunner
5 resolved run records
  • 2026-08-19_dial_url_0_exact_type_recovery_registered
  • 2026-08-19_dial_url_1_finite_trajectory_registered
  • 2026-08-19_dial_url_2_active_discrimination
  • 2026-08-19_dial_url_3_grammar_synthesis
  • 2026-08-19_dial_url_4_withheld_composition_law

Local core code · frozen run records

Simulator-grounded theory extension

Singing-mouse mechanism recovery, observer construction, active probe selection, and latent-mediator extension.

Book location
Chapter 16: Simulator-Grounded Theory Extension
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · completion.json

Evidence boundary. Projection-support recovery does not identify vocal causality; each result remains simulator-relative.

Open the durable experiment record →

4 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-20_sgte1_singing_mice_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_sgte2_singing_mice_observer_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_sgte3_active_probe_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_sgte4_latent_mediator_registered/run.pyrunner
4 resolved run records
  • 2026-08-20_sgte1_singing_mice_registered
  • 2026-08-20_sgte2_singing_mice_observer_registered
  • 2026-08-20_sgte3_active_probe_registered
  • 2026-08-20_sgte4_latent_mediator_registered

Local core code · frozen run records

Corpus-to-testable-theory construction

Generated corpora, alias registration, context covers, active acquisition, sketch extension, and domain transport.

Book location
Chapter 17: From Scientific Documents to Testable Theories
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · registration.json

Evidence boundary. CTTE-2 remains not admitted after missing its localization gate; generated micro-corpora are controlled test objects, not scientific literatures.

Open the durable experiment record →

6 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-20_ctte0_generated_corpus_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_ctte1_alias_registration_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_ctte2_context_cover_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_ctte3_active_acquisition_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_ctte4_sketch_extension_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_ctte5_domain_transport_registered/run.pyrunner
6 resolved run records
  • 2026-08-20_ctte0_generated_corpus_registered
  • 2026-08-20_ctte1_alias_registration_registered
  • 2026-08-20_ctte2_context_cover_registered
  • 2026-08-20_ctte3_active_acquisition_registered
  • 2026-08-20_ctte4_sketch_extension_registered
  • 2026-08-20_ctte5_domain_transport_registered

Local core code · frozen run records

GLP-1 theory and grant construction

Hidden-context pilots, semantic admission, typed IR, cross-field invariants, auditable realization, length bounds, and controlled paraphrase.

Book location
Chapter 17: From Scientific Documents to Testable Theories
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in registered packets
Result artifacts
RESULTS.md · MANUAL_SEMANTIC_AUDIT.md · auditable_proposal_clean.md

Evidence boundary. Formal gate passage is not scientific correctness, fundability, or evidence of improved prose; the negative evaluator results remain explicit.

Open the durable experiment record →

11 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-20_glp1_grant0_program_compilation_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant1_semantic_firewall_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant2_typed_ir_roundtrip_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant3_cross_field_invariants_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant4_1_sanitized_realization_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant4_2_structure_first_realization_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant4_3_bounded_realization_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant4_4_controlled_paraphrase_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_grant4_auditable_realization_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_pilot0_withdrawal_context_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-20_glp1_pilot1_hidden_context_registered/run.pyrunner
11 resolved run records
  • 2026-08-20_glp1_grant0_program_compilation_registered
  • 2026-08-20_glp1_grant1_semantic_firewall_registered
  • 2026-08-20_glp1_grant2_typed_ir_roundtrip_registered
  • 2026-08-20_glp1_grant3_cross_field_invariants_registered
  • 2026-08-20_glp1_grant4_1_sanitized_realization_registered
  • 2026-08-20_glp1_grant4_2_structure_first_realization_registered
  • 2026-08-20_glp1_grant4_3_bounded_realization_registered
  • 2026-08-20_glp1_grant4_4_controlled_paraphrase_registered
  • 2026-08-20_glp1_grant4_auditable_realization_registered
  • 2026-08-20_glp1_pilot0_withdrawal_context_registered
  • 2026-08-20_glp1_pilot1_hidden_context_registered

Local core code · frozen run records

ARTISTIC textual-inversion and natural-image ladder

Concept naming, two-stage repair, component declarations, independent structural observers, and multi-object natural imagery.

Book location
Chapter 18: Generative Visual Languages
Code packages
Entry points
scripts and manifests inside the consolidated packet
Result artifacts
ARTISTIC_TI*_RESULTS.md · ARTISTIC_NI*_RESULTS.md · completion JSON files

Evidence boundary. The results demonstrate bounded visual repair and selective admission, not transformational creativity.

Open the durable experiment record →

74 machine-resolved code surfaces
  • diffusers-upstream:examples/textual_inversion/textual_inversion.py ↗external-dependency
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/build_artistic_ti2d_adaptive.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/build_artistic_ti3_adaptive.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/build_typed_visual_adaptive.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/calibrate_original_mask_observer.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/diagnose_artistic_ni2b_relations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/diagnose_artistic_soccer2.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/download_frozen_model.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4_declarations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4b_proximity_declarations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4c_proximity_declarations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4c_proximity_declarations_v2.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4d_semantic_declarations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/estimate_ti4e_multicue_declarations.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ni0.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ni2a.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ni2b.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ni2c.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti2d.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti3.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti3b.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti3c.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti3d.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4a.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4b.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4b_failed_run.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4c.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4d.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_artistic_ti4e.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_dial_stage2.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_structural_ground_truth.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_substantive.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_typed_visual.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_typed_visual_adaptive.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/evaluate_typed_visual_masks.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/extract_artistic_ni2a_coco_metadata.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/fit_ti4c_source_observers.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_dataset.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_dial_stage2_dataset.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_original_virel_masks.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_ti4_new_objects.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_ti4b_fresh_cohort.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_ti4c_fresh_cohort.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_ti4d_fresh_cohort.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_ti4e_fresh_cohort.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/generate_typed_visual_scenes.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/infer_smoke.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/make_adaptive_contact_sheet.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/make_random_control.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/package_artistic_ni0_sealed.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/prepare_artistic_ni0_public.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/prepare_artistic_ni2a_public.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/prepare_artistic_ni2b_public.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/render_artistic_ni2a_typed_overlay.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/render_artistic_ni2b_typed_overlay.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_external_single.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ni0.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ni2a.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ni2b.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_soccer2.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti2d_inference.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti3_inference.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti3b_retraction.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti3c_componentwise.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti3d_localized.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti4a.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti4b.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti4c.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti4d.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_artistic_ti4e.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_dial_stage2_inference.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_substantive_inference.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_typed_visual_adaptive_inference.pyrunner
  • synthetic-creativity-archive:2026-08-14_artistic_ti0_concept_naming_registered/scripts/run_typed_visual_inference.pyrunner
1 resolved run records
  • 2026-08-14_artistic_ti0_concept_naming_registered

Local core code · frozen run records

ARTISTIC Bongard problems

Rule reuse, composition, declaration extension, blind corpus tests, typed selection, and active oracle queries.

Book location
Chapter 18: Generative Visual Languages
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
packet scripts
Result artifacts
PROGRESS_REPORT.md · RESULTS.md

Evidence boundary. Corpus-4 is an oracle-assisted developmental upper bound, not a blind system comparison.

Open the durable experiment record →

9 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_bp0_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_bp1_compositional_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_bp2_declaration_extension_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus0_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus1_historical/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus2_blind/run_blind.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus2_blind/score_revealed.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus3_typed/run.pyrunner
  • synthetic-creativity-archive:2026-08-15_artistic_bongard_corpus4_active/run.pyrunner
8 resolved run records
  • 2026-08-15_artistic_bongard_bp0_registered
  • 2026-08-15_artistic_bongard_bp1_compositional_registered
  • 2026-08-15_artistic_bongard_bp2_declaration_extension_registered
  • 2026-08-15_artistic_bongard_corpus0_registered
  • 2026-08-15_artistic_bongard_corpus1_historical
  • 2026-08-15_artistic_bongard_corpus2_blind
  • 2026-08-15_artistic_bongard_corpus3_typed
  • 2026-08-15_artistic_bongard_corpus4_active

Local core code · frozen run records

ARTISTIC sports, pose, and scene enforcement

Soccer, cricket, tennis, baseball, scene-graph, layout, slot, pose, sketch-acquisition, and Firefly challenge repairs.

Book location
Chapter 18: Generative Visual Languages
Code packages
Entry points
packet scripts, prompts, and image-editor contracts
Result artifacts
RESULTS.md · prompts.json · typed audit overlays · completion files

Evidence boundary. Three early soccer outputs lack frozen registration and remain qualitative demonstrations; image repair is not itself creative theory extension.

Open the durable experiment record →

21 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-15_artistic_cricket0_v1/run.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_active_admission_cricket2/run.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_layout_enforce_0/aggregate_audit.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_nanobanana_sports0/make_comparisons.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_repair_compiler_a0/evaluate.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_scenegraph_cricket1/run.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/aggregate_generation_audit.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/compile_image_prompt.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/make_blinded_contact_sheets.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/prepare_generation_prompts.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/retrieve.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/run_acquisition.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_sketch_acquire_0/validate_sketch.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_tangent_cricket0/make_trajectory_montage.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_tangent_cricket0/run.pyrunner
  • synthetic-creativity-archive:2026-08-16_artistic_tennis_0/evaluate.pyrunner
  • synthetic-creativity-archive:2026-08-16_dial_artistic_1_mixed_witness_ablation/extract_foreground.pyrunner
  • synthetic-creativity-archive:2026-08-16_dial_artistic_1_mixed_witness_ablation/run_ablation.pyrunner
  • synthetic-creativity-archive:2026-08-16_dial_artistic_2_uncertain_observers/run_observers.pyrunner
  • synthetic-creativity-archive:2026-08-16_dial_artistic_2b_supported_observers/run_test.pyrunner
  • synthetic-creativity-archive:2026-08-16_dial_artistic_2c_prototype_observers/run_test.pyrunner
20 resolved run records
  • 2026-08-15_artistic_cricket0_v1
  • 2026-08-15_artistic_soccer2_diagnostic_v1
  • 2026-08-15_artistic_soccer2_v1
  • 2026-08-15_artistic_soccer_external_v1
  • 2026-08-16_artistic_action_pose_enforce_0
  • 2026-08-16_artistic_active_admission_cricket2
  • 2026-08-16_artistic_layout_enforce_0
  • 2026-08-16_artistic_nanobanana_sports0
  • 2026-08-16_artistic_pose_enforce_0
  • 2026-08-16_artistic_repair_compiler_a0
  • 2026-08-16_artistic_scenegraph_cricket1
  • 2026-08-16_artistic_sketch_acquire_0
  • 2026-08-16_artistic_slot_compose_0
  • 2026-08-16_artistic_tangent_cricket0
  • 2026-08-16_artistic_tennis_0
  • 2026-08-16_dial_artistic_1_mixed_witness_ablation
  • 2026-08-16_dial_artistic_2_uncertain_observers
  • 2026-08-16_dial_artistic_2b_supported_observers
  • 2026-08-16_dial_artistic_2c_prototype_observers
  • 2026-08-17_artistic_firefly_cricket_challenge_0

Local core code · frozen run records

DIAL-CAN structural-obstruction calibration

Bracket calibration, stochastic GAN dynamics, smooth observer ablation, decision regions, and finite-sample estimation.

Book location
Chapter 19: Artistic Theory Extension
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · registration.json

Evidence boundary. The finite-sample controller does not meet the full learned-controller target; style-classifier deviation is not theory extension.

Open the durable experiment record →

5 machine-resolved code surfaces
  • synthetic-creativity-archive:dial_can_0_bracket_calibration.pyimplementation
  • synthetic-creativity-archive:dial_can_1_stochastic_gan.pyimplementation
  • synthetic-creativity-archive:dial_can_1b_smooth_observer.pyimplementation
  • synthetic-creativity-archive:dial_can_2_decision_regions.pyimplementation
  • synthetic-creativity-archive:dial_can_3_finite_estimator.pyimplementation
5 resolved run records
  • 2026-08-18_dial_can_0_bracket_calibration_registered
  • 2026-08-18_dial_can_1_stochastic_gan_registered
  • 2026-08-18_dial_can_1b_smooth_observer_registered
  • 2026-08-18_dial_can_2_decision_regions_registered
  • 2026-08-18_dial_can_3_finite_estimator_registered

Local core code · frozen run records

DILATE paintbrush geometry

Residual calibration, intent-conditioned paintbrushes, algebroid paths, fiber metrics, connections, and held-out transport.

Book location
Chapter 19: Artistic Theory Extension
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · DIAL target cards

Evidence boundary. The admitted PB2.2 package is a controlled reusable operator in the declared testbed, not an autonomous invention comparable to a historical artistic movement.

Open the durable experiment record →

6 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_01_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_0_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_1_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_21_algebroid_path/run.pyrunner
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_22_metric_connection/run.pyrunner
  • synthetic-creativity-archive:2026-08-18_dial_can_paintbrush_2_intent/run.pyrunner
6 resolved run records
  • 2026-08-18_dial_can_paintbrush_01_registered
  • 2026-08-18_dial_can_paintbrush_0_registered
  • 2026-08-18_dial_can_paintbrush_1_registered
  • 2026-08-18_dial_can_paintbrush_21_algebroid_path
  • 2026-08-18_dial_can_paintbrush_22_metric_connection
  • 2026-08-18_dial_can_paintbrush_2_intent

Local core code · frozen run records

DILATE video, codec, and cross-video transport

Egret factorization, semantic and structural codecs, corpus routing, conflict repair, and independent test-movie transport.

Book location
Chapter 19: Artistic Theory Extension
Code packages
  • Infinitesimal Creativity registered experiment archivelocal-curated
Entry points
run.py in each packet
Result artifacts
RESULTS.md · PREREGISTRATION.md · metrics.json

Evidence boundary. Compositional technique discovery is not guaranteed to optimize compression; the two transport errors remain post-admission diagnostics.

Open the durable experiment record →

10 machine-resolved code surfaces
  • synthetic-creativity-archive:2026-08-19_dilate_codec_0_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_codec_1_structural_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_codec_2_sprite_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_codec_3_semantic_track_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_egret_0_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_egret_1_factored_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_egret_2_cross_video_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_movies_0_theory_routing_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_movies_1_conflict_repair_registered/run.pyrunner
  • synthetic-creativity-archive:2026-08-19_dilate_movies_2_independent_transport_registered/run.pyrunner
10 resolved run records
  • 2026-08-19_dilate_codec_0_registered
  • 2026-08-19_dilate_codec_1_structural_registered
  • 2026-08-19_dilate_codec_2_sprite_registered
  • 2026-08-19_dilate_codec_3_semantic_track_registered
  • 2026-08-19_dilate_egret_0_registered
  • 2026-08-19_dilate_egret_1_factored_registered
  • 2026-08-19_dilate_egret_2_cross_video_registered
  • 2026-08-19_dilate_movies_0_theory_routing_registered
  • 2026-08-19_dilate_movies_1_conflict_repair_registered
  • 2026-08-19_dilate_movies_2_independent_transport_registered

Book 04

Discovering Compositional Worlds from Interaction

0 documented experiment families

← Return to the three-book sequenceInspect claim status and evidence →