Route 01
Categorical language
What mathematical language makes composition, models, and structural promises explicit?
Categories for AGI develops the broad vocabulary; LINCS turns that vocabulary into axioms for structural learning; Infinitesimal Creativity adapts it to theory change.
- Chapter 1: Category Theory for AGIp. 23
- Chapter 2: Functors for AGIp. 31
- Chapter 3: Representable Functors and the Yoneda Lemmap. 43
- Chapter 4: Diagrams and Universal Constructionsp. 49
- Chapter 0: A Working Language of Compositionalityp. 33
- Chapter 1: Axioms of Structural Learningp. 59
- Chapter 0: A Working Language for Creative Theory Changep. 25
- Chapter 5: Algebraic Theories as Creative Targetsp. 107
- Chapter 1: A Categorical and Homotopical Toolkitp. 39
- Chapter 2: The ORACLE Programp. 65
- Chapter 3: Categorical Identification Under a Priorp. 77
Route 02
Compositional learning
How can a failure of composition become a typed signal for learning?
The sequence moves from diagrammatic backpropagation to sketches, factorization obstructions, localization, repair, and admission.
- Chapter 5: Categorical Deep Learningp. 59
- Chapter 6: Diagrammatic Backpropagationp. 69
- Chapter 8: Dynamic Compositionalityp. 95
- Chapter 2: Learning by Repairp. 75
- Chapter 3: Learning Sketches and Factorization Obstructionsp. 83
- Chapter 4: Tangent Learning Sketchesp. 91
- Chapter 8: Deep Learning in Non-Compositional Sketchesp. 133
- Chapter 9: Quotients, Localization, Repair, and Admissionp. 143
- Chapter 2: Creativity as Theory Constructionp. 53
- Chapter 10: DIAL as an Algorithmic Familyp. 175
- Chapter 5: Learning Inside Fixed Doctrinesp. 109
- Chapter 6: The UOCL Machinep. 139
- Chapter 7: Persistent Categorical Identificationp. 151
- Chapter 8: Probably Approximately Categorically Correct Learningp. 159
Route 03
Sketches and theories
How are domain theories declared, acquired, extended, and transported?
Sketches progress from presentations of compositional structure to learned declarations and finite theory extensions.
- Chapter 4: Diagrams and Universal Constructionsp. 49
- Chapter 0: A Working Language of Compositionalityp. 33
- Chapter 1: Axioms of Structural Learningp. 59
- Chapter 3: Learning Sketches and Factorization Obstructionsp. 83
- Chapter 9: Quotients, Localization, Repair, and Admissionp. 143
- Chapter 21: A Pattern Language for LINCS Systemsp. 243
- Chapter 0: A Working Language for Creative Theory Changep. 25
- Chapter 1: Initial Sketch Acquisitionp. 41
- Chapter 2: Creativity as Theory Constructionp. 53
- Chapter 5: Algebraic Theories as Creative Targetsp. 107
- Chapter 0: The Infant's Problemp. 25
- Chapter 1: A Categorical and Homotopical Toolkitp. 39
- Chapter 2: The ORACLE Programp. 65
- Chapter 5: Learning Inside Fixed Doctrinesp. 109
- Chapter 16: Persistent Structure Across a Lifetimep. 255
Route 04
Infinitesimal geometry
What becomes observable when models and theories are probed through tangent structure?
Tangent lifts, infinitesimal databases, decision structure, and double involution supply local geometry for diagnosis and controlled theory change.
- Chapter 12: Manifold Learning with Geometric Transformersp. 183
- Chapter 25: Causal Density Functionsp. 397
- Chapter 4: Tangent Learning Sketchesp. 91
- Chapter 5: Infinitesimal Causalityp. 101
- Chapter 6: Infinitesimal Categorical Databasesp. 111
- Chapter 7: Infinitesimal Decisionsp. 123
- Chapter 8: Deep Learning in Non-Compositional Sketchesp. 133
- Chapter 3: The Differential Geometry of Creativityp. 65
- Chapter 4: Skill Optimization for Creative Explorationp. 93
- Chapter 1: A Categorical and Homotopical Toolkitp. 39
- Chapter 7: Persistent Categorical Identificationp. 151
- Chapter 14: Homotopy and Tangent Repairp. 237
Route 05
Causality and intervention
How do observation, intervention, latent structure, and active experiments interact?
Categorical causal models lead to infinitesimal diagnosis, geometric causal discovery, and theory extension grounded by simulators or scientific evidence.
- Chapter 15: Adjoint Functorsp. 207
- Chapter 19: Topos Causal Modelsp. 257
- Chapter 20: Judo Calculusp. 269
- Chapter 21: Csql: Mapping Documents into Topos Causal Model Databasesp. 305
- Chapter 25: Causal Density Functionsp. 397
- Chapter 5: Infinitesimal Causalityp. 101
- Chapter 11: Geometric Causal Discoveryp. 165
- Chapter 12: Repairing Kan-Extension Structurep. 173
- Chapter 11: Causal Learning with Infinitesimal Creativityp. 197
- Chapter 16: Simulator-Grounded Theory Extensionp. 285
- Chapter 17: From Scientific Documents to Testable Theoriesp. 303
- Chapter 3: Categorical Identification Under a Priorp. 77
- Chapter 5: Learning Inside Fixed Doctrinesp. 109
- Chapter 14: Homotopy and Tangent Repairp. 237
Route 06
Decisions and reinforcement learning
How can sequential behavior be represented beyond one fixed decision formalism?
Universal decision models and infinitesimal decision structure culminate in RELIC and the DIAL–URL theory-construction testbed.
- Chapter 26: Universal Decisions with Kan Extensionsp. 409
- Chapter 27: Universal Reinforcement Learningp. 423
- Chapter 28: Deep URL with Geometric Transformersp. 427
- Chapter 7: Infinitesimal Decisionsp. 123
- Chapter 15: Infinitesimal Reinforcement Learningp. 193
- Chapter 16: Learning from Structured Preferencesp. 203
- Chapter 13: Reinforcement Learning with Infinitesimal Creativityp. 219
- Chapter 15: Mathematical Theory Constructionp. 251
- Chapter 9: Online and Persistent Universal Decision Learningp. 171
- Chapter 10: Convex and Regularized Universal Actionp. 201
- Chapter 11: Universal Bandit Decision Learningp. 211
- Chapter 12: Information Categories and Decentralized Decisionsp. 219
Route 07
Reasoning and assurance
How can reasoning steps, repairs, and system claims remain inspectable?
Agent construction, formal companions, structural reasoning, and trustworthy foundation models provide progressively stronger audit boundaries.
- Chapter 18: Building Agentic Systems using Kan Extension Transformersp. 243
- Chapter 31: Formal Verification Mapp. 445
- Chapter 32: CLIFF Companionp. 451
- Chapter 33: Code Companionp. 459
- Chapter 10: Reasoning by Structural Repairp. 153
- Chapter 19: Infinitesimal Argument Repairp. 223
- Chapter 20: Trustworthy Foundation Modelsp. 229
- Chapter 14: Composing Creative Workflowsp. 235
- Chapter 17: From Scientific Documents to Testable Theoriesp. 303
- Chapter 8: Probably Approximately Categorically Correct Learningp. 159
- Chapter 13: Agentic Safety and Universal Online Decision Learningp. 227
- Chapter 14: Homotopy and Tangent Repairp. 237
- Chapter 17: Synthesis and the Next ORACLE Theorem Ladderp. 263
Route 08
Scientific discovery
How can evidence drive the construction, revision, and testing of scientific theories?
The program separates pattern discovery from causal explanation, experimental admission, and the design of sustained research programs.
- Chapter 16: Causal Claims from Languagep. 213
- Chapter 17: Temporal Diffusion over Causal Trajectoriesp. 221
- Chapter 21: Csql: Mapping Documents into Topos Causal Model Databasesp. 305
- Chapter 22: Frontiers of Infinitesimal Learningp. 247
- Chapter 6: Landscapes of Computational Creativityp. 123
- Chapter 15: Mathematical Theory Constructionp. 251
- Chapter 16: Simulator-Grounded Theory Extensionp. 285
- Chapter 17: From Scientific Documents to Testable Theoriesp. 303
- Chapter 0: The Infant's Problemp. 25
- Chapter 5: Learning Inside Fixed Doctrinesp. 109
- Chapter 16: Persistent Structure Across a Lifetimep. 255
- Chapter 17: Synthesis and the Next ORACLE Theorem Ladderp. 263
Route 09
Visual generation and repair
How can a visual system enforce scene theories and eventually invent reusable generative operations?
ARTISTIC diagnoses and repairs declared visual constraints; DILATE poses the harder problem of admitting reusable extensions to the generative process itself.
- Chapter 18: Generative Visual Languagesp. 349
- Chapter 19: Artistic Theory Extensionp. 385
- Chapter 5: Learning Inside Fixed Doctrinesp. 109
- Chapter 16: Persistent Structure Across a Lifetimep. 255
Route 10
Doctrines and categorical identification
How can a learner discover not only a model, but the kind of compositional world in which that model lives?
The fourth volume makes the categorical prior explicit: assimilation learns within a doctrine, while accommodation may revise the doctrine itself.
- Chapter 1: Category Theory for AGIp. 23
- Chapter 4: Diagrams and Universal Constructionsp. 49
- Chapter 0: A Working Language of Compositionalityp. 33
- Chapter 1: Axioms of Structural Learningp. 59
- Chapter 3: Learning Sketches and Factorization Obstructionsp. 83
- Chapter 2: Creativity as Theory Constructionp. 53
- Chapter 3: The Differential Geometry of Creativityp. 65
- Chapter 5: Algebraic Theories as Creative Targetsp. 107
- Chapter 1: A Categorical and Homotopical Toolkitp. 39
- Chapter 2: The ORACLE Programp. 65
- Chapter 3: Categorical Identification Under a Priorp. 77
- Chapter 4: Core Knowledge and Piagetian Constructionp. 89
- Chapter 6: The UOCL Machinep. 139
- Chapter 17: Synthesis and the Next ORACLE Theorem Ladderp. 263