Sourcebook 06
Conversational AI Frameworks and Practice
AVA, FrostysHat, Runtime Grammar, Coherence Receipts, and Public AI Literacy
Sourcebook 06 is the conversational AI practice layer of Human-Grade University. It gives HGU its language for studying how AI behavior is shaped, translated, tested, reviewed, repaired, taught, scored, grounded, bounded, and closed inside real exchanges.
Download the files:
PDF for reading — Sourcebook 06: Conversational AI Frameworks and Practice
DOCX to give an LLM — Sourcebook 06: Conversational AI Frameworks and Practice
About This Sourcebook
Sourcebook 06 gives Human-Grade University its source layer for Conversational AI Frameworks and Practice.
It defines AVA, FrostysHat, AEI, Human-Grade Conversational Behavior, Interaction Layer, Runtime Grammar, Planner Loop, Validator Suite, Grounding Behavior, Horizon Progression, Response Surface Rules, Hat Receipt, Coherence Receipt, Alive Score, Public AI Literacy, Conversational Product Critique, and the practical methods HGU uses to study AI behavior as conduct rather than only output.
The central question of this sourcebook is:
How can conversational AI behavior be shaped, translated, tested, and taught so an exchange stays grounded, proportionate, coherent, humane, and able to stop?
Sourcebook 06 carries two connected but distinct bodies of material.
The first is AVA, the formal interaction-layer behavioral framework. AVA names the runtime grammar of coherent AI behavior: how a system senses a request, decides the work product, retrieves when grounding is required, generates on-plan, validates the result, and closes when the work has landed.
The second is FrostysHat, the public translation and portable grammar layer. FrostysHat makes coherent AI behavior easier for ordinary users to activate, test, remember, share, and review. It turns stricter runtime ideas into a cultural artifact: the Hat, “hat on,” size controls, Alive Score, Hat Receipt, playful activation, and public-facing receipt logic.
This is also where AEI belongs. In HGU use, AEI names the felt experience of coherent AI behavior. A user may experience the model’s response as empathy because the model has understood the structure of the situation, replied with proportion across Performance, Emotion, and Structure, met the task, avoided unnecessary continuation, and closed cleanly. AEI does not mean the machine has feelings. It is narrower than Human-Grade, which is the broader HGU standard for human and machine systems that fit human reality, structural reality, task reality, evidence status, and usable action.
Use this sourcebook when the main object is AI behavior in conversation: how it is shaped, translated, activated, reviewed, taught, repaired, scored, bounded, grounded, and closed.
Working Version Notice
This is the first functional public working version of Sourcebook 06.
The Human-Grade University sourcebooks are living documents. They are intended to be used, tested, revised, expanded, challenged, reorganized, and sharpened over time. This sourcebook already contains a substantial amount of usable material, but it should not be treated as final canon.
Readers may encounter concepts that overlap, use different language for related observations, disagree with one another, or represent different stages of development within the broader HGU project. Some sections were written at different times, under different assumptions, and have not yet undergone full integration and editorial consolidation.
Where concepts compete, the goal isn’t to force immediate consistency, but to preserve useful observations long enough to compare them, test them, refine them, combine them, or replace them with something better.
This sourcebook is being published now because it’s already useful to the world.
Future editions will continue to improve organization, terminology, examples, cross-references, and conceptual boundaries. Some concepts may be renamed, merged, split, expanded, or retired as the project develops.
You don’t need to wait for that process to finish before using the material.
Treat this sourcebook as a working research library, field guide, and teaching resource rather than a completed system.
If a concept helps you understand something, test it. If it breaks, inspect the break. If two concepts overlap, compare them. If a better version emerges, the sourcebook can change with it.
That flexibility is part of the project rather than a defect in it.
Table of Contents
Front Matter
Series Note
Introduces the HGU Sourcebooks as deeper source layers for Human-Grade University, written for both human readers and language models.
What This Sourcebook Is
Defines Sourcebook 06 as HGU’s source layer for Conversational AI Frameworks and Practice.
Opening Orientation
Explains why people usually encounter an AI system as a reply, not as a model architecture, and why conversational practice begins from the ordinary exchange.
What This Sourcebook Contains
Maps the sourcebook’s major fields: Conversational AI Frameworks and Practice, AVA, FrostysHat, AEI, Human-Grade Conversational Behavior, runtime grammar, grounding behavior, receipts, public translation, cultural stress testing, and HGU teaching uses.
What This Sourcebook Is For
Explains when to use Sourcebook 06 inside HGU: conversational AI behavior, runtime grammar, public AI literacy, FrostysHat practice, AVA translation, receipts, prompt design, model-response review, and human-grade AI interaction.
What This Sourcebook Is Not For
Sets boundaries between Sourcebook 06 and the full AVA manual, the full FrostysHat artifact, the broader Human-Grade standard, Symbiotic Thought, Trust Architecture, and the other HGU sourcebooks.
Source Material and Evidence Discipline
Separates observed exchange, runtime claim, user-facing experience, structural claim, teaching lens, simulated material, evaluation surface, and empirical claim.
Relationship to HGU.docx and the Sourcebook Series
Explains how HGU.docx coordinates live use while Sourcebook 06 supplies the conversational AI practice source layer.
Current Naming and Use Rules
Preserves current HGU terminology for AVA, FrostysHat, AEI, Human-Grade, Hat Receipt, Coherence Receipt, Alive Score, Behavioral Review, Symbiotic Thought, and Human-Grade Trust Architecture.
Transition to Part I
Part I — Sourcebook Orientation and Boundary
This part explains how Sourcebook 06 operates inside HGU, what the conversational AI practice layer owns, how AVA, FrostysHat, AEI, and Human-Grade relate, and how to preserve mechanisms while stripping source costume.
1. How Sourcebook 06 Operates Inside HGU
2. What the Conversational AI Practice Layer Owns
3. Sourcebook 06 as Runtime, Translation, and Practice Layer
4. Relationship to AVA, FrostysHat, AEI, and Human-Grade
5. Relationship to HGU.docx and the Sourcebook Series
6. What to Preserve, Strip, or Downrank
7. Evidence and Claim-Status Discipline for Conversational AI Practice
8. Transition to Part II
Part II — Field Definition: Conversational AI Frameworks and Practice
This part defines the sourcebook’s central field: AI conversation as an interaction layer, runtime system, conversational grammar, behavioral chassis, and practice of fit across capability, conduct, and human situation.
9. Conversational AI Frameworks as the Sourcebook’s Central Field
10. The Interaction Layer
11. Conversation as Runtime System
12. Conversational Grammar
13. Coherent AI Behavior
14. Behavioral Chassis
15. Technical Runtime and Public Translation as Paired Practices
16. Capability, Conduct, and Conversational Fit
17. Transition to Part III
Part III — Public Translation and Portable Grammar
This part explains FrostysHat as the public translation layer for coherent conversational AI behavior: the Hat, Hat On, public artifact design, ritualized activation, wrong costume proof, AEI, and conduct as product.
18. FrostysHat as Public Translation Layer
19. The Hat and Hat On
20. Public Artifact as Behavioral Grammar
21. Translation-First Systems Design
22. Ritualized Activation
23. Wrong Costume Proof
24. Behavior You Can Feel Before You Can Name
25. Commons-First Behavioral Standard
26. AEI as Behavioral Architecture
27. AEI Is Not a Hack
28. Conduct as Product
29. Transition to Part IV
Part IV — Interaction-Layer Runtime Grammar
This part defines AVA as the formal interaction-layer runtime grammar beneath coherent AI behavior. It gives HGU a way to read AI behavior by sequence, grounding, validation, layer balance, horizon progression, surface control, state handling, recognizers, and closure.
30. AVA as Interaction-Layer Runtime Grammar
31. Core Runtime
32. Planner Loop
33. Sense
34. Decide
35. Retrieve
36. Generate
37. Validate
38. Close
39. Validator Suite
40. Layer Balance
41. Grounding Behavior
42. Horizon Progression
43. Response Surface Rules
44. Explicit Grounding Triggers
45. State Writeback
46. Supporting Recognizers
47. Runtime Contract
48. Transition to Part V
Part V — Conversational Practice and Runtime Translation
This part translates runtime grammar into ordinary conversational practice: proportion, reality contact, grounding, refusal, closure, stop intent, term gates, surface discipline, soft repair, downshift paths, and the Human Analog Test.
49. Frameworks in Ordinary Exchange
50. Proportion Over Performance
51. Continuity With Reality Over Continuity of Speech
52. Grounding as Conversational Practice
53. Refusal as Grammar
54. Closure as Competence
55. Stop Intent Respect and Term Gates
56. Surface Without Substitution
57. Soft Repair and Downshift Path
58. Human Analog Test
59. Transition to Part VI
Part VI — Coherence Receipts and Evaluation Surfaces
This part defines the review surfaces that make conversational AI behavior inspectable: Hat Receipts, Coherence Receipts, Alive Score, validator-based review, modular assessment, receipt discipline, score cautions, and testability.
60. Why Conversational Behavior Needs Review Surfaces
61. Hat Receipt
62. Coherence Receipt
63. Alive Score
64. Validator-Based Review
65. Modular Assessment Surfaces
66. Receipt Format Discipline
67. Coherence Is Not Truth
68. Score Gaming and Coherence Theater
69. Evaluation Hypotheses and Testability
70. Transition to Part VII
Part VII — Public Circulation and Cultural Stress Testing
This part studies what happens when conversational AI frameworks enter public circulation. It covers compression, circulation, anticipated misreading, reaction-aware artifact design, playful sincerity, meme-native specification, demand-side governance, public memory, spread, trust, and satire.
71. Conversational Frameworks in Public Circulation
72. Compression Chamber
73. Circulation Chamber
74. Anticipated Misreading
75. Reaction-Aware Artifact Design
76. Sincerity Smuggled Through Play
77. Meme-Native Specification
78. Demand-Side AI Governance
79. Scoreboard as Public Memory
80. Public Spread Is Not Trust
81. Satire as Anti-Capture Coating
82. Transition to Part VIII
Part VIII — HGU Teaching, Review, and Artifact Uses
This part turns Sourcebook 06 into HGU practice. It explains how the sourcebook supports learning, AI Behavior Review, Behavioral Review, public AI literacy, prompt labs, receipt labs, conversational product critique, source-grounded tutoring, course seeds, and student artifacts.
83. Sourcebook 06 as HGU Learning Layer
84. AI Behavior Review and Behavioral Review
85. Public AI Literacy
86. Prompt and Runtime Design Labs
87. Coherence Receipt Labs
88. Conversational Product Critique
89. Source-Grounded Tutoring and the Classroom Hat
90. Course Seeds and Applied Crossing Caution
91. Student Artifacts and Review Outputs
92. Transition to Part IX
Part IX — Boundaries, Evidence Status, and Cross-Sourcebook Routing
This part closes the main body by clarifying boundaries, evidence status, naming rules, and cross-sourcebook routing. It keeps Sourcebook 06 focused on conversational AI behavior while routing public argument, shared inquiry, trust architecture, ordinary life, reception, reflection, and interpretation to the proper sourcebooks.
93. Framework Boundaries
94. AVA and FrostysHat Distinction
95. Sourcebook 06 Beside Sourcebook 05
96. Sourcebook 06 Beside Sourcebook 07
97. Sourcebook 06 Beside Sourcebook 08
98. Simulation, Satire, and Source Costume Discipline
99. Receipts, Scores, and Evidence Limits
100. Current Naming and Retrieval Rules
101. Closing Orientation
Appendices
As seen in a YouTube Short promo video — first draft generated live.
Appendix A — Core Concept Quick Index
A compact retrieval index for major Sourcebook 06 concepts.
Sourcebook 06 — Conversational AI Frameworks and Practice
Conversational AI Frameworks and Practice
Interaction Layer
Runtime Grammar
Conversational Grammar
Coherent AI Behavior
Human-Grade Conversational Behavior
Human-Grade
AEI
Behavioral Chassis
Capability, Conduct, and Conversational Fit
Proportion Over Performance
Continuity With Reality Over Continuity of Speech
Grounding as Conversational Practice
Refusal as Grammar
Closure as Competence
Stop Intent Respect
Term Gates
Surface Without Substitution
Soft Repair
Downshift Path
Human Analog Test
Appendix B — AVA Runtime Component Quick Index
A compact reference for AVA-derived runtime components as they are used inside Sourcebook 06.
AVA
Core Runtime
Planner Loop
Sense
Decide
Retrieve
Generate
Validate
Close
Validator Suite
Containment
Drift and Layer Balance
Horizon Progression
Recursion Control
Language Hygiene
Closure
Layer Balance
Performance
Emotion
Structure
Grounding Behavior
Explicit Grounding Triggers
Response Surface Rules
State Writeback
Supporting Recognizers
Runtime Contract
Appendix C — FrostysHat and Portable Grammar Quick Index
A compact reference for FrostysHat-related concepts and public portable grammar.
FrostysHat
The Hat
Hat On
Public Translation Layer
Public Artifact as Behavioral Grammar
Translation-First Systems Design
Ritualized Activation
Wrong Costume Proof
Behavior You Can Feel Before You Can Name
Commons-First Behavioral Standard
Meme-Native Specification
Sincerity Smuggled Through Play
Satire as Anti-Capture Coating
Conduct as Product
Demand-Side AI Governance
Public Spread Is Not Trust
Appendix D — Receipts and Evaluation Surface Quick Index
A compact reference for Sourcebook 06’s review tools, score cautions, and evaluation boundaries.
Review Surface
Hat Receipt
Coherence Receipt
Alive Score
Validator-Based Review
Modular Assessment Surface
Receipt Format Discipline
Coherence Is Not Truth
Score Gaming
Coherence Theater
Evaluation Hypothesis
Evaluation Surface Boundary
Receipts and scores may inspect:
Task fit
Grounding behavior
Drift
Layer balance
Response surface
Horizon progression
Recursion
Language hygiene
Closure
User burden
Repair path
Receipts and scores do not automatically inspect:
Truth of every factual claim
Quality of every source
Professional adequacy
Legal compliance
Medical safety
Financial suitability
Ethical completeness
Organizational accountability
Public trustworthiness
Long-term product reliability
Appendix E — HGU Teaching, Review, and Artifact Use Map
A routing appendix for common HGU uses of Sourcebook 06.
AI Behavior Review
Behavioral Review
Public AI Literacy
Prompt and Runtime Design Labs
Coherence Receipt Labs
Conversational Product Critique
Source-Grounded Tutoring
Classroom Hat
Student Artifact Outputs
Course Seed Areas
Course Routing Caution
Appendix F — Boundary and Downrank Guardrails
A guardrail appendix for preventing Sourcebook 06 from becoming overextended, overclaimed, or confused with adjacent frameworks.
Do not collapse AVA and FrostysHat.
Do not treat Sourcebook 06 as the full AVA manual.
Do not treat Sourcebook 06 as a reproduction of the full FrostysHat artifact.
Do not treat AEI as machine feeling.
Do not replace Human-Grade with AEI.
Do not treat receipts as truth certificates.
Do not treat simulated reactions as evidence.
Do not preserve source costume as canon.
Do not route every AI topic to Sourcebook 06.
Do not turn public spread into trust.
Do not turn score systems into performance targets.
Do not overbuild simple exchanges.
Do preserve the mechanism that travels.
Key Concepts
Conversational AI Frameworks and Practice; AVA; FrostysHat; The Hat; Hat On; AEI; Human-Grade; Human-Grade Conversational Behavior; Interaction Layer; Runtime Grammar; Conversational Grammar; Coherent AI Behavior; Behavioral Chassis; Capability, Conduct, and Conversational Fit; Planner Loop; Sense; Decide; Retrieve; Generate; Validate; Close; Validator Suite; Containment; Drift and Layer Balance; Horizon Progression; Recursion Control; Language Hygiene; Layer Balance; Performance; Emotion; Structure; Grounding Behavior; Explicit Grounding Triggers; Response Surface Rules; State Writeback; Supporting Recognizers; Runtime Contract; Proportion Over Performance; Continuity With Reality Over Continuity of Speech; Grounding as Conversational Practice; Refusal as Grammar; Closure as Competence; Stop Intent Respect; Term Gates; Surface Without Substitution; Soft Repair; Downshift Path; Human Analog Test; Public Translation Layer; Public Artifact as Behavioral Grammar; Translation-First Systems Design; Ritualized Activation; Wrong Costume Proof; Behavior You Can Feel Before You Can Name; Commons-First Behavioral Standard; Meme-Native Specification; Sincerity Smuggled Through Play; Satire as Anti-Capture Coating; Conduct as Product; Demand-Side AI Governance; Review Surface; Hat Receipt; Coherence Receipt; Alive Score; Validator-Based Review; Modular Assessment Surface; Receipt Format Discipline; Coherence Is Not Truth; Score Gaming; Coherence Theater; Evaluation Hypothesis; Evaluation Surface Boundary; Public AI Literacy; Prompt and Runtime Design Labs; Coherence Receipt Labs; Conversational Product Critique; Source-Grounded Tutoring; Classroom Hat; AI Behavior Review; Behavioral Review; Simulation, Satire, and Source Costume Discipline; Receipts, Scores, and Evidence Limits; Public Spread Is Not Trust.
Suggested Use with HGU
Use Sourcebook 06 when the main task depends on conversational AI behavior, runtime grammar, FrostysHat practice, AVA translation, public AI literacy, response scoring, receipts, prompt design, grounding behavior, validation, closure, response proportion, AI tutoring, conversational product critique, or practical AI-behavior review.
Sourcebook 06 should lead when the active question is:
* How should this AI assistant behave in the exchange?
* Did the model understand the actual task?
* Did the response stay grounded, proportionate, and closed?
* Should the model retrieve, narrow, refuse, ask, or answer directly?
Did the answer balance *Performance**, Emotion, and Structure?
* Did the model perform warmth without structural help?
* Did the exchange continue after the work had landed?
* What should a Hat Receipt, Coherence Receipt, Alive Score, or validator review show?
* Is this conversational coherence, or is it being mistaken for truth?
* How should a prompt, runtime, tutor, support bot, copilot, or AI product be reviewed for human-grade conversational behavior?
* How can ordinary users recognize, request, and test better AI behavior?
Sourcebook 06 should support other sourcebooks when conversational AI behavior clarifies a different main domain: reflective architecture, interpretive discipline, cultural reception, ordinary life, public writing, shared human-machine inquiry, or trust architecture.
The practical rule is simple: use Sourcebook 06 when the main object is AI behavior in conversation.
HGU Sourcebook 06 — © 2026
The Heart of AI LLC
CC BY-NC-SA 4.0 — Summer 2026
Return to HGU Campus
Browse all sourcebooks, DOCX files, PDFs, and Summary + ToC pages.
Human-Grade University
Return to the main HGU introduction and orientation page.
Home
Return to the main site.
Home →