LEARN: 24-course guided journey, open exploration, integrity harness and Agent lenses

[BryanHarrisScripts]

Goal

Reorganize PlotPickle LEARN into a human-scale 3-year / 6-semester / 24-course learning journey without deleting, rewriting, condensing, hiding, or gatekeeping any canonical curriculum.

The default experience should feel like a coherent degree-style progression while the learner remains free to browse, jump ahead, revisit, or complete material in any order.

This issue also defines the next LEARN Agent contract: the Human may use / to switch among qualified Agents that answer through different skills and input/output contracts so the learner can deliberately seek multiple informed perspectives.

Product principle

Guided, never gatekept.

PlotPickle should provide structure, orientation and recommended progression, but access to knowledge must remain open.

• Curriculum supplies the knowledge.
• The Journey supplies a recommended sequence.
• The Navigator supplies orientation.
• The integrity harness protects completeness.
• Agents supply different expert lenses.
• The Human decides what to learn, when to learn it, and which advice to use.

Existing curriculum authority to preserve

The current canonical LEARN archive remains authoritative:

• 12 curriculum topics;
• 81 archived lessons;
• 95 bundled teaching/source documents;
• 88 presentation lessons after the existing Foundations presentation expansion.

Do not rewrite or migrate this knowledge merely to create the new journey. The new program is an orchestration/indexing layer above the existing curriculum.

All existing source provenance, lesson content, integrated teaching material, RAG inventory and current LEARN validation boundaries remain intact unless a separate curriculum-content issue explicitly changes them.

24-course program model

Present the recommended curriculum as:

• Year 1 / Semester 1: Courses 01-04
• Year 1 / Semester 2: Courses 05-08
• Year 2 / Semester 3: Courses 09-12
• Year 2 / Semester 4: Courses 13-16
• Year 3 / Semester 5: Courses 17-20
• Year 3 / Semester 6: Courses 21-24

The default LEARN surface should show the learner four courses at a time for the current semester rather than confronting them with the entire 88-lesson presentation set at once.

Each course should be a curated grouping of existing lessons/references, not a copy of curriculum text.

Recommended sequencing should move from intuitive story literacy toward applied craft and finally toward less immediately intuitive professional/industry material. In particular, deeper Industry/business/professional knowledge should generally sit later in the journey, after the learner has created enough story material for that knowledge to be relatable and useful.

Learning progression

The exact 24 course names/mapping must be derived during implementation from the canonical lesson inventory, but the progression should follow these educational phases:

1. story foundations and basic visual/story literacy;
2. premise/theme, character, world/genre and introductory structure;
3. PlotPickle application and the 24/96 model;
4. deeper character, dialogue, scene movement, visual storytelling and drafting;
5. revision, collaboration and responsible AI/provenance;
6. advanced application, ownership, industry and professional context.

Do not force an artificial equal lesson count per course. Optimize the grouping for progressive comprehension and application while preserving all source material.

Deterministic program map

Add one canonical program-map owner outside the UI, conceptually similar to:

learn/program-map.json

The exact path/schema may follow existing repository conventions, but there must be one deterministic owner that maps:

• 24 stable course IDs;
• year and semester;
• course order;
• course title and purpose;
• lesson IDs included;
• recommended prerequisites/dependencies;
• PlotPickle application targets;
• optional estimated effort/read time;
• status/version of the mapping.

The map points to existing lessons. It does not duplicate lesson bodies or bundled source text.

Recommended sequence is advisory, not access control

Course order and prerequisites are guidance only.

The learner must be able to:

• open any semester from day one;
• open any course from day one;
• open any lesson/reference through Explore/Index from day one;
• move backward or forward freely;
• complete Course 18 before Course 06 if they choose;
• revisit completed material;
• ignore the recommended path entirely.

Do not use lock icons, disabled course rows, completion gates, or hidden curriculum as prerequisites.

A later course may display a gentle advisory such as Usually taken after Structure Fundamentals, but it remains selectable.

Progress follows the Human's actual learning history, not the recommended order.

LEARN navigation model

LEARN should eventually expose two complementary navigation modes.

Journey

The primary guided view:

LEARN -> Year -> Semester -> 4 Courses -> Lessons -> Apply in PlotPickle

The current semester should be the default focus, with progress and recommended next work visible.

Explore / All Curriculum

An unrestricted reference/index view across the complete curriculum.

Search/browse should support useful dimensions such as topic, course, lesson, concept and PlotPickle application area without changing canonical curriculum authority.

The learner should never need to abandon the Journey to prove completion; Journey and Explore are two views over the same curriculum/progress state.

Course rhythm: learn and apply

Courses should not become passive reading playlists.

Preferred repeated educational rhythm:

LEARN -> APPLY -> CHECK -> CONTINUE

Application should create or refine something meaningful in the learner's active PlotPickle project where appropriate: character decisions, world material, structural choices, 24/96 work, dialogue, draft material, revision decisions, collaboration/provenance records, etc.

The Human remains creative/canon authority. Completing an exercise or receiving Agent advice must not silently mutate PPF/canon.

Curriculum Navigator

Add/reuse a LEARN navigator that can answer, deterministically:

• Where am I in the recommended journey?
• What four courses are in this semester?
• What have I actually completed?
• What is the recommended next course/lesson?
• What will this course teach?
• Why is it placed here?
• Where will I apply it in PlotPickle?
• What material have I already completed out of sequence?

Navigator logic must be based on the program map + real progress, not Agent improvisation.

Curriculum Integrity Harness

Create or extend deterministic LEARN validation so reorganizing the experience can never silently lose knowledge.

At minimum the harness must prove:

• 81/81 archived lessons remain present;
• 95/95 bundled teaching/source documents remain present;
• 88/88 current presentation lessons remain reachable;
• 24/24 courses exist in the program map;
• every mapped course points only to valid canonical lessons;
• every canonical presentation lesson is reachable through Journey and/or explicitly classified reference coverage;
• no unintended orphaned curriculum;
• no duplicate ownership of lesson content;
• all advisory prerequisite/course references resolve;
• Explore can still reach the full canonical curriculum;
• restructuring does not alter source hashes/content unless an explicit curriculum-content change is intended.

The harness protects completeness and mapping. It does not decide subjective pedagogy by itself.

Sage role

Sage remains a teacher/guide, not the curriculum authority.

When relevant, Sage may receive bounded learner context such as:

• current year/semester/course/lesson;
• courses/lessons actually completed;
• relevant existing project application state;
• canonical retrieved curriculum for the question.

Sage may explain continuity (you learned X earlier; this extends it into Y) and may teach material out of sequence when asked.

Sage must never refuse a legitimate curriculum question because the learner has not completed a recommended prerequisite.

/ Agent switching and multiple expert opinions

The existing / Agent-selection interaction should evolve into a real LEARN/creative perspective mechanism.

Selecting another Agent must change more than tone/personality. Each eligible Agent should have an explicit skill package and input/output contract so different Agents can answer the same question from genuinely different professional lenses.

Example lenses may include:

• Sage / curriculum teacher;
• Fresh Reader / audience experience;
• Structure specialist / 24/96 architecture;
• Character specialist / want, need, contradiction, stakes, agency;
• Dialogue specialist / objective, tactics, subtext, voice;
• Producer/Industry specialist / professional/commercial communicability.

Exact registered Agents must come from existing canonical Agent configuration/roster rather than a second hard-coded Agent database.

All Agents may receive the same permitted project evidence and relevant canonical curriculum, but their skills, evaluation criteria, context shaping and response contract may differ.

Disagreement is valid and useful. PlotPickle must not force creative Agent answers to converge into one synthetic opinion.

A future /panel mode may ask several appropriate Agents independently and summarize agreements/disagreements, but /panel is optional/future unless existing Agent infrastructure makes it trivial. Do not block this issue on panel synthesis.

Agent authority boundaries

Agents are expert lenses over shared knowledge/evidence, not competing sources of truth.

• LEARN curriculum remains shared canonical knowledge.
• PPF/project evidence remains shared project truth/canon boundary.
• Agents may interpret, critique and recommend differently.
• Agents must not rewrite canonical curriculum.
• Agents must not silently mutate canon/project state.
• Agent disagreement is advisory evidence for the Human.
• Provider/model choice must continue to follow existing Agent/provider authority; do not create a new hidden provider path for LEARN.

Staged Skin V1 UI/UX requirement

This issue should be implemented in a deliberately staged way so the architecture is visible before deeper curriculum content is exposed.

Current #1915 Writer's Craft navigation already provides the first Skin V1 submenu and deliberately stops before lesson/module navigation. Preserve that compatibility boundary while extending it intentionally.

For the first implementation slice:

• add a clearly named LEARN Journey / 24-Course Program destination from the Writer's Craft navigation hierarchy;
• make the destination genuinely wired and keyboard reachable;
• render the 3-year / 6-semester / 4-course-per-semester structure as a truthful preview/navigation surface;
• show course shells/titles/status placeholders from the deterministic program map where available;
• do not yet expose/open full lesson curriculum content from these new course rows;
• do not pretend unfinished course-to-content navigation is connected;
• preserve Escape/Back navigation and Skin V1 keyboard-directory semantics;
• preserve the existing nine Writer's Craft collection names/compatibility unless the implementation explicitly proves a non-breaking migration path.

The desired first visible state is therefore:

Dashboard -> Writer's Craft -> LEARN Journey -> 6 Semesters -> 4 Course shells each

with the journey/navigation wired, but the underlying course content intentionally unavailable until the subsequent content-wiring phase.

Recommended implementation slices

Slice A — inventory + program-map design

• inventory current 81 archive / 88 presentation / 95 source coverage;
• derive a proposed 24-course grouping without changing curriculum text;
• explicitly place later professional/Industry knowledge where it makes pedagogical sense;
• record mapping rationale and advisory dependencies;
• add deterministic schema/validation.

Slice B — integrity harness

• prove complete lesson/source preservation;
• prove all course references resolve;
• prove no orphaned/duplicated knowledge ownership;
• preserve existing LEARN validation.

Slice C — visible Skin V1 journey preview

• expose the new journey through Writer's Craft submenu navigation;
• show Year/Semester and four course rows per semester;
• keyboard/shortcut/back navigation works;
• course rows remain truthful preview/unavailable content destinations at this stage;
• no lesson body/content is newly exposed from the journey yet.

Slice D — Journey + Explore navigation/content wiring

• connect course shells to their mapped canonical lessons;
• add unrestricted Explore/All Curriculum;
• keep all courses selectable regardless of progress;
• persist actual progress independently of recommended order.

Slice E — Apply/Check integration

• connect course application targets to existing PlotPickle project workflows;
• no automatic canon mutation;
• provide course-level completion/progress evidence.

Slice F — Agent-aware learning context

• pass bounded journey/progress context to Sage/eligible Agents;
• prove / switching selects genuinely different skill/input/output contracts;
• preserve existing provider/tool/canon authority.

Non-goals

• deleting or condensing curriculum to make LEARN look smaller;
• rewriting all lessons into 24 new monolithic documents;
• locking later semesters until earlier courses are complete;
• requiring Sage or any Agent to grant permission to access knowledge;
• making an Agent decide canonical course order at runtime;
• making one Agent the final creative authority;
• forcing different Agents to agree;
• exposing unfinished course lesson content during the initial Skin V1 preview slice;
• changing PPF canon authority;
• changing PlotPickle Score authority;
• adding a second Agent registry/provider framework;
• introducing another copy of curriculum source text.

Acceptance criteria

1. A developer brief captures the complete 3-year / 6-semester / 24-course design and all authority boundaries.
2. One deterministic program map defines exactly 24 courses grouped 4 per semester across six semesters.
3. Existing canonical curriculum remains intact: 81 archived lessons, 95 bundled sources and current 88 presentation lessons remain accounted for.
4. The integrity harness fails closed on lost/orphaned/invalid curriculum mapping.
5. Recommended ordering never becomes access control.
6. A learner can browse any semester/course/lesson regardless of completion once content wiring is enabled.
7. Progress records actual completion out of sequence rather than coercing a linear state machine.
8. Journey and Explore are views over the same curriculum/progress authority.
9. Industry/professional content is sequenced primarily later unless specific lesson dependencies justify an earlier placement.
10. Courses support the LEARN -> APPLY -> CHECK -> CONTINUE educational rhythm without automatic canon mutation.
11. Sage may use bounded learner/journey context but cannot gatekeep access or rewrite curriculum.
12. / Agent selection uses the canonical Agent roster and can produce materially different expert analysis via distinct skills/input-output contracts, not personality-only wrappers.
13. Agent disagreement remains visible/advisory; the Human decides which advice to use.
14. Initial Skin V1 implementation exposes a wired LEARN Journey submenu/preview showing six semesters and four course shells per semester while deliberately withholding new lesson-content navigation.
15. Initial preview truthfully marks unavailable content destinations rather than showing false connected status.
16. Existing Skin V1: restore menu indicators and simplify Dashboard navigation #1915 Writer's Craft navigation/keyboard compatibility is preserved or explicitly migrated with focused regression evidence.
17. Focused tests cover program-map integrity, no gatekeeping, Skin V1 journey navigation, curriculum preservation and Agent authority boundaries.
18. Convergence reports CONVERGED before merge.
19. PR Gate and Windows Product Gate are green on the exact same head before merge.

Required developer brief

Create and maintain:

docs/developer-briefs/<issue>-learn-24-course-open-journey.md

The brief must preserve the full design discussion, especially: 24 courses / 6 semesters / 4-course blocks, gradual human knowledge ingestion, later Industry placement, full curriculum preservation, guided-not-gated access, Journey + Explore, deterministic navigator and harness, LEARN/APPLY/CHECK/CONTINUE, Sage as teacher not authority, / Agent switching with real skill differences, Human authority, and the staged visible-but-content-unavailable Skin V1 submenu.

[BryanHarrisScripts]

Gradual build plan — keep each merge small

This issue is intentionally too broad for one PR. Build it as a sequence of bounded, independently useful slices. Do not move to the next slice until the current slice is green on focused tests and does not weaken existing LEARN/#1915 behavior.

Phase 0 — baseline inventory only

Goal: freeze what exists before reorganizing anything.

• Record the canonical 12 topics / 81 archived lessons / 95 bundled sources / 88 presentation lessons.
• Add a deterministic report for lesson/source counts and current hashes.
• Record the existing nine Writer's Craft collection rows from Skin V1: restore menu indicators and simplify Dashboard navigation #1915 as a compatibility boundary.
• No UI changes.
• No curriculum changes.
• No Agent changes.

Exit: baseline evidence is reproducible and existing LEARN validation remains green.

Phase 1 — 24-course program map only

Goal: design the journey without changing what the user sees.

• Add one canonical program-map schema/owner.
• Define exactly 24 stable course IDs grouped 4 per semester across 6 semesters.
• Map existing lesson IDs into the courses; do not copy lesson bodies.
• Record course purpose, year/semester, advisory prerequisites and intended PlotPickle application target.
• Sequence intuitive story/craft knowledge earlier and deeper Industry/professional material primarily later.
• Do not add access locks.

Exit: 24/24 courses validate; every mapped lesson ID resolves; no source/curriculum content changed.

Phase 2 — curriculum integrity harness

Goal: prove that the new organization cannot lose knowledge.

• Extend LEARN validation for complete archive/source/presentation coverage.
• Detect orphaned lessons, duplicate ownership, invalid course references and broken advisory prerequisites.
• Assert 81/81 archived lessons, 95/95 bundled sources and 88/88 presentation lessons remain accounted for.
• Assert the 24-course map is an index/orchestration layer, not another curriculum store.

Exit: deliberate deletion/mis-mapping fixtures fail closed; existing curriculum validation still passes.

Phase 3 — visible Skin V1 Journey shell only

Goal: let us see and evaluate the UX before wiring lesson content.

• Preserve the current Skin V1: restore menu indicators and simplify Dashboard navigation #1915 Writer's Craft submenu contract.
• Add a wired LEARN Journey / 24-Course Program destination beneath that hierarchy.
• Show 3 years / 6 semesters / four course shells per semester.
• Keyboard, shortcuts, Escape and Back work.
• Course rows are visibly preview/unavailable for lesson content at this stage.
• Do not expose lesson bodies yet.
• Do not show green/connected state for content destinations that are not actually wired.

Exit: the full 6 x 4 structure is visible and navigable, but curriculum content remains untouched and unavailable through the new course rows.

Phase 4 — wire one semester as the vertical slice

Goal: prove the model on four courses before wiring all 24.

• Connect Semester 1's four course shells to their mapped canonical lessons.
• Preserve unrestricted direct access to the rest of the existing curriculum through the existing path/index.
• Prove no copied content, no altered lesson text and no gatekeeping.
• Add actual progress tracking for the four courses that does not assume sequential completion.

Exit: one complete four-course semester works end-to-end and out-of-order completion is proven.

Phase 5 — expand Journey to all six semesters

Goal: scale only the already-proven pattern.

• Wire Semesters 2-6 using the same course component/navigation contract.
• Keep every course selectable regardless of prior completion.
• Recommended-next status is advisory only.
• Progress follows actual Human completion, including out-of-order work.

Exit: 24/24 courses are reachable without changing curriculum authority or introducing locks.

Phase 6 — Explore / All Curriculum

Goal: make self-directed learning first-class rather than a fallback.

• Add unrestricted Explore/All Curriculum over the same canonical lesson/progress state.
• Support useful browsing/search dimensions such as topic, course, lesson, concept and PlotPickle application area.
• Journey and Explore must never create separate completion truth.

Exit: a learner can deliberately ignore the recommended journey and still access the complete curriculum cleanly.

Phase 7 — LEARN -> APPLY -> CHECK -> CONTINUE

Goal: connect learning to PlotPickle without turning learning into automatic canon mutation.

• Add course application targets to existing PlotPickle workflows.
• Start with one or two representative course/application paths before broad rollout.
• Human explicitly accepts any project/canon change.
• CHECK records learning/application evidence, not an opaque AI score.

Exit: the educational loop works on representative courses with Human authority preserved.

Phase 8 — Sage journey awareness

Goal: improve teaching continuity only after the deterministic journey is stable.

• Give Sage bounded current year/semester/course/lesson and actual-completion context.
• Sage may reference earlier learning and teach out of order when asked.
• Sage cannot unlock, deny, reorder or rewrite curriculum.

Exit: Sage improves continuity without becoming curriculum authority.

Phase 9 — / Agent lenses

Goal: allow genuinely different expert opinions using existing Agent authority.

• Reuse the canonical Agent roster and existing / selection path.
• Define explicit skill/input/output contracts for eligible LEARN/creative Agents.
• Prove that switching Agent changes the expert lens, not merely personality wording.
• Allow disagreement; do not synthesize one mandatory creative answer.
• Preserve provider, tool, PPF/canon and Human-approval boundaries.

Exit: at least two or three qualified Agents can answer the same question through materially different skill contracts.

Phase 10 — optional panel synthesis only after Agent lenses are proven

Goal: compare perspectives without creating a new authority.

• Optional /panel can ask several Agents independently and summarize agreement/disagreement.
• Do not block LEARN: 24-course guided journey, open exploration, integrity harness and Agent lenses #1918 completion on this if it expands scope materially.

Merge discipline

Prefer one PR per phase, and split a phase again if it starts touching too many owners. No phase should simultaneously redesign curriculum data, navigation, progress persistence, Agent contracts and project application behavior.

At every phase:

• build from current main;
• run focused regression first;
• run existing LEARN/Skin V1: restore menu indicators and simplify Dashboard navigation #1915 compatibility tests;
• preserve curriculum hashes/content unless the phase explicitly requires otherwise;
• stop and fix regressions before expanding scope;
• require convergence plus exact-head PR Gate/Product Gate where the repository requires them.

The priority is safe accumulation: first map, then prove preservation, then show the shell, then wire one semester, then scale. Do not attempt the whole 24-course/Agent experience in one build.

[BryanHarrisScripts]

Phase 0 has started in PR #1919 (feat/1918-phase-0-learn-baseline).

This first PR intentionally stops at the baseline contract: current LEARN inventory/hashes, 88-presentation-lesson evidence, #1915 Writer's Craft compatibility, guided-never-gatekept invariants, deterministic validation, focused regression, convergence evidence and PR Gate wiring. It does not yet define the 24-course map or touch the UI.

The focused #1918 LEARN baseline check is already green in PR Gate. #1918 remains open for Phase 1+.

[BryanHarrisScripts]

Pipeline status restored

Phase 0 is accepted and merged.

Current sequence:

• Phase 0 — DONE / PROVEN
  • Baseline inventory, hashes, Skin V1: restore menu indicators and simplify Dashboard navigation #1915 Writer’s Craft compatibility, deterministic validation, focused regression, convergence evidence, PR Gate and Product Gate are complete.
• Phase 1 — NEXT
  • Build the deterministic 24-course program map only.
  • Define exactly 24 stable course IDs grouped 4 per semester across 6 semesters.
  • Map existing canonical lesson IDs without copying lesson bodies or changing curriculum content.
  • Record course purpose, advisory prerequisites and PlotPickle application targets.
  • Keep Industry/professional material primarily later in the journey.
  • No UI changes yet.
• Phase 2 — BLOCKED ON PHASE 1
  • Do not begin or accept Phase 2 until Phase 1 is green and merged.
  • Phase 2 must then prove 24/24 course integrity, valid mappings, no orphaned curriculum, no duplicate ownership, valid advisory prerequisites, preserved 81/95/88 coverage, and deliberate deletion/mis-mapping fixtures failing closed.

Do not skip ahead to Phase 3 UI work until both Phase 1 and Phase 2 have independently satisfied their exit criteria.

The original ten-phase gradual build plan remains authoritative for #1918. This issue should stay open until the intended follow-on phases are either completed here or explicitly split into linked issues with equivalent acceptance criteria and sequencing safeguards.