canon/WorkingNotes.txt
Start Humanity Ch 02-15 Cascade — Phase 1 Session A (Ch 02-03 Flores). Verify memory entries 7 (cascade plan), 8 (session types), 9 (staging docs), 10 (execution defaults) are loaded. If missing, stop and ask. Per CASCADE EXECUTION STANDARDS in Project 1 instructions, produce: 1. ChapterDesign_Ch_02-15_AncientEarly_v2_PHASE1.md — Ch 02 + Ch 03 only 2. ResearchBasis_Ch_02-15_AncientEarly_v2_PHASE1.md — Ch 02 + Ch 03 only Both open with `# PHASE 1 — Ch 02-03 FLORES` header. Pull from existing Ch 02-15 ChapterDesign and ResearchBasis docs in Project 1. Refine to strict TAP Section 4.1/4.2 ownership. Route overflow to Sessions B and C — flag what's collected for each. End response with the standard handoff: next-message prompt + new-session prompt. Begin. what's my full book list split by chapter for chapters 16-25, please include where to purchase them or download them. preference is for PDF versions so I can copy and past easier. Also we don't need ones that you're already capable of reading fully without me getting for you. purpose of the book list and book processing is so we can fill out those region building pages so that the region build page can be used for projects 4A, 4B, 5, 6, & 7 [NOT IN PROJECT] The 2026 ARPG list you pasted. Only Elden Ring appears in the canonical comparables (GDD_Core Section 1). Path of Exile 2, Blood of the Dawn Walker, Mortal Shell 2, Lords of the Fallen 2, Code Vein II, and Alabaster Dawn are not referenced in any connected document. The Architecture Expansion Reference lists its own design-lesson sources: Super Mario 64, GoldenEye 007, Pokémon R/B/Y, RDR2, Witcher 3, Sonic Adventure 2, Super Monkey Ball, Mario Party, CoD, Halo, Battlefield, C&C, StarCraft, WoW, ARC Raiders, Rust, Minecraft. If you want to add any of the 2026 ARPGs as design references, that's a new Project 1 decision — not something currently documented. UX/UI flow documentation. I haven't seen wireframes, HUD spec, or dialogue UI flow anywhere in the docs I've searched. This becomes critical at UE5 Phase 2 when you build GA_DialogueOption and the voice pipeline — without UX docs, the implementation will bake in decisions that should be design-level. Before book processing, a 10-15 page UX doc should exist covering: main HUD layout, dialogue UI (6 options + voice indicators), inventory/equipment screen (16 slots + vril-seat visualization), map/Vimana UI, journal UI (healer's child's journal), world-state indicator, integrity indicator, familiar slot UI. Accessibility spec. 2026 GDD templates universally include accessibility. The only accessibility-adjacent content I've found in Humanity is the six-option dialogue system (which is good) and the voice-acted protagonist (also good). Missing: colorblind support for integrity/vril visual indicators, subtitle spec, difficulty customization separate from integrity path, motor accessibility for the combat system. Not urgent, but should exist as a stub document before any combat UE5 work begins. Pre-production prototype milestones. Industry practice is to validate the core loop in prototype form before production begins. Your Year 1 milestones (Y1-A through Y1-J) do this correctly for the Ch 2 pipeline proof. Good. But the one critical playable-proof you don't yet have scheduled is the full three-path ending reveal — the Ch 76-77 waterfall Grand Sage moment. Because this is the game's thematic payload and it must land, a Year-2 or Year-3 standalone prototype of just that scene would be worth building even before the chapters between it and Ch 2 exist. Finding out after Year 8 that the reveal doesn't land as designed would be catastrophic. Your Production Playbook already specifies Cursor + Claude Code + Claude Desktop MCP, which aligns perfectly with 2026 best practice. What isn't yet in your playbook but should be: Recommendation A — Add a .claudeignore file to your UE5 project StraySpark's guide explicitly calls this out as table-stakes for game projects: *"Use .claudeignore — similar to .gitignore, this file tells Claude Code which directories and files to skip when scanning your project. For Unreal projects, you almost always want to ignore: Binaries/, Intermediate/, Saved/, DerivedDataCache/, .vs/, *.uasset, .umap. This prevents Claude from trying to read binary files and generated content that would waste context and confuse analysis." This saves massive context and prevents the "Claude spent 40+ messages trying to figure out which classes affected X" failure pattern documented on the Epic forums. Should be in your Phase 1 Foundation setup in the Production Playbook. Recommendation B — Codebase map MCP The Epic forum thread from 3 weeks ago describes an open-source MCP tool specifically built to solve the problem of "Claude reading UE5 files one at a time, can't follow .uasset or Blueprint references, and eventually hallucinates a dependency." Worth evaluating when you get to Phase 2. Search: "unreal codebase map MCP open source" on the Epic forums. Recommendation C — The "Teach Claude to see inside Unreal" pattern The dev.to article from colton_willey (March 2026) describes the exact workflow you're going to want: build a custom MCP plugin that exposes specific Unreal operations Claude can call directly, then iterate — every gap Claude can't handle, add a new tool. Their specific examples: writing blueprints, animation montages, anim graphs, blendspaces. The key insight: "every gap I encounter I immediately cover." For Humanity specifically, the operations worth exposing early are: Data Table row operations (since your architecture is "chapters are data entries only"), quest generation templates, dialogue-data population, and PCG biome creation. Your Production Playbook's Phase 2 through Phase 4 already assume this kind of tooling exists — the explicit MCP development work should be on the Phase 1.5 or 2 checklist. Static vs. agentic AI in solo RPG experiences (OpenAI Developer Community, 2025) describes multi-agent ReAct architectures for procedural narrative. Worth reading — relevant to your infinite side quest system's GPU-local generation architecture. Solo RPG Revolution (Basunat 2nd Edition, 2025) covers hybrid approaches combining human creative direction with AI procedural generation. Your current AI agent team structure (Architect, Engineer, World Builder, Music, Asset, QA) is exactly this hybrid model, and your creative-director-as-sole-decision-maker rule matches current best practice. What's genuinely new in 2026 worth evaluating Four tools/patterns from the 2026 research that aren't in your current stack: AI-driven playtesting (Juego Studios' 2026 guide): automated test suites that play through chapters to identify bugs, unreachable content, and pacing issues. For a 77-chapter game, this is nearly mandatory for QA at your scale. Unreal Engine 5.5+ has built-in support; the open-source tooling is maturing fast. Put on Years 6-7 milestones. Memory-first AI for NPCs (Jenova 2026): "Players will expect NPCs to remember their play style, their previous choices, and even their tone of voice across multiple sessions." Your faction tracking system already does this at the regional level. The 2026 expectation is that individual NPCs do it too — the babalawo who recognized the child in Ch 12 should remember them when they're revisited in Ch 39. Your infinite side quest system architecture supports this, but the per-NPC memory spec isn't explicitly documented anywhere I searched. Worth writing up as a section in GDD_Systems before UE5 Phase 3. Worldbuilding consistency tools (Chronicler, Feb 2026): plain-text offline Markdown wikis with auto-generated broken-link reports, wiki-style infoboxes, and cross-embedding. For a 20+ document project like Humanity, a Chronicler or Obsidian vault with link consistency reports would catch the kind of drift we've been finding in these sweeps automatically. Obsidian's Dataview plugin can query "all references to Chapter X" or "all familiar names" and surface inconsistencies. Google Docs and Notion don't do this. Worth piloting even just as a consistency-check mirror of your Drive structure. Can wait until after book processing but should happen before UE5 Phase 2 Write the UX/UI document. 10-15 pages covering HUD, dialogue UI, inventory/equipment, map/Vimana, journal, world-state and integrity indicators, familiar slot UI. Do this before GA_DialogueOption is implemented. Write a Risk Register. Six to ten high-impact risks with likelihood/impact/mitigation. Reviewed monthly in the Project 1 review alongside consistency. Stub an Accessibility doc. Even a one-pager. Colorblind spec, subtitle spec, motor accessibility considerations, separating difficulty from integrity path. Add .claudeignore to the UE5 project spec in the Production Playbook. Before Phase 1 begins. Ten minutes. Document the NPC memory spec. How does an individual NPC (the Ch 12 babalawo, the healer's child, the Mesopotamian Architect's successive manifestations) remember the player across chapters and across decades of in-game time? This is a genuine design gap that will cause implementation pain if left until Phase 3. Larger strategic recommendations (Year 2-3) Build the Ch 76-77 waterfall prototype. Not a full chapter — a 15-minute standalone Unreal scene with final-quality voice and visual effects that proves the Grand Sage reveal lands emotionally. This is the single highest-risk creative moment in the entire game, and the whole project's thematic weight depends on it working. Prototyping it early — with placeholder everything around it — is cheap insurance against discovering at Year 8 that the reveal feels flat. Evaluate Gemini CLI 1M-context for monthly cross-doc consistency sweeps. Your Production Playbook already specifies this agent role; actually using it against the full docset each month as an automated pass would catch the kinds of drifts we've been fixing manually.