Golden Nuggets

# Golden Nuggets — Platform Vision Notes

Captured from the long-form architecture discussion so the highlights don’t get lost while experimenting.

## 1. Sessions as Universal Work Containers
- Sessions represent time-boxed focus, not just calendar events.
- Multiple activities per session (`reading`, `analysis`, `translation`, etc.).
- Focus modes (`deep-work`, `exploration`, `collaboration`, `tidy-up`).
- Sessions chain/branch; memory accumulates (decisions, insights, questions).

## 2. Grid Composer
- Structured canvas (CSS grid) with reusable cell types.
- Cells can host signals, workflows, AI panels, resource lists, block editors.
- Connections between cells (data-flow/reference/sequence).
- Grid + session are inseparable (session = grid layout + memory).

## 3. Universal Blocks
- Blocks can reference any node (signal/resource/session/task).
- Slash commands, `@` mentions, `[[` backlinks, live queries, interactive canvases.
- Blocks mark AI-generated content explicitly.
- Progressive enhancement: start with signals, expand to compositions/notes.

## 4. Node/Edge Everywhere
- Everything becomes a node with flexible props.
- Edges are first-class (strength, confidence, metadata, temporal validity).
- Allows merging of Convex + Supabase data under a single mental model.

## 5. Progressive Disclosure UX
- Level 1: Simple buttons (New note, New session).
- Level 2: Power gestures (drag PDF → extract, `/` commands).
- Level 3: Pro actions (Cmd+K, batch ops, graph views).
- Users opt in/out; classic navigation stays intact.

## 6. Magic Moments (per persona)
- Student: “Show what matters this week.”
- Facilitator: “One click spins up week + sessions + assignments.”
- Analyst: “Drop file → get glossary + PRD + tasks.”
- Each persona gets a clear “wow” action as soon as Phase 0 is stable.

## 7. Context Engine
- Derive context from session, grid, recent nodes, active queries.
- Provide context to UI + AI without forcing features.
- AI context includes session memory, grid state, intentions, glossaries.

## 8. Hybrid Storage
- Supabase: auth, file storage, high-volume domain tables.
- Convex: node graph, session data, real-time interactions.
- Local seed: fallback + fixture generation.
- Persistence adapter pattern keeps UI decoupled.

## 9. Visibility & Ownership Rules
- Child visibility <= parent visibility (tighten allowed, widen denied).
- Sections own sessions/resources/assignments/discussions.
- Collections reference without changing structural parents.
- Sessions prefer Section parent but can fall back to “Unassigned”.

## 10. Feature Rollout Strategy
- Feature flags (server-driven truth, client mirrors for UI).
- Dual-mode navigation (classic + session) with opt-in prompts.
- Dual-write during migration; virtual sessions for legacy content.
- Migration dashboard tracks data → session → blocks progress.

## 11. AI Philosophy
- AI suggests, human decides; everything is badged.
- Start with simple actions (“Suggest title”, “Summarize session”).
- Context-aware AI emerges once session memory + grid exist.

## 12. Implementation Order
1. Invisible foundation (session layer, flags, adapters).
2. Optional enhancements (session rooms, universal blocks in beta).
3. Gentle introduction (context hints, “continue session” prompts).
4. Gradual migration (power users first, classic mode always available).

---

Use this document as the source of truth when spinning new beta experiments or turning ideas into implementation prompts. If an idea lands in a beta page, map it back here (or update this file) so we can retire beta code without losing the insight.