Discovery Sequence View - Outside In

DomainsDOL EnglishProduct1.603 words8 min read

active

Updated: 2026-03-27 Status: active

Why this view exists

Keep discovery aligned with macro user journey first.
Prevent diving into deep internal modules before entry flow is stable.
Ensure later module decisions are anchored to platform-level routing logic.

Discovery order (locked)

Stage A - Entry and Access Foundation
- Scope: public entry points, login/register gateway, auth return path.
- Outputs: agreed entry intents, auth boundary, no-dead-end rules.
Stage B - Onboarding Intent and Goal Capture
- Scope: optional goal setup, skip behavior, deferred completion nudges.
- Outputs: goal-state model and missing-goal fallback behavior.
Stage C - Home Router and Mode Switching
- Scope: post-login root resolution, program-first shell entry, and scoped aggregate exceptions.
- Outputs: routing contract, restore/fallback ladder, module entry priorities.
Stage D - Learning Core (Self-study)
- Scope: bank -> attempt -> result -> next recommendation.
- Outputs: recommendation logic, scoring interpretation, progression evidence.
Stage E - Course and Operations Surfaces
- Scope: teacher/CS/academic/admin roles for course lifecycle only.
- Outputs: operational boundary, permissions, handoff logic.
Stage F - Cross-cutting Growth and Retention
- Scope: streak/reactivation/motivation, KPI and analytics.
- Outputs: measurement framework and lifecycle interventions.
Stage G - Vocabulary Learning Loop Hardening
- Scope: vocabulary default entry, daily queue behavior, overload protection.
- Outputs: lean daily loop with deterministic queue and intake guardrails.
Stage H - AI Tutor Core Logic and Guidance Integrity
- Scope: AI Tutor role baseline, grounded knowledge contract, active-attempt integrity, output artifact taxonomy.
- Outputs: decision-backed AI Tutor state doc + UX extraction targets for AIT lane.
Stage I - Smart Search AI Inline Intelligence Surface
- Scope: AI inline quick-insight/catalog UX, boundary with AI Tutor, and escalation handoff contract.
- Outputs: decision-backed inline boundary + export contract + Smart Search UX extraction targets.

Current stage

Active mode: cross-stage hardening and consistency pass (no unresolved stage gate blockers).
Stage A is closed at backbone + edge-rule level (DEC-0011).
Stage A2 is opened to finalize authentication detail, profile model baseline, package taxonomy, and entitlement boundaries.
Stage A2 baseline model (single-user profile + Free/Pro/Pro Max + course a la carte boundary) is captured in DEC-0032.
Stage A2 auth-method and tier-entitlement baseline are captured in DEC-0033 and refined by DEC-0044/DEC-0048.
Stage A2 paywall/checkout/post-payment return policy is captured in DEC-0034.
Stage A2 unified result-page lock behavior and upgrade entry policy are captured in DEC-0035.
Stage A2 feature-based upgrade offer-set and fast-path policy are captured in DEC-0036.
Stage A2/D monetization and AI entitlement experience baseline is further locked in DEC-0067.
Stage A2 payment cycle/method overlap lifecycle is refined in DEC-0047.
Stage A2 payment cycle taxonomy is refined in DEC-0081 (12-month moves to annual auto-renew, one-time focused on 3/6).
Stage A2 course-vs-subscription entitlement and course-credit model are refined in DEC-0046.
Stage A2 account security/recovery/deletion lifecycle is refined in DEC-0050.
Stage A2 age-gate, duplicate-account warning UX, and auth precision are closed by DEC-0056/DEC-0064.
Stage A now also has outside-in low-friction baseline refinement captured in DEC-0043 (pre-auth browse + auth gate at attempt/persist + minimal signup + home-first post-signup).
Stage B is closed with contextual goal nudge policy and daily cap captured in DEC-0041 (superseding fixed weekly popup policy from DEC-0006).
Stage B goal modeling quality is further locked by DEC-0071 (program-specific goal scale and assessment-form-aware target input).
Stage B execution details are locked by DEC-0072 (goal depth by skill, primary/secondary structure, and goal versioning).
Stage B fatigue handling is locked by DEC-0074 (soft-snooze after repeated dismiss/ignore to reduce popup pressure while keeping goal optional).
Stage C is closed with home defaults and edge policies captured in DEC-0012 and refined by DEC-0049/DEC-0055/DEC-0065.
Stage C shell topology is now locked by DEC-0096:
- no standalone post-login overview shell,
- Program Workspace is the default logged-in shell reality,
- cross-program aggregation survives only as explicit scoped views,
- deep objects require canonical ownership and must not drift into pseudo-program routing.
Stage C also has non-mandatory-goal alignment captured in DEC-0040.
Stage C/M first-week activation guidance is locked by DEC-0079 (Home inline ladder for new-user first 7 days, non-blocking and no-popup).
Stage C/M activation relevance hardening is locked by DEC-0080 (program-adaptive starter resolution + continuity + deterministic fallback).
Stage C still keeps detailed IA extraction follow-up open for:
- aggregate-eligible view list beyond schedule / calendar,
- Từ vựng top-level prominence validation,
- mobile shell compression,
- canonical object ownership readiness in data/model layers,
- persistence safety for last meaningful program workspace.
Latest Stage C review also surfaced build-ready interaction gaps:
- program switcher = global state needs clearer UX contract in owner IA docs,
- Hôm nay vs Tiến độ still needs hard boundary if both remain top-level,
- critical-flow context switching and accessibility semantics must be designed at IA layer, not deferred to visual polish.
Cross-stage ideation continuation is active in draft-only mode via EVT_2026-02-22_cross-stage-ux-optimization-idea-pack-draft.md (no decision promotion yet; open tie-breaker/min-set/confidence-display questions parked in Open Questions view).
Cross-stage ideation refinement loop is captured in EVT_2026-02-22_cross-stage-ux-optimization-iteration-loop-draft.md (adaptive-practice-copilot draft reached internal DoD pass; still draft-only, no DEC promotion yet).
Cross-stage hardening defaults are drafted in EVT_2026-02-22_cross-stage-recommendation-hardening-defaults-draft.md (adds proposed defaults for tie-breaker, thin-inventory floor, confidence visibility, cold-start transition, and rollout guardrails; pending confirmation).
Cross-stage candidate lock refinement is drafted in EVT_2026-02-24_cross-stage-recommendation-hardening-candidate-lock-draft.md (narrows to 3 numeric confirmation knobs: 60% mixed-program threshold, 45-day inactivity degrade, +5% rollout gate).
Stage D is closed with recommendation baseline from DEC-0003 and refined by DEC-0042 + DEC-0045.
Stage D attempt execution reliability baseline (timed/untimed, draft safety, submit integrity) is locked by DEC-0068.
Stage D deep execution defaults (timed background/no-pause, draft TTL, submit retry/support) are locked by DEC-0069.
Stage D goal-comparison compatibility baseline is locked by DEC-0071 (direct/normalized/not-comparable modes by score profile map).
Stage D confidence/quality gate for numeric goal-gap and same-assessment-form fallback guidance is locked by DEC-0072.
Stage D recommendation freshness baseline is locked by DEC-0075 (freshness quota, topic cap, and light diversification after repeated manual refresh with no action).
Stage F/D cross-module handoff is locked by DEC-0076 (reactivation reminder intent seeds first recommendation set without adding comeback mode complexity).
Stage D/A monetization-safe actionability is locked by DEC-0077 (recommendation set prioritizes available-now items and caps lock teaser behavior).
Stage D recommendation explainability and trust baseline is locked by DEC-0078 (single primary-reason priority order plus deterministic confidence tiers and low-confidence cap/placement guardrail).
Stage D/G taxonomy baseline is locked by DEC-0073 through canonical assessment-form IDs and program baseline profiles.
Stage G vocabulary daily-loop baseline (focus queue, quick-start subset, overload governor, non-punitive carry-forward) is locked by DEC-0070.
Stage E is closed with CS trigger threshold, contact SLA, and visibility catalog captured in DEC-0013.
Stage F has decline detection and light intervention policy captured in DEC-0014.
Stage F has verdict cadence and 3-level outcome model captured in DEC-0015.
Stage F has rescue-pack touchpoint and re-prompt policy captured in DEC-0016.
Stage F has rescue-pack composition/cooldown/inline prompt policy captured in DEC-0017.
Stage F has new-decline/cooldown-reset policy captured in DEC-0018.
Stage F has similarity-ladder/recovery-quality/reset-scope policy captured in DEC-0019.
Stage F has simple-threshold/scale-guardrail policy captured in DEC-0020.
Stage F has numeric-threshold defaults and A/B cutoff captured in DEC-0021.
Stage F has evaluation-timing/transparency/review-cadence policy captured in DEC-0022.
Stage F has recovery-status visibility and escalation intervention policy captured in DEC-0023.
Stage F has AI Tutor 7-day plan structure/regeneration/exit policy captured in DEC-0024.
Stage F has AI Tutor plan sourcing/fallback/program-switch policy captured in DEC-0025.
Stage F has AI Tutor daily execution policy (missed-day/completion/optional-role) captured in DEC-0026.
Stage F has AI Tutor daily reminder policy (channel/timing/stop-rule) captured in DEC-0027.
Stage F has AI Tutor reminder edge policy (permission/cap/login-behavior) captured in DEC-0028.
Stage F has AI Tutor reminder consistency policy (timezone/precheck/multi-device) captured in DEC-0029.
Stage F has AI Tutor reminder quiet-hours/plan-singleton/deeplink policy captured in DEC-0030.
Stage F has AI Tutor day-boundary/unavailable-fallback/submit-timestamp policy captured in DEC-0031.
Stage F notification governor and AI-score-ready one-shot delivery are refined in DEC-0052.
Stage F template governance and safety fallback are refined in DEC-0053.
Stage F KPI minimum pack is refined in DEC-0054.
Stage F KPI operational contract (cadence/guardrails/escalation) is refined in DEC-0066.
Stage F is closed at current discovery scope.
Stage H baseline is now closed through Wave 2 with DEC-0082..DEC-0092 (role/grounding/integrity/output + rendering/provenance/sandbox exception).
Stage H current focus is rollout quality gates and implementation telemetry tuning.
Stage I baseline is now closed through DEC-0086..DEC-0089 (layer boundary + handoff + mobile compression + full-test mapping).
Stage I current focus is runtime optimization tuning under existing contracts (no new baseline blocker).

Stage gate checklist

Each stage closes only when:
- Core logic is documented in 20_DECISIONS or 30_STATE.
- Open questions are either resolved or explicitly marked deferred.
- UX extraction targets are mapped in 50_UX_EXTRACTION.

Discovery Sequence View - Outside In

Why this view exists

Discovery order (locked)

Current stage

Stage gate checklist

Related references