# Plan 001 — "Chung cư sở hữu 50 năm" : Vietnamese report + interactive HTML showcase

Date: 2026-08-26 · Project: Diep.Wealth (fresh — first content artifacts) · Status: DRAFT — awaiting GO

## 1. Outcome

Two artifacts, one storyline, legally accurate:

1. **Native Vietnamese report** (~1,800–2,500 words, báo chí explainer register, Northern Vietnamese, zero AI tells): "CHUNG CƯ 'SỞ HỮU 50 NĂM' – CÓ PHẢI 50 NĂM SAU LÀ MẤT NHÀ?" — dismantles the myth by separating *quyền sở hữu căn hộ* (property right, no expiry) from *thời hạn sử dụng công trình* (building lifespan), covers current law (Luật Nhà ở 2023 + sửa đổi 2025, Luật Đất đai 2024, NQ 21) and the stuck-old-apartment problem.
2. **Self-contained interactive HTML deck** presenting the report to a broad audience — scrollytelling one-pager, mobile-first (most VN readers are on phones), opens from `file://` with zero build.

Success = a reader who believed "50 năm nữa mất nhà" finishes either artifact understanding: the clock counts the *building*, not their *ownership* — and what actually happens when a building reaches end-of-life.

## 2. Skill combo (the stack, per phase)

| Phase | Skill / agent | Role |
|---|---|---|
| P0 Fact base | **`research-grounding`** (doctrine: no claim from memory) + **`antigravity:multi-source-search`** (external — consumed via `get_skill`, followed inline: cross-validated evidence ledger with source diversity, confidence, conflicts, gaps), executed by `researcher` agent with wigolo/exa/firecrawl rotation per RULES [60] | Live-source verification of every legal claim |
| P1 Report | **`vn-author`** (`--register bao-chi`, full pipeline, fact sheet as input data) | Author the Vietnamese report from scratch |
| P2 Editorial | **`vn-editor`** (polish pass) | Independent fresh-editor read (author ≠ judge) |
| P3 Fact validation | **`agent-validator`** (fresh context) running **`fact-check-framework`** | Claim inventory → classify verifiability → each claim ↔ P0 ledger citation → flag unsupported |
| P4 Deck | **`ak-show-off`** (activates `ak:frontend-design`; mandatory Design Read + Self-Review Gate) + **`ak-motion-design`** reference for scroll choreography | Build the self-contained showcase HTML |
| P5 Deck blind test | Fresh `general-purpose` agent (RULES [57] blind tester) driving **`ak-agent-browser`** | Real-browser sweep (desktop + 390px): click every interaction, screenshots, console errors — raw defect report |
| P6 Wrap | native | Place artifacts, final report, journal |

Rejected alternatives (checked, not defaults): `slides` (presentation-deck orientation — Chart.js patterns may inform deck charts, but ak-show-off owns the showcase build); `vn-canu-reporting`/`vn-bctt-report` (government briefing generators — wrong genre, heavy scope.json machinery); `vn-validator` (scoped to government-style briefing drafts); `antigravity:plotly` (new dep, YAGNI for 2–3 charts); `ego-browser` (targets user's real Chrome profile — local-file QA needs no profile); `antigravity:scientific-writing` (overlaps P0+P3 with heavier machinery).

**Optional add-on (default OFF):** PPTX companion of the deck, via the project-local `Diep.Wealth/SKILLS/pptx-deck-creation` (narrative + design profiles) and `pptx-official/references/html2pptx.md` (HTML→PPTX bridge) — only if he wants an email/print channel later.

## 3. Phases

### P0 — Legal fact base (~25–35 min, delegated)
RULES [11]: none of this ships from memory — all verified live, this session epoch.
1. **Luật Nhà ở 2023 + Luật sửa đổi 103/2025/QH15 (hiệu lực 1/8/2025)**: "thời hạn sử dụng nhà chung cư" concept — how determined (cấp công trình, hồ sơ thiết kế, kiểm định); rules on selling when >50% design life used; disclosure duty of chủ đầu tư; end-of-life path (tiếp tục sử dụng / cải tạo / phá dỡ xây mới) and owner protections (bồi thường, tái định cư, ưu tiên mua căn mới). Exact article numbers.
2. **Quyền sở hữu**: where the law anchors "quyền sở hữu nhà ở là quyền tài sản, không đặt thời hạn" for VN individuals buying commercial apartments.
3. **Luật Đất đai 2024**: đất ở đô thị do cá nhân sử dụng → ổn định lâu dài; where 50-year land terms actually apply (đất dự án thương mại, kết cấu hạ tầng) — the origin of the myth.
4. **NQ 21**: verify exact number, issuer, date, and what it actually says about chung cư hết niên hạn / tái phát triển chung cư cũ (user's text cites it without a date — do not ship an unverified date).
5. **Stats**: số liệu khu chung cư/tập thể cũ xuống cấp cần cải tạo (Bộ Xây dựng), реform progress numbers — each with source + date (RULES [14]).

Deliverable: `plans/reports/2026-08-26-001-fact-ledger.md` — claim ↔ source URL ↔ quote ↔ date, with per-claim confidence and open gaps, in the `antigravity:multi-source-search` ledger format. Deck and report cite only from this ledger.

### P1 — Report (~30–45 min)
Run **`vn-author`** inline (`✍️` flow), `--register bao-chi`, full pipeline (multi-pass review incl. final AI-tell sweep — not `--fast`). Inputs: user's seed text (as reference data, `<van_ban_nguon>`), P0 fact sheet (as sourced data — vn-author forbids invented figures/quotes, so every number must trace to the ledger). Structure:
1. Mở: cảnh thang máy (the overheard conversation)
2. Hai khái niệm bị trộn成一: quyền sở hữu căn hộ vs thời hạn sử dụng công trình
3. Luật hiện hành nói gì (2023/2025 sửa đổi, Đất đai 2024 — dẫn điều khoản)
4. Con số "50 năm" từ đâu而来 (đất có thời hạn / niên hạn thiết kế / công trình hạ tầng)
5. Vòng đời công trình hết hạn thì sao: kiểm định → 3 phương án, quyền của chủ sở hữu
6. Bài toán thực: chung cư cũ xuống cấp, cải tạo giậm chân tại chỗ
7. NQ 21: định hướng khơi thông
8. Kết: đồng hồ đếm công trình, không đếm quyền sở hữu
Output: `content/chung-cu-so-huu-50-nam/bao-cao.md`

### P2 — Editorial pass (~10–15 min)
**`vn-editor`** polish on P1 output: register check (Northern, bro-adjacent báo chí), AI-tell sweep, coherence. One pass, edits in place.

### P3 — Fact validation (~10–15 min)
Spawn **`agent-validator`** (fresh context) applying the **`fact-check-framework`**: enumerate every claim in the report, classify verifiability, check each against the P0 ledger only; verdict per claim (CONFIRMED / UNSUPPORTED / WRONG) — no rewriting. Any UNSUPPORTED → fix from ledger or cut the claim.

### P4 — HTML deck (~60–90 min)
**`ak-show-off`** with the final report as mission input. Preferences: `--no-publish` this run (publishing is a separate explicit step), screenshots on, languages vi-only. Deck sections:
1. **Hero** — big countdown "50:00:00…" → twist line: "đồng hồ này đếm cái gì?"
2. **Hai chiếc đồng hồ** — sở hữu (∞, không tắt) vs công trình (50 năm, tắt dần) — interactive side-by-side
3. **Myth/fact flip cards** — 4–5 hiểu lầm phổ biến (lật thẻ)
4. **Vòng đời tòa nhà** — scroll-driven timeline: xây → già hóa → kiểm định → 3 ngã rẽ
5. **Số liệu** — animated counters (khu cũ cần cải tạo, tiến độ) from P0 ledger
6. **Mini quiz** — "50 năm nữa bạn có mất nhà?" 3 câu, kết luận cá nhân hóa
7. **NQ 21 & định hướng** — what changes
8. **Kết + nguồn** — full citations
Constraints: semantic HTML, keyboard nav, `prefers-reduced-motion`, VN diacritics render (incl. inside canvas charts), zero console errors.
Output: `content/chung-cu-so-huu-50-nam/showcase/index.html`

### P5 — Blind test (~15–20 min)
Fresh agent, told nothing, driving **`ak-agent-browser`**: open `index.html` cold (desktop + 390px mobile), click through every interaction, report defects verbatim (incl. console errors and VN diacritics inside canvas charts). Fix loop until clean. Plus ak-show-off's own screenshots for the user.

### P6 — Wrap (~5 min)
Place artifacts, final summary with preview instructions, journal note (auto-captured). No publish, no git (not a repo) unless ordered.

## 4. Assumptions (defaults chosen — say otherwise to override)

1. Register: `bao-chi` explainer (educated general reader), not `hanh-chanh` briefing.
2. Deck language: Vietnamese only (ak-show-off default is vi+en; vi chosen for the audience).
3. Length: 1,800–2,500 words; deck mirrors it, doesn't replace reading.
4. Placement: `content/chung-cu-so-huu-50-nam/` in Diep.Wealth (new dirs — first content).
5. The user's seed text is a source skeleton, not the final prose — vn-author rewrites natively around the same argument.

## 5. Risks

- **Legal nuance**: the myth has ≥3 origins (đất có thời hạn, niên hạn thiết kế, công trình hạ tầng). Report must dismantle each — replacing one oversimplification with another is the top failure mode.
- **NQ 21 identity**: verify number/issuer/date/content live before citing (P0 step 4).
- **Chart fonts**: Chart.js canvas text can drop VN diacritics — P5 checks glyph rendering explicitly.
- **Self-containment**: deck must work offline from `file://` (CDN fallbacks checked in P5).

## 6. Effort

Total ≈ 2.5–3.5 h of in-session agent work: P0 25–35 m · P1 30–45 m · P2 10–15 m · P3 10–15 m · P4 60–90 m · P5 15–20 m · P6 5 m.

## 7. Acceptance criteria

- Report: native Northern Vietnamese, no AI tells (vn-editor pass clean), every legal claim and number traced to P0 ledger (agent-validator: zero UNSUPPORTED).
- Deck: loads standalone `file://`, zero console errors, usable at 390px, all interactions fire, all VN text correct with diacritics, blind test clean.
- Consistency: same numbers and conclusions in both artifacts; deck links/cites the same sources as the report.