# English, the 80/20 Way — From Vietnamese L1 to Functional Fluency > **20 minutes a day. 180 days. ~90 high-frequency scenarios.** This repo is a > speaking- and writing-first fluency lab built for the **Vietnamese L1** learner. > No grammar encyclopaedia, no rare vocabulary — just the **20% of English that > carries ~85–90% of real conversation**, drilled until it leaves your mouth > without translation. --- ## 🟢 Start here **Open [`./index.html`](./index.html) in your browser.** That dashboard is the front door of the whole repo. Pick a card → open a bundle → practice. > **The `.html` is what you practice with. The `.md` and `_corpus.md` are references.** > | File you see in a bundle | What it is | When you open it | |---|---|---| | `{name}.html` | **The practice player** — your daily workout. | **Every day. This is the artifact.** | | `{NAME}.md` | The readable guide: what + why + L1 pitfalls + cheat sheet. | When you want the *why* behind a chunk. | | `{name}_corpus.md` | Ground truth: every real native example + source URL + IPA. | For the curious, or to audit a claim. | You will live inside the `.html`. The other two files exist so that **nothing you say is invented** — every line you drill traces back to a real native utterance (Cambridge, COCA, YouGlish, …). > ⚠️ **Honest note — the dashboard fills in over time.** Bundles ship > **incrementally**; a card's link only goes live once that bundle is built. If a > card is greyed out, it is coming soon. See the > [“Bundles ship incrementally”](#honest-notes) section below. --- ## What “80% native” actually means We say it plainly, because overpromising fluency is the oldest scam in language learning. **“80% native” is functional, structural fluency in the high-frequency zones — not a flawless native accent.** Concretely, after 180 days you can: - Handle **~90 common scenarios** fluidly, using native-like **chunks** (not word-by-word translation from Vietnamese). - Be **intelligible without repetition** — a native listener understands you the first time, even if your accent is still Vietnamese. - **Switch register** (casual ↔ professional) and **mode** (speak ↔ write) on demand. - Produce the **~2,000 most-frequent spoken word families**, the **~50 speech acts** that organise daily talk, and the **~10 pronunciation fixes** that unblock Vietnamese-speaker intelligibility. That ~2,000 + ~50 + ~10 is the Pareto core. It covers an estimated **85–90% of real conversation**. The last 10–15% — accent refinement, rare idioms, literary register — is a lifetime project, and it is **out of scope** here on purpose. --- ## The six mindset principles (why this works) These six ideas decide *what we drill and what we ignore*. If a lesson violates one of them, it doesn’t belong in this repo. | # | Principle | What it means in practice | |---|---|---| | 1 | **Pareto 20→80** | Attack the high-frequency zones. Ignore rare grammar, exotic idioms, anything a native speaker uses less than once a week. | | 2 | **Chunks, not words** | Fluency = retrieving multi-word patterns (*“How’s it going?”, “I was wondering if…”, “Let’s circle back”*). We never drill single words in isolation. | | 3 | **Output only counts** | You only “know” what you can **say** or **write** under time pressure. Every single day you produce — record yourself, or write 3–5 sentences. Reading is not enough. | | 4 | **Real attestation** | Every example is a **cited** native utterance (dictionary / corpus / YouGlish). No invented “this sounds native to me” sentences — ever. | | 5 | **L1-aware** | Every pitfall targets a real **Vietnamese → English** interference: dropped final consonants, missing tense, pro-drop, /θ/→/t/. This is where Vietnamese learners actually lose points. | | 6 | **Daily dose beats binge** | 20 min/day × 180 days beats 6 hours on Sunday. Spaced retrieval is the mechanism; the `.html` player schedules it for you. | --- ## The bundle model — `.html`-first A **bundle** is the smallest unit of practice. It is a **triple** of files that share one stem (e.g. `greetings_intros`) and live together in a phase subfolder: | File | Role | Hard rule | |---|---|---| | `{name}_corpus.md` | **Ground truth.** Every real example + source URL + IPA + frequency rank. | Every line cited. **No invented sentences.** | | `{NAME}.md` | **Readable guide (reference).** What + why + the Vietnamese→English pitfalls table + a ≤8-chunk cheat sheet. | Quotes the corpus verbatim under `> From {name}_corpus.md:` callouts. | | `{name}.html` | **Primary learner artifact.** The interactive practice player you open every day. | Every chunk/dialog line traces back to the corpus. | > **The one rule of this repo:** *every example that appears in a `.md` or `.html` > is a real, **cited** attestation recorded in `{name}_corpus.md`.* Nothing is > invented. If you can’t find a native source for a sentence, it doesn’t go in. ```mermaid graph TD DASH["english/index.html
(dashboard — the front door)"] -->|click a card| HTML["{name}.html
PRACTICE PLAYER
(what you open daily)"] HTML -->|curious about the why?| MD["{NAME}.md
guide + L1 pitfalls + cheat sheet"] HTML -->|audit a claim| CORPUS["{name}_corpus.md
every example, cited"] MD -.->|quotes verbatim| CORPUS HTML -.->|every chunk traces to| CORPUS style DASH fill:#eaf2f8,stroke:#2980b9,stroke-width:3px style HTML fill:#eafaf1,stroke:#27ae60,stroke-width:3px style MD fill:#fef9e7,stroke:#f1c40f style CORPUS fill:#f4ecf7,stroke:#8e44ad ``` **Read the diagram left-to-right:** you start at the dashboard, you *practice* inside the `.html`, and the `.md` / `_corpus.md` are the reference layer that makes the practice trustworthy. --- ## The 6-phase roadmap (180 days, 90 bundles) One phase = one subfolder = one cluster of skills. Each bundle is **2 days** of paced work. Day ranges, counts, and stems below are **canonical** — every doc in this repo uses these exact values. | Phase | Folder | Days | Bundles | Theme | |---|---|---|---|---| | **0 — Pronunciation** | `pronunciation/` | 1–20 | 10 | Fix the sounds that block intelligibility for Vietnamese L1: final consonants, /θ/ /ð/, clusters, vowel length, stress, linking, reductions, intonation, thought groups. | | **1 — Core Speech Acts** | `speech_acts/` | 21–60 | 20 | The daily conversational functions: greet, thank, apologise, request, agree/disagree, interrupt, clarify, close, schedule, phone. | | **2 — Workplace Speaking** | `workplace/` | 61–90 | 15 | Meetings, status updates, short presentations, SBI feedback, STAR interviews, negotiating, networking, video-call specifics. | | **3 — Writing** | `writing/` | 91–130 | 20 | Email anatomy, register, requests, apologies, bad-news, follow-ups, IM/Slack, LinkedIn, cover letters, CV bullets, editing, complaints. | | **4 — Discourse & Nuance** | `discourse/` | 131–160 | 15 | Hedging, humour, politeness, top-frequency idioms, phrasal verbs, collocations, register switching, storytelling, discourse markers, fillers. | | **5 — Capstone** | `capstone/` | 161–180 | 10 | Integration under pressure: impromptu talks, debating, live feedback, speaking under pressure, timed writing, sustained monologue, full simulations. | | **Total** | 6 subfolders | **1–180** | **90** | Speaking- and writing-first functional fluency. | > Want the **day-by-day map** with every bundle’s title and one-liner? Open > [`./CURRICULUM.md`](./CURRICULUM.md). It doubles as a progress tracker > (`- [ ]` checkboxes per bundle). --- ## The daily ritual (~20–25 min) This is the **20% effort** that produces the 80% result. Do it every day, in this order, inside the bundle’s `.html` player: ```mermaid graph LR R["📖 READ
5 min
the bundle guide"] --> S["🎙️ SHADOW
7 min
drill chunks + dialog aloud"] S --> P["✍️ PRODUCE
8 min
record a variant OR write 3–5 sentences"] style R fill:#eaf2f8,stroke:#2980b9 style S fill:#eafaf1,stroke:#27ae60 style P fill:#fef9e7,stroke:#f1c40f,stroke-width:3px ``` 1. **READ (5 min)** — skim the bundle guide (`{NAME}.md`) for the *what* and the Vietnamese→English pitfalls. 2. **SHADOW (7 min)** — drill the ≤8 survival chunks **and** the dialog aloud, with the YouGlish clips open. Mimic rhythm, not just words. 3. **PRODUCE (8 min)** — **speak** (record a variant of the dialog with your own details, via the player’s tap-to-record) **or** **write** (3–5 sentences using the chunks). This is the step that builds fluency. Skip it and you’re memorising, not learning. > 🗓️ **Every 7th day = Sunday integration.** No new bundle. Instead, run **one > ~10-minute conversation simulation** that combines the week’s functions > (self-recorded). This is where isolated chunks start to chain into real speech. > The `.html` player for that week’s last bundle has a simulation lane for this. --- ## How to use this repo 1. **Open [`./index.html`](./index.html)** in any modern browser. This is your dashboard — bookmark it. 2. **Start at Phase 0, Bundle 01** (`pronunciation/final_consonants`). Do **not** skip pronunciation even if you “already know some English” — it is the #1 intelligibility lever for Vietnamese L1 and unblocks everything after it. 3. **Spend 2 days per bundle.** Day 1: READ + SHADOW. Day 2: PRODUCE (speak or write) + review yesterday's recording. 4. **Mark each bundle finished** via the button inside its `.html` — it syncs to the dashboard through shared `localStorage`, so your progress persists between sessions on the same browser. 5. **Every 7th day, run the integration simulation** instead of opening a new bundle. 6. **Follow the phase order.** Each phase assumes the previous one’s chunks are already semi-automated. Jumping ahead to Workplace before Core Speech Acts will feel hollow. 7. **Track the full 180 days** in [`./CURRICULUM.md`](./CURRICULUM.md) — tick a box every time you mark a bundle finished. --- ## Folder layout ``` english/ ├── README.md ← you are here (mindset + roadmap) ├── HOW_TO_RESEARCH.md ← builder law: how every bundle is produced (subagents) ├── CURRICULUM.md ← day-by-day map of all 90 bundles (progress tracker) ├── _SPEC.md ← internal coordinator spec (not learner-facing) ├── index.html ← the dashboard (front door — open this first) │ ├── pronunciation/ ← Phase 0 · Days 1–20 · 10 bundles │ ├── final_consonants_corpus.md │ ├── FINAL_CONSONANTS.md │ ├── final_consonants.html │ ├── th_sounds_corpus.md │ ├── TH_SOUNDS.md │ ├── th_sounds.html │ └── … │ ├── speech_acts/ ← Phase 1 · Days 21–60 · 20 bundles ├── workplace/ ← Phase 2 · Days 61–90 · 15 bundles ├── writing/ ← Phase 3 · Days 91–130 · 20 bundles ├── discourse/ ← Phase 4 · Days 131–160· 15 bundles └── capstone/ ← Phase 5 · Days 161–180· 10 bundles ``` Inside any phase subfolder, **one bundle = three files sharing a stem**: `{name}_corpus.md` (ground truth) · `{NAME}.md` (guide) · `{name}.html` (the player you practice with). --- ## For builders If you are **building** bundles (not learning from them), the law lives in [`./HOW_TO_RESEARCH.md`](./HOW_TO_RESEARCH.md). It encodes: - The **one rule** (every example cited; nothing invented). - The **triple-per-bundle** model and each file’s hard rules. - The **worker prompt template** + the mandatory Tailwind v4 head + shared `@theme` dark palette every `.html` must use verbatim. - The **verification sweep** every bundle must pass before it ships. > **The golden rule of building:** the coordinator **never hand-writes a bundle.** > Every bundle is produced by a subagent (one worker per bundle), launched in > parallel, then verified. Fresh context per bundle keeps bundle #90 as rigorous > as bundle #01. --- ## Honest notes A few things we will not pretend: - **Bundles ship incrementally.** This repo is built bundle-by-bundle by subagents. The dashboard (`index.html`) is the front door, but **a card’s link only goes live once that bundle is built.** If a card is greyed out, the bundle is queued, not missing — check back, or track progress in [`./CURRICULUM.md`](./CURRICULUM.md). - **The `.html` player needs internet.** Styling uses the **Tailwind v4 browser CDN** (`cdn.jsdelivr.net/npm/@tailwindcss/browser@4`), and audio clips stream from **YouGlish/YouTube**. Per Tailwind’s own docs, the Play CDN is a **dev-only** convenience — fine for a personal learning repo on a connected machine, but **pages will render unstyled offline** and audio links will be dead. No audio files are bundled; everything is streamed from the cited native source. - **Accent is flagged per chunk (US/UK),** not chosen globally. Real native usage varies; the corpus records which variety each example is. - **This is not a grammar course.** We teach the chunks and speech acts that carry daily conversation. Deep grammar study is a different (and much slower) path — take it after, not instead of, this one. --- ## Next steps - **Learner?** Open [`./index.html`](./index.html) → Phase 0 → Bundle 01. - **Want the full day map?** [`./CURRICULUM.md`](./CURRICULUM.md). - **Building a bundle?** [`./HOW_TO_RESEARCH.md`](./HOW_TO_RESEARCH.md). > **Sources & further reading:** authoritative references used across this repo > — Cambridge, Oxford Learner’s, Collins, Merriam-Webster, Macmillan; > COCA / BNC (english-corpora.org) and wordfrequency.info (spoken sub-corpus); > YouGlish and Forvo for audio; the Manchester Academic Phrasebank for writing. > These are the corpora every `{name}_corpus.md` cites from.