Record := {
  Z: Z0|Z1|Z2|Z3
  Skill: capability target
  FailureModes: top break patterns
  Sensors: how to detect (fast)
  Repairs: smallest reliable loop
  OutputTest: pass/fail condition (P3)
}

Skill: Word_Retrieval_P3(word)
Meaning: recall word instantly + correct meaning + correct usage frame

FM1: Recognition-only (P1) — “I know it when I see it”
FM2: Slow recall (P2) — “I can get it… give me time”
FM3: Wrong meaning slice — partial/incorrect semantic mapping
FM4: Wrong collocation — unnatural word pairing
FM5: Wrong register/tone — formal/informal mismatch
FM6: Spelling/phonology mismatch — can say but can’t spell / vice versa

S1: RecallTime(word) > 2s (no prompting)  → not P3
S2: DefinitionDrift (explains vaguely)     → mapping weak
S3: CollocationErrorRate high             → usage frame missing
S4: SpellingErrorRate high                → orthography unstable
S5: SpeechHesitation mid-sentence         → retrieval not automatic

R0a: 2-Sentence Anchor
  - write 2 sentences: neutral + emotional
  - include collocation pair
R0b: 3-Recall Drill (spaced)
  - recall at T+0, T+10min, T+24h
  - must hit <2s each time
R0c: Speak-Then-Write Transfer
  - say sentence aloud → write same sentence

Pass if:
  RecallTime < 2s AND
  sentence natural AND
  correct collocation AND
  correct tone AND
  spelling correct (writing context)

Skill: Chunk_Retrieval_P3(phrase/collocation)
Meaning: retrieve multi-word unit as one object

FM1: Word-by-word assembly (slow, clunky)
FM2: Wrong preposition / wrong particle
FM3: “Synonym substitution” breaks meaning
FM4: Overuse of generic verbs (do/make/get)
FM5: Repeated same phrase (low variation)
FM6: Misuse idioms (literal confusion)

S1: OveruseGenericVerbs rate high
S2: PrepositionErrorRate high
S3: CollocationGap (writes correct word, wrong pair)
S4: VariationIndex low (same adjectives/verbs repeated)

R1a: Collocation Pair Bank
  - For each target word, store 3 strong pairs
  - e.g. "make" → make an effort / make a decision / make progress
R1b: Swap-Frame Drill
  - keep meaning constant, change frame:
    "very tired" → exhausted / drained / worn out
R1c: Pattern Sentence Templates
  - store patterns:
    "Not only..., but also..."
    "As soon as..., ..."

Pass if:
  writes 5 sentences with target chunk family
  with natural variation + zero preposition errors

Skill: Sentence_Build_P3(structure)
Meaning: produce correct structure automatically under load

FM1: Run-on sentences / fragments
FM2: Tense drift
FM3: Subject-verb agreement failures
FM4: Weak connectors (and/then/so)
FM5: Missing clarity (who did what)
FM6: Over-complex sentences collapse

S1: ErrorPer100Words high
S2: ConnectorDiversity low
S3: ClarityTest fail (reader unsure actor/action)
S4: SentenceLength variance extreme (all too long / too short)

R2a: 1-Structure-per-Week
  - choose 1 structure and over-train:
    relative clauses / contrast / cause-effect / dialogue tags
R2b: Connector Ladder
  - replace "and/so" with precise connectors
  - train 10 connectors in grouped meaning families
R2c: Rewrite Weakest Sentence
  - identify worst sentence → rewrite 3 ways

Pass if:
  timed paragraph has
  - zero tense drift
  - clear actor/action
  - ≥ 4 connector types
  - ≤ 2 grammar errors per 120 words

Skill: Exam_Composition_P3
Meaning: stable narrative + rich language under time pressure

FM1: “Blank page” freeze
FM2: Vocabulary collapses under timer (P3→P2)
FM3: Repeats same words/phrases
FM4: Flat verbs + flat adjectives
FM5: Weak emotion/imagery
FM6: Story coherence breaks (logic gaps)

S1: FreezeEvents per writing session
S2: RewriteDelta huge (first draft weak)
S3: VariationIndex low
S4: TimeOverrun (cannot finish)
S5: ParagraphWeakness concentration (one paragraph collapses)

R3a: 5-Box Plan (story skeleton) + vocab targets
R3b: Timed Micro-Drills (8–12 min paragraphs)
R3c: “Weakest Paragraph Rewrite” every session
R3d: Targeted Vocab Injection:
     choose 8 words + 4 chunks + 2 sentence structures per comp

Pass if:
  completes full comp on time
  with stable structure + visible variation
  and ≥ 12 deliberate vocabulary upgrades used correctly

Input:
  age/level (Pri/Sec/Adult)
  goal (exam / speaking / writing / comprehension)
  symptom (what fails)
  time budget (10/20/40 min per day)

If symptom = "I recognise but can't use"
  RootCause = P1 only (no forced output)
  Repair = R0b + R0c daily 10 min
If symptom = "I pause mid-sentence"
  RootCause = retrieval latency (P2) + load sensitivity
  Repair = timed speaking drills + chunk training (R1a)
If symptom = "My writing uses basic words"
  RootCause = missing chunk bank + generic verbs
  Repair = Collocation Pair Bank + Swap-Frame Drill (R1a, R1b)
If symptom = "I keep repeating same words"
  RootCause = low variation index (thin synonym families)
  Repair = 3 synonym families per week + inject into paragraph
If symptom = "Grammar collapses when I try to sound advanced"
  RootCause = Z2 structure not automated
  Repair = 1-Structure-per-Week + rewrite 3 ways (R2a, R2c)
If symptom = "I blank out in exams"
  RootCause = stress + no load training (P3 not stable)
  Repair = timed micro drills + 5-box plan (R3b, R3a)
If symptom = "I forget words after learning"
  RootCause = no spaced activation
  Repair = 3-recall drill T+0/T+10m/T+24h + weekly recycle

1) 3-recall drill (5 words)
2) 2-sentence anchor (1 word)
3) Speak-then-write (1 sentence)

1) 5 collocations (bank)
2) 8-minute timed paragraph (inject 6 upgrades)
3) rewrite weakest sentence

1) input (10 min story excerpt)
2) output (12 min paragraph)
3) correction + rewrite weakest paragraph (12 min)
4) spaced review (6 min)

Memorising word lists produces recognition (P1),
not usable vocabulary (P3),
because it skips output, context, and load training.

Test: Can student use the word correctly in:
  - speaking within 2 seconds?
  - writing under time pressure?
  - correct collocation + tone?
If NO → word list did not work.

WordList → isolated symbol storage
→ weak context binding
→ retrieval latency
→ load causes freeze
→ exam output reverts to basic words
→ student concludes “I’m bad at English”

Replace WordList with:
  Word+Chunk+Sentence+TimedUse+Rewrite loop

Primary_Vocab_OS {
  Goal: move children from P1→P3 for common + story-rich vocabulary
  CoreLoop:
    Input (stories) →
    Anchor (2 sentences) →
    Output (talk + write) →
    Feedback →
    Spaced recall
  Minimum Standard:
    1000+ P3 words + 300 P3 chunks by end Pri 6 (target band)
}

Primary_Sensors {
  RecallTime < 2s (common words)
  VariationIndex in comp ≥ 0.6 (simple proxy)
  Connector diversity: because/so/but/when/if
  Generic verbs count per paragraph trending down
}

Secondary_Vocab_OS {
  Goal: exam-stable vocabulary under time pressure (P3-load stable)
  Engine:
    ChunkBank (collocations) +
    SentenceStructure automation +
    Timed paragraph drills +
    Weakest paragraph rewrite
}

Secondary_Sensors {
  FreezeEvents per timed session
  UpgradesUsedCorrectly >= 12 per comp
  Grammar error rate <= 2 per 120 words
  VariationIndex trending up weekly
}

VOCABULARY_OS_SPEC v1.0 {
  Vocabulary := retrievable symbols enabling thought↔meaning↔output
  Phases: P0 absent | P1 passive | P2 active-weak | P3 active-stable
  Law1: Input != Output (must train output)
  Law2: Context binding required (word = symbol+meaning+usage frame)
  Law3: Decay without activation (phase drift)
  Law4: Load reveals truth (P3 must survive time pressure)
  CoreLoop: Input → Output → Feedback → Repair → Spaced recall
  Gate: P3 = <2s recall + correct usage + correct tone + correct collocation
}