How Secondary 3 English Tuition Works Or Does Not (Almost-Code, Singapore)

SEC3.ENGLISH.TUITION.OS (SINGAPORE) — ALMOST-CODE v0.1
Scope: Secondary 3 (Upper Sec) | Track: O-Level readiness (1184) | Output: stable language under load
================================================================================
A) SYSTEM INTENT (WHAT THIS “TUITION” REALLY IS)
Intent:
  Tuition := a closed-loop upgrade system that increases:
    - BindIntegrity (ideas connect)
    - CorridorStability (can produce under time + stress)
    - TransferReliability (works across prompts, texts, contexts)
One-line Lock:
  "Secondary 3 English tuition works when it converts vocabulary+grammar into
   repeatable corridors for writing, comprehension, listening and oral—stable under load."
================================================================================
B) OFFICIAL CONSTRAINTS (SYLLABUS + EXAM MAP)
B1) MOE Syllabus Aims (installed as operating objectives):
  AO-MOE-1: Listen / Read / View critically and accurately across text types
  AO-MOE-2: Speak / Write / Represent in Standard English for purpose/audience/context
  AO-MOE-3: Use Standard grammar + vocabulary accurately; understand how language makes meaning/impact
  AO-MOE-4: Use English with impact, effect, affect
B2) O-Level 1184 Assessment Map (Tuition must pre-build these corridors):
  Paper1:
    S1 Editing (grammar error correction)
    S2 Situational Writing: 250–350 words (purpose/audience/context + visual stimulus)
    S3 Continuous Writing: 350–500 words (choose 1 of 4 topics; narrative/argument etc)
  Paper2:
    Comprehension (literal/inferential/evaluative)
    Vocabulary-in-context + language-for-effect
    Summary ≈ 80 words (exclude provided intro words)
  Paper3:
    Listening (varied tasks; informational note-taking task included)
  Paper4:
    Oral Planned Response (video prompt; up to ~2 min)
    Spoken Interaction (discussion on related topic)
================================================================================
C) VOCABULARY REQUIREMENTS (WHAT “GOOD VOCAB” MEANS IN SEC 3)
VocabularyRequirement := NOT word-count.
VocabularyRequirement := (PrecisionTokens + RegisterControl + CollocationControl + BindTokens)
Define VocabKit.SEC3 :=
  NodeSets:
    N1 AcademicVerbs (analyse, evaluate, justify, infer, qualify, synthesise…)
    N2 Stance/Modal tokens (may, must, likely, arguably, however, therefore…)
    N3 Precision nouns (policy, consequence, trade-off, incentive, reliability…)
    N4 Descriptive gradients (subtle→obvious; tentative→definitive; beneficial→harmful…)
    N5 Topic banks (Sec3 typical: tech/media, environment, health, school, community, ethics)
  BindTokens (mandatory):
    B1 cause→effect
    B2 contrast→counterpoint
    B3 concession→qualification
    B4 example→generalisation
    B5 sequence→process
    B6 claim→evidence→explanation
  OutputRule:
    VocabularyCounts only when it improves:
      - comprehension of vocab in context
      - clarity + shades of meaning in writing
      - oral fluency + precision
================================================================================
D) DIAGNOSTICS (WEEK 0) — FIND THE REAL FAILURE
InputTests (baseline):
  T1 WritingSnapshot:
    - 1 Situational (short plan + 1 paragraph)
    - 1 Continuous (intro + 2 body paras)
  T2 ComprehensionSnapshot:
    - inference questions + vocab-in-context + short summary attempt
  T3 EditingSnapshot:
    - 20–30 sentence-level edits (common grammar patterns)
  T4 OralSnapshot:
    - 1 planned response (2 min) + 5 min discussion
  T5 ReadingSpeed+Accuracy:
    - timed passage: main idea / detail / tone
DiagnosisTags:
  FP1 NODE_RICH_BIND_POOR           # knows words but can’t explain
  FP2 CORRIDOR_OVERFIT              # only works on “expected” formats
  FP3 SENTENCE_UNSTABLE             # grammar breaks under speed
  FP4 SUMMARY_COLLAPSE              # cannot compress without copying
  FP5 ORAL_FLUENCY_GAP              # ideas exist; delivery collapses
================================================================================
E) CORE TRAINING LOOPS (THE TUITION ENGINE)
Loop = Teach → Model → Drill → Stress → Feedback → Repair → Re-test
E1) LOOP.WRITING (Paper 1 readiness)
  Corridor.W1 Situational:
    Plan := Audience + Purpose + Context + RequiredPoints + Tone + VisualUse
    Output := 250–350 words; points covered; clear organisation
  Corridor.W2 Continuous:
    Plan := Thesis/Angle + 2–3 BodyClaims + Evidence + Counter/Qualification
    Output := 350–500 words; cohesion; paragraph logic
  SensorPack.W:
    S-W1 TaskFulfilmentScore
    S-W2 CohesionScore (topic sentences + linking)
    S-W3 PrecisionScore (verbs + stance + collocations)
    S-W4 ErrorRate (grammar/punct/spell) under time
  RepairRules.W:
    IF S-W4 high THEN run EditingMicroLoop daily (10 min)
    IF S-W2 low THEN run BindParagraph template (claim→why→example→link)
    IF S-W3 low THEN install “upgrade swaps” (weak word → precise token)
E2) LOOP.READING+COMPREHENSION (Paper 2 readiness)
  Corridor.R1 Understand:
    main idea → supporting details → inference → writer’s intent/tone
  Corridor.R2 VocabInContext:
    meaning from surrounding binds (contrast, example, definition, tone)
  Corridor.R3 Summary (~80 words):
    extract points → group → rephrase → compress without copying
  SensorPack.R:
    S-R1 InferenceAccuracy
    S-R2 VocabInContextAccuracy
    S-R3 SummaryCompressionRatio
    S-R4 OwnWordsIndex (avoid lifting)
  RepairRules.R:
    IF Summary lifts text THEN enforce ParaphraseConstraint:
      "No 3 consecutive words from source"
    IF Inference weak THEN run “EvidenceLine” drill:
      claim must point to a line + explain why
E3) LOOP.GRAMMAR+EDITING (Paper 1 Section A readiness)
  GrammarTargets.SEC3:
    - subject–verb agreement
    - tenses + consistency
    - pronoun reference
    - modifiers
    - sentence boundaries (run-ons/fragments)
    - parallelism
  SensorPack.G:
    S-G1 EditAccuracy
    S-G2 ErrorTypeHistogram
  RepairRules.G:
    IF ErrorType repeats 3x THEN create 5-sentence minimal pair drill
E4) LOOP.LISTENING (Paper 3 readiness)
  Corridor.L1:
    identify gist → details → inference; note-taking structure
  SensorPack.L:
    S-L1 DetailAccuracy
    S-L2 NoteTakingCompleteness
E5) LOOP.ORAL (Paper 4 readiness)
  Corridor.O1 Planned Response (2 minutes):
    Hook → Point → Reason → Example → Link (H-P-R-E-L)
  Corridor.O2 Spoken Interaction:
    answer → elaborate → qualify → question back
  SensorPack.O:
    S-O1 IdeaOrganisation
    S-O2 VocabularyRange (appropriate, not fancy)
    S-O3 Fluency+Pronunciation
  RepairRules.O:
    IF rambling THEN enforce “2-point max” rule + timed delivery
================================================================================
F) WEEKLY SCHEDULER (SEC 3 DEFAULT)
WeeklyCycle (2 tuition sessions + daily micro):
  Session A (90–120m): Writing corridor + Editing repair
  Session B (90–120m): Comprehension + Summary + Oral drill
  Daily Micro (15–25m):
    - 10m Editing OR Sentence upgrade
    - 10m Reading (1 short article) + 5-sentence summary
    - 3 “Bind swaps”: however/therefore/although + claim/evidence link
================================================================================
G) Z0–Z6 BIND (SO IT SCALES AS A SYSTEM)
Z0: Sentence stability (grammar + precise token)
Z1: Paragraph corridors (binds: cause/contrast/qualify)
Z2: Paper corridors (P1/P2/P3/P4 readiness)
Z3: School load (time limits + common task patterns)
Z4: National standard (assessment objectives + exam constraints)
Z5: Cross-context transfer (novel prompts, new topics, new text genres)
Z6: Real-world capability (explain, persuade, summarise, negotiate under load)
================================================================================
H) VERIFICATION (PROVE IT WORKS)
Every 4 weeks run ReTestSuite:
  - 1 Situational + 1 Continuous (timed)
  - 1 Compre set + Summary (timed)
  - 1 Oral planned response (timed) + discussion
  - 1 Editing set
PassCondition (Sec3 target):
  CorridorStability := output quality does not collapse under time + prompt shifts
  TransferReliability := same skills survive new topics/text types
================================================================================
END

SEC3.ENGLISH.TUITION — HOW IT DOES NOT WORK (NEGATIVE VOID) v0.1
Scope: Singapore Secondary 3 | Target: O-Level English Language (1184) readiness
Anchors: MOE EL Syllabus 2020 aims include listening/reading/viewing; speaking/writing/representing; accurate grammar+vocabulary.
         SEAB 1184 (Y26) scheme includes Editing; Situational 250–350; Continuous 350–500; Paper 2 includes vocab-in-context + language for effect + ~80-word summary; plus Listening & Oral.
Entities:
- MOE
- SEAB
================================================================================
0) THE OBJECTIVE (WHAT “WORKING” WOULD HAVE MEANT)
WorkingTuition := raises
  - BindIntegrity (ideas connect)
  - CorridorStability (output holds under time/stress)
  - TransferReliability (skills survive new prompts/text types)
across Paper1/2/3/4 constraints.
NonWorkingTuition := increases activity/time but does NOT increase the above 3.
================================================================================
1) CORE FAILURE: VOCAB AS INVENTORY (NOT MECHANISM)
PATTERN NV1: WORDCOUNT_OPTIMISATION
Belief:
  "Vocabulary = how many words you can list."
Behavior:
  - memorise synonyms
  - replace simple words with “fancy” words
  - collect word banks without usage constraints
Result:
  nodes↑ binds↓ corridors↓
Output:
  - “wordy but empty”
  - wrong register, wrong collocations, awkward tone
Exam effect:
  - Situational writing misses audience/purpose/context despite “good vocab”
  - Continuous writing becomes incoherent under time
FailureTrace:
  inventory_nodes → weak_binds → corridor collapse → truncation under load
================================================================================
2) TEMPLATE OVERFIT (MODEL ESSAY = CRUTCH)
PATTERN NV2: CORRIDOR_OVERFIT_TEMPLATE
Belief:
  "If I memorise formats, I’m safe."
Behavior:
  - copy model essays
  - memorise introductions/conclusions
  - force every prompt into same 2–3 story/argument shapes
Result:
  works only on expected prompts; fails on novel phrasing / unusual angle
Exam effect:
  - Continuous writing: irrelevant content, forced theme, poor task fit
  - Oral: rehearsed monologue collapses during interaction (cannot adapt)
FailureTrace:
  narrow corridor → prompt shift → corridor breaks → panic writing/talking
================================================================================
3) GRAMMAR DRILLS WITHOUT TRANSFER (EDITING ISOLATED)
PATTERN NV3: GRAMMAR_ISLAND
Belief:
  "Do enough worksheets and grammar will appear in writing."
Behavior:
  - drill rules in isolation
  - do Editing questions but never re-run the same error inside real writing
Result:
  grammar holds in slow practice but breaks under time
Exam effect:
  - Paper 1 Editing may improve slightly, but writing still error-heavy under load
FailureTrace:
  rule knowledge (slow) ≠ production stability (fast) → error rebound in timed tasks
================================================================================
4) COMPREHENSION AS GUESSWORK (NO EVIDENCE BINDS)
PATTERN NV4: ANSWER_WITHOUT_ANCHOR
Belief:
  "Compre is vibes / intuition."
Behavior:
  - answer inference/evaluative questions without pointing to lines + explaining why
  - vocab-in-context treated as dictionary recall, not context binds
Result:
  low reliability; marks fluctuate; cannot justify
Exam effect:
  - Paper 2: inference + language-for-effect weak; vocab-in-context inconsistent
FailureTrace:
  no evidence-bind → weak justification corridor → marks volatility
================================================================================
5) SUMMARY FAIL MODE: COPYING (NO COMPRESSION CORRIDOR)
PATTERN NV5: LIFTING_TEXT
Belief:
  "Summary = shorten by deleting."
Behavior:
  - copy phrases
  - keep original sentence structure
  - no grouping of points
Result:
  low “own words” control; poor organisation; unclear meaning
Exam effect:
  - Summary (~80 words; intro words excluded) becomes penalty zone
FailureTrace:
  no compression corridor → lifting → loss of clarity/precision
================================================================================
6) NO STRESS TESTING (PRACTICE WITHOUT LOAD)
PATTERN NV6: NO_LOAD_TRAINING
Belief:
  "If I can do it untimed, I can do it in exam."
Behavior:
  - write slowly with unlimited time
  - redo only “comfortable” questions
  - avoid speaking on camera / timed oral
Result:
  corridors never become load-stable
Exam effect:
  - Paper 1/2 timing causes collapse
  - Oral planned response + interaction becomes rambling or blanking
  - Listening note-taking misses details under real-time constraints
FailureTrace:
  no load exposure → first real load = exam → corridor fracture
================================================================================
7) NO REPAIR LOOP (FEEDBACK WITHOUT RE-TEST)
PATTERN NV7: FEEDBACK_DECAY
Belief:
  "Teacher marks; student reads; improvement happens."
Behavior:
  - corrections are seen once, not converted into drills
  - no error log; no re-test gate
Result:
  repeated same errors for months
Exam effect:
  - persistent grammar slips in writing
  - repeated weak question types in Paper 2
  - oral fillers + unclear stance remain
FailureTrace:
  feedback (information) without repair (behavior change) → no system update
================================================================================
8) WRONG SEQUENCING (BUILDING LATE-STAGE ARTEFACTS FIRST)
PATTERN NV8: START_WITH_POLISH
Belief:
  "Start with fancy vocabulary, then structure later."
Behavior:
  - decorate sentences before installing binds/corridors
Result:
  pretty fragments, incoherent paragraphs
Exam effect:
  - Situational writing fails task fulfilment (purpose/audience/context)
  - Continuous writing lacks cohesion
FailureTrace:
  surface upgrade without corridor base → collapse under real constraints
================================================================================
9) PCCS GAP (PRE-CAREER BINDS NEVER INSTALLED)
PATTERN NV9: PCCS_BIND_MISSING
Belief:
  "School alone can install reasoning language."
Behavior:
  - home life lacks “why/because/therefore”, sequencing talk, repair scripts
Result:
  student’s bind tokens are underused; explanations feel unnatural
Exam effect:
  - oral interaction weak elaboration/qualification
  - writing lacks reasoning connectors
FailureTrace:
  weak PCCS binds → weak life corridors → weak academic transfer
================================================================================
10) NON-WORKING TUITION: SENSOR PROFILE (HOW TO SPOT IT FAST)
NonWorkingSignature :=
  - ScoreVariance high (same student swings wildly between attempts)
  - TransferCollapse: new topic/text type → performance drops sharply
  - TruncationEvents: paragraphs become lists under time
  - ErrorTypeRepeats: same 3 grammar errors reappear every script
  - SummaryLifts: copied phrases; low own-words control
  - OralOverrun: cannot hold 2-minute planned response structure
================================================================================
11) MINIMUM REPLACEMENT (ONE SCREEN; OPTIONAL INSERT)
ReplaceWith :=
  - vocab = precision tokens + bind tokens + corridor reps (not lists)
  - every correction must become a drill + re-test gate
  - practice must include load (timed) + perturbation (novel prompts)
  - comprehension must include evidence binds (line→why)
  - summary must include compression corridor (group→rephrase→trim)
================================================================================
END