O-Levels English OS (Almost-Code, V1.1) — Full Pillar Page (Paste-Ready)

OLevels_English_OS := a reliability-and-variance-control system that ensures
a student can repeatedly produce exam-grade outputs under:
  time pressure + topic variation + strict marking,
by upgrading input→processing→output→feedback→repair loops.

P3(O-Level English) = stable grade outcomes across:
  - unseen comprehension (literal + inferential)
  - summary (compression without meaning loss)
  - situational writing (purpose + tone + audience)
  - continuous writing (structure + language variation)
with low variance across topics and papers.

Score = Skill × StabilityUnderVariation
Most students fail because StabilityUnderVariation is low.

Variation = {
  unfamiliar topics,
  different passage styles,
  changing question traps,
  time pressure,
  fatigue,
  stress,
  stricter marking
}

OLevels_English_OS {
  Inputs:
    - reading exposure (narrative + nonfiction)
    - vocabulary/chunk/structure training
    - exam-format timed practice
  Engines:
    - Vocabulary_OS (P3 retrieval under timer)
    - Comprehension_OS (evidence discipline)
    - Summary_OS (compression + paraphrase control)
    - Writing_OS (situational + continuous)
    - ControlLayer (timing, planning, checking, anti-freeze)
  Sensors:
    - variance sensors (std dev, collapse events)
    - error tags (misread / over-infer / vocab gap / tone mismatch)
    - timing sensors (finish rate)
  RepairLoops:
    - targeted drills + rewrite + post-mortem
  Output:
    - stable exam performance (reduced variance)
}

Goal: accurate answers + stable inference without guessing.

- can answer with evidence lines
- wrong answers are rare and explainable
- inference remains consistent after re-read

Goal: capture required points within word limit, paraphrased cleanly.

- points captured ≥ target band
- zero meaning leakage
- minimal repetition
- obeys word limit consistently

Goal: write for audience/purpose with correct register and structure.

- addresses task fully
- tone correct (formal/informal/persuasive)
- clear paragraphing
- language upgrades appear naturally

Goal: coherent story/argument with variation, imagery, and control.

- plan quickly
- write on time
- maintain coherence
- vocabulary variety + strong verbs
- low grammar drift

FM1: Evidence-free answering (guessing inference)
FM2: Misread details under speed
FM3: Summary leaks meaning / misses points
FM4: Tone/register mismatch in situational writing
FM5: Writing collapses under timer (freeze events)
FM6: High variance across papers/topics (unstable system)

INF_Sensors {
  WrongAnswerTags distribution:
    - misread
    - vocab gap
    - over-infer
    - evidence not used
  ReReadStability (same answer after re-read)
  EvidenceLineHabit (Y/N)
}

SUM_Sensors {
  PointsCaptured / PointsRequired
  MeaningLeakEvents
  WordLimitCompliance rate
  RepetitionRate
}

W_Sensors {
  TaskCompletionScore
  ToneMatchScore
  CoherenceScore
  GrammarErrorRate
  UpgradeCountCorrect (>= 12 per full piece)
}

CTRL_Sensors {
  TimeToStartWriting (<2 min)
  FinishRate (completes on time)
  FreezeEvents (count)
  Last10MinQualityDrop (Y/N)
}

VAR_Sensors {
  ScoreStdDev across 5 timed sets
  CollapseEvents under pressure
  TopicSensitivity (drop on unfamiliar topics)
}

ControlLayer {
  - fast planning
  - time boxing
  - question-type recognition
  - evidence anchoring
  - check-and-fix loop (last 5–8 minutes)
  - anti-freeze start routine
}

StartRoutine {
  1) write 3 bullet points immediately
  2) choose first sentence template
  3) begin writing even if imperfect
}

Loop_Evidence {
  do unseen set
  for each Q:
    - answer
    - underline evidence line(s)
    - rewrite answer in own words
  tag wrong answers
  repair dominant tag weekly
}

Loop_Summary {
  1) list required points (no paraphrase)
  2) compress points into sentences
  3) paraphrase using synonym families
  4) check meaning drift
  5) trim to word limit
}

Loop_Situational {
  write 1 task
  rewrite intro + conclusion only
  check:
    - audience markers
    - formality markers
    - persuasive markers
}

Loop_Continuous {
  5-box plan (2–4 min)
  timed full piece
  inject upgrades:
    - 8 words + 4 chunks + 2 structures
  repair:
    - rewrite weakest paragraph
}

Loop_TimedSet {
  do timed exam section
  post-mortem:
    - tag failure modes
    - choose one dominant leak
    - run one repair loop immediately
}

- build chunk bank + paraphrase families
- start evidence discipline
- timed paragraphs 2×/week

- summary drills 3×/week
- situational tone weekly
- weakest paragraph rewrite every writing session

- weekly timed set + post-mortem
- eliminate dominant wrong-answer tag
- increase speed without losing accuracy

- practice across varied topics
- train late-session fatigue stability
- lock in start routine + time boxing

More worksheets → recognition rises (P1)
→ weak evidence discipline
→ summary still leaks
→ tone errors persist
→ exam collapse events remain
→ variance stays high
→ grade does not improve reliably

O-Level English improves when you repair the dominant leak:
evidence discipline / compression / tone / load stability,
then train under variation.

Secondary English OS builds the engines.
O-Levels English OS trains stability: low variance under exam variation.

If vocabulary disappears under the timer, the student is not P3 yet.
Use Vocabulary OS to convert recognition into load-stable retrieval.

The same closed-loop repair logic applies:
diagnose leaks → train structure → train load → transfer to exam.