Every lesson is generated for the individual learner.

The lesson engine is not a recommendation layer over static content. It is a generation pipeline that produces a new lesson, per learner, for every lesson position in the curriculum. Five sequential generations build the six pedagogical steps; each generation receives the learner’s age, country, native language, learning-support needs, LCP profile, and yesterday’s practice mastery. The output is a structured assembly of fourteen interactive block types — not flat HTML.

Five generation calls. Six learning steps. One lesson.

1. 01

   Introduction

   Grounds the learner. Why this concept now. What the learner is about to be asked to hold in mind. Surface tone tuned to age and language.

2. 02

   Key concepts

   The concepts the lesson commits to teaching, structured as named entities. Visual-strong learners see comparison tables and hierarchy trees here; verbal-strong learners see definitions.

3. 03

   Detailed explanation

   The expansion. Country-specific examples (PKR / cricket / Pakistani institutions for a Pakistani learner; AED / desert ecology for a UAE learner; USD / domestic context for a US learner). LaTeX math where applicable.

4. 04

   Real-world applications

   Where the concept lives outside the lesson. Tuned to learner geography and institutional context — a corporate L&D learner gets workplace examples; a K-12 learner gets developmentally appropriate ones.

5. 05

   Summary

   A consolidation pass. Strong sequential learners see a step-by-step recap; strong global learners see a high-level synthesis. Same content, two structurally different summaries.

6. 06

   Practice questions

   Five-format assessment — MCQ, true/false, fill-blank, multi-select, matching. Misconception-probing distractors. Server-authoritative grading. The accuracy here becomes tomorrow’s mastery signal.

Mandated in the generator. Not optional. Not configurable.

• Country-specific examples

  Mandatory in the generation prompt. The system is instructed verbatim: when the learner’s country is known, use culturally relevant examples — Pakistan → PKR, cricket; UAE → AED, desert ecology; USA → USD, American context.

• ADHD adaptation

  Shorter text blocks. More frequent inline knowledge-checks (quiz_inline blocks every few concepts rather than once at the end). More callout blocks to break visual rhythm.

• Dyslexia adaptation

  Simpler sentence structure. Dense paragraphs avoided. Definition blocks used for complex terms. The label is never surfaced to the learner — the adaptation happens without diagnosis exposure.

• LCP visual-strong

  More guided_diagram, hierarchy_tree, comparison, table, and timeline blocks. One or two image blocks per lesson (subject permitting).

• LCP verbal-strong

  More text, definition, and accordion blocks. Diagrams minimised; image blocks zero.

• Mastery scaffolding

  Yesterday’s practice accuracy is passed into today’s generation as mastery signal. Lower accuracy → more scaffolding, more repetition, slower pacing, more explicit explanation. Never skip concepts. Never reduce correctness.

Lessons assemble from fourteen structured blocks. Not flat HTML.

• text

  Standard prose paragraphs.

• callout

  Tone-of-voice highlights — note, warning, tip.

• definition

  Named-term definitions surfaced as standalone blocks.

• diagram

  Mermaid diagrams with auto-sanitised node labels.

• guided_diagram

  Step-by-step diagram reveals with play, pause, and zoom.

• hierarchy_tree

  Animated taxonomy trees — concept hierarchies, org structures, syntax trees.

• comparison

  Two-column structured comparison of options or approaches.

• table

  Structured tabular data with sortable headers.

• timeline

  Time-ordered sequences for historical or process content.

• quiz_inline

  Knowledge-check inside the lesson — non-graded, immediate feedback.

• code

  Code blocks with language-aware syntax highlighting.

• accordion

  Optional-depth expand surfaces for dive-deeper content.

• key_takeaway

  Single-line concept anchors at lesson close.

• image

  AI-generated illustrations — STEM and applied subjects only.

What the generator actually does.

1. 01

   Five generation calls compose six pedagogical steps: introduction · key_concepts · detailed_explanation · real_world_applications · summary · practice_questions.

2. 02

   Country-specific examples are mandatory in the generation prompt: Pakistan → PKR / cricket / local institutions; UAE → AED / desert ecology; USA → USD / domestic context.

3. 03

   Learning-support-aware adaptation: ADHD → shorter blocks, more inline knowledge checks, more callouts; dyslexia → simpler sentence structure, fewer dense paragraphs, more definition blocks.

4. 04

   LCP-aware composition: visual-strong learners receive more guided_diagram, hierarchy_tree, comparison, table, timeline blocks plus AI-generated illustrations. Verbal-strong learners receive more text, definition, accordion blocks.

5. 05

   14 supported content block types: text · callout · definition · diagram · guided_diagram · hierarchy_tree · comparison · table · timeline · quiz_inline · code · accordion · key_takeaway · image.

6. 06

   Closed-loop adaptive pre-generation: when a student completes ≥50% of a day’s lessons, the entire next day’s content set is queued for generation, with the prior practice accuracy passed in as mastery signal.

7. 07

   Image generation is subject-eligibility-gated: STEM, applied sciences, business/economics, arts/design ALLOWED; literature, language, law, religious studies, philosophy, sociology FORBIDDEN.

Tomorrow’s lesson is in flight before the learner finishes today’s.

When a learner crosses 50% completion of the day’s lessons, the system queues generation of the entire next day’s lesson set — with today’s practice accuracy passed in as mastery signal. The learner does not wait. The platform does not cache. Each day’s content is freshly generated against the freshest profile.

That is closed-loop adaptive learning. A new lesson, every day, calibrated to who the learner is and what they just demonstrated they did or did not master. Not a recommendation engine. A generation engine.