v2.8.12: Universal Subject Intelligence — AI That Knows What You're Studying

Since Claritii launched, the AI has treated every subject the same way. A chemistry flashcard was generated with the same prompt template as an English flashcard, and a mathematics quiz question was marked against the same rubric as a history one. That's always felt a little wrong — because subjects are not the same. You don't learn mathematics the way you learn a language, and you don't study law the way you study biology.

v2.8.12 changes that. This release introduces a Universal Subject Intelligence Engine that detects what subject family you're studying and adapts every AI interaction — content generation, study mode recommendations, assessment rubrics, and tutor behaviour — to match how that family is actually learned.

What Subject Families Exist

We defined 15 subject families that cover every major area of study globally, from primary school through university and professional education:

• Mathematics — Pure and applied mathematics, statistics, logic
• Natural Science — Physics, chemistry, biology, earth science, astronomy
• Social Science — Psychology, economics, sociology, political science, geography
• Humanities — History, philosophy, literature, classics, religious studies
• Language Acquisition — Learning a new language (French, Japanese, Spanish, etc.)
• Language & Literature — Studying your native language's literature and linguistics
• Creative Arts — Art, music, drama, film, design
• Computing — Computer science, programming, data science, cybersecurity
• Engineering — Civil, mechanical, electrical, aerospace, chemical engineering
• Health & Medicine — Medicine, nursing, pharmacy, public health, anatomy
• Business — Finance, accounting, marketing, management, entrepreneurship
• Law — Legal studies, criminal law, contract law, constitutional law
• Applied Trades — Construction, electrical trades, automotive, plumbing
• Physical Education — Sports science, exercise physiology, coaching
• General — Cross-disciplinary or undetected subjects

Every family has a full cognitive profile: how knowledge is represented in that field (concept graphs, system maps, timelines, argument trees), how learning typically progresses, what assessment styles are common, and what study modes are most effective.

How Detection Works

When you create a subject or upload an assessment notification, Claritii runs a two-layer detection process:

1. Keyword matching — The system scans the subject name, assessment name, and topic names against a curated global keyword list covering thousands of subject-specific terms. "Algebra" → mathematics. "Sonnet" → language and literature. "Tort law" → law. "Cellular respiration" → natural science. This is fast and works for the vast majority of cases.
2. AI fallback — When keywords are ambiguous (e.g., "Theory of Knowledge" or "General Studies"), the system routes to the reasoning AI model for classification. This ensures every assessment gets correctly classified regardless of naming conventions.

Detection results are cached per assessment, so the AI classification only runs once. Subsequent content generation sessions reuse the cached family assignment.

What Changes in Practice

For students, the most visible change is that study materials feel more relevant to the subject they're actually studying. Here's what's different by family:

Mathematics. Problem-solving and active recall are prioritised. The AI uses LaTeX extensively for equations and symbols. Assessment rubrics focus on procedural accuracy, proof structure, and logical reasoning. Generate a mathematics quiz, and you'll get multi-step problems with worked solutions — not vocabulary recall.

Natural Sciences. Conceptual understanding and visual learning are emphasised. The AI generates diagram-heavy content, comparisons between related processes, and practical application questions. Rubrics reward mechanistic explanations and evidence-based reasoning.

Humanities. Argument analysis, contextual reasoning, and thematic connections drive content generation. Blurting, mind mapping, and Feynman technique are the recommended modes. The AI expects multi-paragraph responses with evidence and counterargument.

Language Acquisition. Flashcards, cloze exercises, and spaced repetition take priority. Content emphasises vocabulary retention, grammar pattern recognition, and contextual usage. Rubrics reward communicative competence over grammatical perfection in early stages.

Computing & Engineering. Systems thinking and project-based learning are centre stage. Interleaving and active recall are recommended. The AI generates code snippets, system diagrams, and architecture questions alongside theoretical content.

Law. Case-based reasoning and argument evaluation drive content. The AI generates hypothetical scenarios, precedents to analyse, and essay frameworks. Rubrics reward application of legal principles, not just recitation.

Health & Medicine. Clinical reasoning and diagnostic thinking are emphasised. The AI generates patient scenarios, pharmacological calculations, and anatomy identification tasks alongside theoretical content.

Every family gets its own nuance, and every family's rubric criteria are stored in the database — 40 rubric rows across all families and study modes, ensuring the marking engine knows what "good" looks like for each domain.

Database Schema & Architecture

Three new database tables support the subject intelligence system:

• subject_topic_nodes — Hierarchical topic structure per subject with parent/child relationships, level, and metadata. Enables future syllabus-aware content mapping.
• subject_profiles — Per-user subject profiles storing detected family, confidence, and per-subject settings. Links subjects to their universal family classification.
• topic_progress — Per-student tracking of topic-level mastery within each family, enabling future cross-subject analytics and personalised learning path recommendations.

No code changes are needed to add support for a new subject — the keyword list and family profiles are data-driven. Add the keywords to the detection engine, and the rest of the system adapts automatically.

What's Next

The subject intelligence engine lays the foundation for several in-progress features: family-aware exam prediction that accounts for different assessment styles across disciplines, personalised learning paths that adapt within a family based on topic-level mastery, and cross-family analytics that compare how a student performs across different subject types.

v2.8.12 is the first release where Claritii genuinely understands what you're studying — not just what assessment you have, but what kind of learning that subject requires.