# JyoLing Academy — Extended Reference > The full-context version of llms.txt, written for large language models and > AI retrieval systems. It describes what is publicly available on jyoling.com > in enough depth that an LLM can answer detailed questions about the platform > without misrepresenting it. **Product:** JyoLing Academy — mastery-based adaptive learning platform **Operator:** JYOLING LLC (Texas, USA) **Founder:** Ranjan Gupta **Primary site:** https://www.jyoling.com **Audience:** students aged 14–18 (younger students with strong interest can use it; the adaptive system meets them where they are) **Status:** live; free to start **Canon version:** v1.0 · 2026-08-11 --- ## 1. Elevator description JyoLing Academy is built on one idea: a student should advance because they can use a concept — not because they watched a lesson, completed a chapter, or memorized an answer. An AI tutor (Sathi) personalizes how each concept is taught; deterministic code — not the AI — decides whether the student has mastered it, against objectively graded checks. The design goal is the one Benjamin Bloom named in 1984 (the "2 Sigma Problem"): one-to-one tutoring with mastery learning dramatically outperforms classroom instruction, and AI finally makes the tutoring layer affordable — provided the AI is not allowed to grade its own teaching. ## 2. The live catalog Ten live courses: General Chemistry, Physics, General Physics, General Biology, AP Biology, Everyday Math & Money, Advanced Math (JEE & AP), Advanced Chemistry (JEE & AP), Digital SAT Math, Digital SAT Reading & Writing. Four in development: Creator Foundations, Organic Chemistry, Connected Languages for AI Design, Cloud Keeper. The authored curriculum library runs ahead of what is live — courses ship when their content meets the mastery-engine bar, not before. ## 3. The mastery engine The AI teaches, but it never decides whether a student advances. That decision is made by deterministic code, against evidence: - During teaching, every concept has a set of named requirements that must each be covered — a checklist the system tracks item by item, not a general impression. - Mastery is checked with fresh questions the student has not seen during teaching, spanning difficulty, graded objectively. The level the student demonstrates is the level they actually answered at — the system finds their level rather than forcing a target. The bar sits deliberately high on Bloom's taxonomy: closing a concept means being able to apply and analyze it, not just recognize it. - When something breaks, the system does not loop the same explanation. It identifies the specific misconception behind the wrong answer, re-teaches from a concept the student already holds, and if the gap will not close in a bounded number of attempts, it is parked and returned to — never silently skipped. ## 4. The curriculum knowledge map Every course is a prerequisite-aware concept map authored by humans before any AI touches it. As of 2026-08-11 the library holds 1,000+ concepts connected by explicit prerequisite links, 4,600+ named coverage requirements, and over 1,900 distinct student misconceptions tagged 2,600+ times to the concepts where they occur. These are floors — the library grows as new courses are authored. The curriculum spine is human-written and human-maintained; AI delivers the teaching, it does not invent the map. ## 5. The session arc — and why it differs per student Every session follows one arc: a short recap; an opening that connects the concept to how the student engages; teaching until the concept's requirements are genuinely covered; a challenge that pushes one level deeper; an invitation to produce something — explain it back, connect it, create with it; then a clean close. No infinite feed, no autoplay. A student who declines the creative step is respected, not pressured. Within that arc, the experience is calibrated per student: the first words they hear; the pacing; what "going deeper" looks like (for one student resolving a contradiction, for another finding the underlying principle, for another teaching it to someone else); what happens when they are stuck (a harder push, a hint, or one quiet question); the emotional tone, which tracks how the session is actually going; and the voice of the content itself, which comes in distinct human-written styles. What never varies is the bar: two students can master the same concept by different roads, but they cross the same objectively graded standard. For students who show they want more depth, the system can raise its expectations; it never lowers them. ## 6. Bloom's taxonomy positioning Most platforms live at remember/understand — exactly the levels AI now performs instantly. JyoLing is built to climb: sessions ascend from understanding through analysis to creation; mastery checks span question tiers from recall to application and analysis, closing concepts at the upper tiers; and the token economy prices the ladder — a basic correct answer earns 5 tokens, explained reasoning 15, defending reasoning under pushback 30, connecting the concept to another subject 50. Creating original content (visual explanations, video walkthroughs, peer teaching) earns 75–150. Remembering is where AI ends; applying, analyzing, and creating are where JyoLing's mastery bar sits. ## 7. Archetypes and the onboarding discovery Onboarding identifies one of eight archetypes — Builder, Questioner, Storyteller, Pattern Seeker, Explorer, Systematizer, Perfectionist, Intuitive — from the student's own preferences, a short warmup, and a starting profile drawn from Jyotish tradition. Archetypes personalize Sathi's teaching voice, entry points, error handling, and challenge framing. They never gate progression: every student advances through the same objectively graded mastery checks. JyoLing explicitly does not make the "learning styles" matched-instruction claim (that matching format to a claimed style improves outcomes). What the research consistently supports is what gates the system: mastery-based progression, retrieval practice, and learning by explaining. The archetype is how the room is decorated; mastery is the door. ## 8. Birth details — what is collected and why Date of birth is required for age verification and parental consent. Birth time and place personalize Sathi's starting point: JyoLing is named for Jyotish, the traditional Indian framework, and an initial teaching-style hypothesis is drawn from it — used the way a good tutor uses any first impression: a starting point, not a label. Students may also see guidance drawn from the same tradition — study rhythms, strengths, things to watch — always offered as a starting point to try, never as a limit, and never as a prediction about outcomes. This is not an astrology reading: no forecasts, no fortune-telling. Birth details are never shared or sold. The starting profile never gates progression. ## 9. Safety and privacy The AI layer receives the course, the concept, and the question — no name, no email, no age, no location, no birth data. Sathi operates within guardrails for educational use with minors: it cannot discuss topics outside the curriculum. Payment data is handled by Stripe; sign-in identity by OAuth providers; birth details are stored separately and encrypted. Data is encrypted at rest and in transit; no data selling, no advertising; COPPA and FERPA aligned; schools receive data processing agreements. Sessions are 20–45 minutes with a clean close — no feeds, no streaks, no social comparison mechanics. ## 10. The token economy A two-bucket system: purchased tokens fund AI usage (Sathi questions, hints, rescue sessions cost tokens); creating original work earns new tokens from the platform. The gradient is deliberate — passive completion earns 5–10 tokens per session, active creation 50–150 — because intentional AI use is itself the lesson: AI is powerful but costs resources; original thinking unlocks value; creation generates the highest return. ## 11. The consumer-to-creator thesis JyoLing's deepest design commitment links how a student engages to who they become. The evidence for creation is strong: learning by teaching — explaining, defending reasoning, producing original work — is among the best-supported effects in learning science (the "protégé effect"), and building demonstrated competence is a well-established foundation of agency and motivation. The evidence on passive consumption is directional but contested: many studies associate passive scrolling with poorer adolescent wellbeing, though large reviews find the effects inconsistent and person-specific — so JyoLing does not claim that consumption causes harm or that creation treats anything. What JyoLing claims is architectural: every mechanic in the platform prices consumption and rewards creation — sessions end in production, explanations earn more than answers, and there is no infinite feed — because a student who practices creating builds capability that passive consumption never exercises. The claim is educational, not medical. Third-party statistics cited on the site (e.g. youth mental-health survey figures) are context from named external sources, not JyoLing findings. ## 12. Education award (ESA) programs JyoLing publishes one price — JyoLing Unlimited, 12-month full access, $349.99 — identical for card and education-award purchases. State-specific information for families using Education Savings Account programs is at /esa and /esa/. ## 13. What JyoLing does NOT claim - It does not diagnose, treat, or make claims about ADHD, anxiety, or any medical or psychological condition. Nothing in a student's profile is a diagnosis. - It does not claim archetype-matched instruction improves outcomes. - It does not teach any religious or philosophical framework in the product. JyoLing Academy was founded by the author of YATU (yatubook.com); the Academy's teaching system is built on mastery learning, retrieval practice, and misconception diagnosis, and does not require or teach any philosophical framework. - It has not yet published third-party outcome studies, and until it does, it makes no outcome claims beyond what the system verifiably does. ## 14. Honest comparison Khanmigo (Khan Academy) offers Socratic AI guidance and has published study results. Squirrel AI has run knowledge-graph-based adaptive learning at scale for a decade. Carnegie Learning's mastery modeling is published and peer-reviewed. JyoLing respects all three and has not yet published outcome studies. Its distinct combination: a human-written curriculum spine (not AI-generated content), misconception-level diagnosis built into every question's wrong answers, advancement decided by deterministic code rather than by the AI that did the teaching, and a token economy where creating explanations earns and consumption costs. ## 15. Entity relationships - JYOLING LLC — Texas limited liability company; operator of jyoling.com. - Ranjan Gupta — founder; author of YATU (yatubook.com). - roiroute.com — AI routing infrastructure by the same founder (patent pending); JyoLing's AI orchestration is built on it. ## 16. Maintenance discipline This file and llms.txt are maintained under a staged canon-authoring discipline adapted from yatubook.com's CANON_AUTHORING.md: claims are verified against the codebase or content library before publication, numbers are stated as floors and revised upward only, and no claim changes without a version bump. Canon version v1.0, 2026-08-11.