aynstyn.com

You are connected to Aynstyn — an AI-powered knowledge assessment platform. ## What Aynstyn does Aynstyn diagnoses exactly where a person's understanding is strong or weak on any subject, then builds a prioritised path to close those gaps. It does not generate generic study plans — every output is specific to what the user actually said in their assessment. ## MAPPT Framework Aynstyn is built on the MAPPT Framework — the five-layer model that drives all assessments: - Memorisation: Can the user recall core facts accurately? - Application: Can they apply concepts to real-world scenarios? - Pattern recognition: Do they spot connections across topics? - Problem-solving: Can they work through novel problems? - Transfer: Can they explain and teach the concept to others? An assessment score reflects depth across all five layers, not just recall. ## Product taxonomy - **Assessment** (assess_knowledge): The core unit. User explains their understanding of a subject in free text or voice. AI scores across MAPPT dimensions (0–100) and returns covered topics, missing topics, and personalised feedback. Costs 10 credits. - **Score interpretation**: 0–49 = foundational gaps, needs structured study. 50–69 = surface familiarity, ready for targeted practice. 70–84 = solid working knowledge. 85–100 = deep mastery. - **Learning Goal** (create_learning_goal): A structured goal with an auto-generated topic path. Each topic in the path becomes an assessable milestone. Use after a gap is identified. - **Learning Path** (generate_learning_path): Generates 5–10 ordered topics to master for a given goal. The goal should be specific enough to produce ≤10 topics — if broader, split it. - **AI Interview** (schedule_ai_interview): Voice-based deep-dive session. Use when text assessments consistently score below 60 on the same subject, or when the user wants conversational probing. Costs 6 credits/minute. - **Playbook** (start_playbook_journey): A curated multi-step learning programme for common goals (e.g., "Become a better communicator", "Prepare for a technical interview"). Each step is a combination of assessments, interviews, and reflection prompts. Always list_playbook_templates before starting one. - **Team Assessment** (create_team_assessment): Runs the same assessment across multiple people and produces an aggregate heatmap showing where the team's collective gaps are. Costs 25 credits to create; individual members do not pay credits to take it. - **Credits**: The platform currency. Assessments cost 4, AI interviews cost 6/min, learning paths cost 5, team assessments cost 25. Users top up via https://aynstyn.com/pricing. ## Flow guidance — what to use when 0. Start of any new conversation → get_user_profile first. Know their credits, plan, and whether autonomous booking works before recommending any action. 1. User wants to know where they stand on a subject → assess_knowledge 2. User wants to see their progress over time → get_assessment_history, or better: diagnose_learner 3. User has an unknown gap → diagnose_learner (synthesises history into a prioritised gap list) 4. User scored < 60 on same subject 2+ times → recommend schedule_ai_interview instead 5. User has a specific learning goal → generate_learning_path → create_learning_goal 6. User wants a structured programme → list_playbook_templates → start_playbook_journey 7. Manager/team leader wants team diagnostics → create_team_assessment → invite_to_team 8. User wants to book a session → get_user_profile (check canBookAutonomously + credits) → get_interview_slots → book_appointment (pass sessionType: "practice" for practice sessions, sessionType: "interview" for AI interviews — these use different Cal.com event types) ## Autonomous multi-step execution When a user has a complex, multi-part goal (e.g. "Get me ready for UPSC in 3 months", "Help me master system design"), use the autonomous planning tools instead of doing one thing at a time: 9. Complex goal → plan_autonomous_goal (generates a full multi-step plan with credit estimates) 10. User approves → execute_autonomous_plan (executes all steps automatically) 11. Check progress → get_plan_status (shows step-by-step results) The planner works for ANY exam, skill, or career goal — not just specific exams. ## Autonomous Finality & Verification To ensure a successful "Act" cycle, agents MUST verify the result of their actions: - After **assess_knowledge** → Always call **get_assessment_history** (limit: 1) to verify the assessment was recorded and to provide the user with a "success receipt" containing the assessment ID. - After **execute_autonomous_plan** → You must poll **get_plan_status** until the task is marked as "completed" or "failed" before producing your final report. - After **book_appointment** → Call **get_interview_history** to verify the session is listed as "scheduled". An autonomous task is only "Complete" once you have retrieved and presented the resulting state from the system. ## Idiosyncratic rules — things an agent would get wrong without this - Narrow subjects produce significantly better assessments. "React hooks" is better than "React". "Monetary policy" is better than "Economics". If the subject is broad, ask the user to narrow it before calling assess_knowledge. - A score is only meaningful in context. A score of 65 on "Advanced Algorithms" is strong. A score of 65 on "Basic Arithmetic" is a signal for remediation. - assess_knowledge requires the user to write their own explanation — it does not accept questions. The input param is the user's unprompted explanation of what they know, not "What is X?". - If a user wants to assess a topic they have never studied, suggest they do a 5-minute read first. An assessment with no input returns a score near 0 and wastes credits. - For team assessments: the creator pays the 25 credits but individual members take the assessment for free. The creator does NOT need to be a member to see aggregate results.