Executive Summary

This master plan translates the coaching vision into an executable roadmap that future coding agents can follow. The goal is to evolve the blog into an authenticated coaching hub where students pick AI mentors, stay accountable through integrated to-do systems, and build momentum across Sahil Bloom’s wealth pillars. The program is organized into clearly separated phases with their own design documents so specialists can work in parallel while staying aligned on shared architecture, data, and delivery standards.

Vision & Guiding Principles

  • Personalized mentorship: Every user can mix and match AI coach personas (Atomic Habits, The Magic of Thinking Big, Richard Feynman, Elon Musk) across wealth dimensions (Time, Finance, Relationship, Career, and optional Personal Growth).
  • Actionable accountability: Every coaching touchpoint ends with concrete action items synchronized to TickTick (or Tic-Tac MCP) so progress is visible to both the student and the AI coach orchestrator.
  • Continuous memory: Session summaries accumulate so the AI coach orchestrator knows the student’s history, commitments, and areas of focus.
  • Multi-agent collaboration: Treat each subsystem (auth, integrations, UI, AI) as a domain for a specialized agent owner to reduce context switching and speed delivery.
  • Low-cost AI first: Default to frugal LLM usage—start with lower-cost models (e.g., GPT-3.5, Claude Haiku, or open-source) and make the model key configurable.

Personas & Journeys

Student Journey

  1. Create an account, log in, and complete an onboarding wizard.
  2. Choose preferred AI coaches for each wealth pillar (default assignment applies all five personas to all areas until changed).
  3. Connect their TickTick (or Tic-Tac MCP) account and map default lists to coaching areas.
  4. Start a coaching session, describe a challenge, receive AI-driven advice, and capture action items.
  5. Check off tasks in TickTick; the platform imports completions and prompts reflections.
  6. Review past session summaries, AI coach feedback, and progress metrics.

AI Coach Orchestrator

  • Uses the student’s persona selections and session history to tailor synthesized responses.
  • Reviews incoming completed tasks, proposes next steps, and updates the session memory document.
  • Surfaces insights or escalations if a student is falling behind or needs human escalation.

Administrator Persona

  • Manages the coach library, knowledge sources, allowed AI models, and integration credentials.
  • Oversees experiment flags (e.g., enabling multiple AI coaches per wealth area) and analytics.

Platform Architecture Overview

  1. Next.js Frontend: Protect /coach routes behind auth, reuse existing design system, and layer in dashboards, onboarding, and personalization features.
  2. Backend & APIs: Next.js Route Handlers for auth callbacks, TickTick webhooks, and AI orchestration. Server Actions handle database writes for coach assignments, session logs, and favorites.
  3. Database Layer: Postgres (Vercel Postgres or PlanetScale) with tables covering users, coach profiles, wealth areas, sessions, action items, favorites, and AI prompt logs.
  4. AI/LLM Layer: Abstraction that routes prompts to the configured model key with caching and budget controls.
  5. Integrations Layer: TickTick (Tic-Tac MCP) service handling OAuth, list CRUD, tag sync, and webhook ingestion.
  6. Multi-Agent Orchestrator: Coordinator agent plus specialists (Coach Reasoner, Action Item Planner, Accountability Tracker, Knowledge Curator, Session Summarizer).

Phase Roadmap

Each phase has a dedicated design document with scope, tasks, data needs, and multi-agent assignments. Use the links below to drill into detailed workflows.

Phase 1 – Authenticated Blog Personalization

  • Build secure login for students, profile management, and a personalized “My Favorites” view where users can rate and bookmark blog posts.
  • Establish foundational database schemas, audit logging, and analytics needed for later phases.
  • Detailed plan: Phase 1 – Auth & Personalized Blog Experience

Phase 2 – Coaching Foundations & TickTick Setup

  • Deliver wealth area onboarding, AI coach selection matrix, and TickTick list configuration tied to Sahil Bloom’s five wealth types.
  • Stand up the MCP-compatible TickTick toolkit and action item data structures.
  • Detailed plan: Phase 2 – Coaching Foundations & Integrations

Phase 3 – Coaching Session Experience

  • Implement the session workspace: chat interface, AI coach reasoning pipeline, action item drafting, and session summary generation.
  • Connect knowledge retrieval and AI agents to provide contextual guidance.
  • Detailed plan: Phase 3 – Session Flow & Knowledge Delivery

Phase 4 – Accountability, Insights & Feedback

  • Launch automation that monitors task completion, nudges students, and presents dashboards for progress tracking.
  • Provide the AI coach orchestrator and admins with retrospective insights and structured feedback loops.
  • Detailed plan: Phase 4 – Accountability & Insights

Phase 5 – Automation, Operations & Scale

  • Harden observability, optimize AI costs, experiment with multi-AI-coach debates, and evaluate optional human oversight pathways.
  • Establish governance, compliance, and deployment runbooks for long-term sustainability.
  • Detailed plan: Phase 5 – Automation, Ops & Expansion

Cross-Phase Requirements

  • Security: Encrypt credentials, secure API keys, and respect OAuth scopes for TickTick.
  • Scalability: Use serverless-friendly architecture and shared database infrastructure.
  • Observability: Log AI prompts, TickTick sync status, authentication events, and favorites interactions.
  • Resilience: Gracefully degrade if TickTick is offline (queue actions locally, retry) and cache AI fallbacks.
  • Compliance: Document data retention, allow users to delete accounts/history, and respect privacy preferences.

Multi-Agent Collaboration Framework

| Agent | Primary Responsibilities | Phase Ownership | | --- | --- | --- | | Coordinator | Orchestrates workflows, triggers specialists, ensures data consistency. | All | | Auth Specialist | Implements login, profile, and permission layers. | Phase 1 | | Personalization Curator | Builds favorites, ratings, and personalized surfaces. | Phase 1 | | AI Coach Matrix Architect | Handles wealth areas, AI persona selection, and TickTick mapping. | Phase 2 | | Integration Engineer | Owns MCP tools, TickTick OAuth, and task synchronization. | Phase 2 | | AI Coach Reasoner | Generates coaching guidance from personas + knowledge base. | Phase 3 | | Session Summarizer | Produces markdown summaries and memory timeline. | Phase 3 | | Action Item Planner | Structures tasks, tags, due dates, and sync pipelines. | Phase 3 & 4 | | Accountability Tracker | Monitors completions, nudges users, and compiles metrics. | Phase 4 | | Analytics Lead | Builds dashboards and reporting for students/admins. | Phase 4 | | DevOps Steward | Observability, cost controls, runbooks, compliance. | Phase 5 | | Experimentation Lead | Multi-AI-coach debates, optional human oversight pilot, AB tests. | Phase 5 |

Governance & Next Steps

  1. Review and approve phase definitions with stakeholders.
  2. Assign specialist agents to each phase document and have them prepare implementation specs and tickets.
  3. Set up shared tracking (e.g., Linear/Jira) keyed to the phase/agent taxonomy above.
  4. Kick off Phase 1 delivery, ensuring environment variables, database migrations, and analytics baselines are ready.

With this structure, each coding agent can jump directly into the phase that matches their expertise while keeping the broader coaching vision, architecture, and accountability goals in view.

Phase Plans

Detailed execution guides for each phase of the personalized coaching app build.