The Inspiration

We've all been there—staring at ChatGPT, typing "I need advice on..." and getting generic responses that could apply to anyone. The AI doesn't know you. It doesn't know your values, your projects, or what keeps you up at night.

This frustration sparked Clarity Coach. We asked ourselves: What if AI could truly know you? Not just remember your last message, but deeply understand your identity, values, and goals—then provide advice that's authentically yours.

The vision was simple but powerful: Build a personal AI operating system, not just another chatbot.


What We Built

Clarity Coach is a context-aware AI coaching platform that understands who you are at your core.

Core Features:

  • 🎯 Deep Context Vault
    Users answer thoughtful questions during onboarding about their values, beliefs, current projects, and future vision. This context becomes the foundation of every conversation—making each response deeply personalized.

  • 🤖 Four Specialized Coaches
    Each coach uses a different LLM optimized for its specialty:

    • Clarity Coach – Cut through confusion and find focus (GPT-4)
    • Productivity Coach – Build systems that actually work (Claude)
    • Creativity Coach – Unlock creative potential and ship faster (Gemini)
    • Wellness Coach – Sustainable habits for mind and body (GPT-3.5-turbo)
  • ✨ Custom Coach Creation
    Users can create unlimited custom coaches by answering a few questions. Our AI selects the best LLM and configures the coach automatically. Even better—coaches can be shared via codes, creating a community-driven coach marketplace.

  • 💾 Persistent Conversation History
    Every conversation is saved per coach. Users can return anytime, review past sessions, or continue where they left off.

  • 🔄 Living Context System
    User context isn't static—it evolves. Users can update their values, add projects, or refine their goals anytime in Settings. These updates instantly improve all coaching responses.


How We Built It

Architecture

We chose Flutter for cross-platform consistency and BLOC for predictable state management. The architecture follows clean principles with clear separation:

lib/
├── features/          # Feature modules (auth, coaching, onboarding)
├── core/             # Shared utilities, theme, widgets
├── services/         # Firebase, AI, local storage
└── domain/           # Entities and business logic

Multi-Model AI Strategy

This was our secret weapon. Instead of one-size-fits-all AI:

  • Clarity Coach uses GPT-4 for nuanced decision-making
  • Productivity Coach uses Claude for structured thinking
  • Creativity Coach uses Gemini for divergent ideation
  • Wellness Coach uses GPT-3.5-turbo for empathetic responses

For custom coaches, we built an AI model selector that analyzes the coach's purpose and automatically picks the optimal LLM.

Context Injection System

Every message sent to any coach includes:

  • User's core identity and values
  • Active projects and goals
  • Past decisions and learned lessons
  • Current life season and priorities

This transforms generic AI into a personal advisor who truly knows you.

Tech Stack

  • Frontend: Flutter + BLOC
  • Backend: Firebase (Auth, Firestore, Analytics)
  • AI Layer: Multi-model approach (OpenAI, Anthropic, Google)
  • Storage: Local-first with cloud sync
  • Monetization: RevenueCat (ready for premium tiers)

Challenges We Faced

  1. Context Without Overwhelm
    Early prototypes asked too many questions upfront. Users felt interrogated, not understood.
    Solution: We redesigned onboarding as a conversation, not a form. Progressive disclosure—ask only what's needed now, offer depth optionally. The result? 85% completion rate.

  2. Prompt Engineering at Scale
    Each coach needed unique system prompts that reference user context without being repetitive or robotic.
    Solution: Built a PromptBuilder utility that dynamically injects relevant context based on conversation topic. If discussing career decisions, it emphasizes values. If discussing daily tasks, it highlights current projects.

  3. Multi-Model Cost Management
    Using multiple LLMs could get expensive fast.
    Solution: Smart caching, conversation summarization after 10 messages, and strategic model selection (GPT-4 only when nuance matters, GPT-3.5 for routine responses).

  4. Making Context Feel "Alive"
    Static profiles feel dead. We needed context to evolve naturally.
    Solution: Implemented "context suggestions" where the AI notices patterns ("You've mentioned burnout 3 times this week—want to add wellness to your priorities?") and gentle reminders ("It's been 30 days since you updated your context").

  5. Coach Sharing Mechanics
    How do you let users share custom coaches without compromising privacy?
    Solution: Generate unique coach codes that contain only the coach's configuration (prompts, model choice, personality)—never the creator's personal context. When imported, the coach adapts to the new user's context automatically.


What We Learned

Technical Lessons

  • Local-first architecture is powerful: Users never lose data, even offline
  • BLOC scales beautifully: Complex state becomes predictable
  • Multi-model beats single-model: Each AI has strengths—use them strategically
  • Context is everything: Same question, different context = wildly different advice

Product Lessons

  • Onboarding is make-or-break: Get this wrong, and users bounce
  • People want to be understood: Generic AI feels cold; personalized AI feels magical
  • Simplicity scales: Four coaches beat fifty mediocre ones
  • Community creates value: Shared coaches could become the app's killer feature

Personal Growth

We went from "let's build a chatbot" to "we're creating a personal AI operating system." That mindset shift changed everything—the architecture, the UX, the ambition.


What's Next

Clarity Coach is just the beginning. We're building toward:

  • 📊 Context Analytics
    "Your top value is Growth—but 80% of your time goes to Stability tasks. Here's the disconnect."

  • 🤝 Accountability Partners
    Share context with a friend. Check in on each other's progress. Hold space together.

  • 🎙️ Voice Mode
    Coaching conversations while walking, driving, or cooking.

  • 🏪 Coach Marketplace
    Discover community-created coaches. Download a "Startup Founder Coach" or "Marathon Training Coach" with one tap.

  • 🔗 Integration Ecosystem
    Connect with calendars, note apps, and project management tools. Context flows everywhere.


Why This Matters

We're living in the age of AI, but most AI tools treat us like strangers. They optimize for engagement, not understanding. They remember your last prompt, not your life story.

Clarity Coach is different. It's not about more AI—it's about better AI. AI that knows you're a builder who values integrity. AI that remembers you're balancing a startup launch with family time. AI that speaks to you, not some statistical average.

This is personal AI, done right.
This is what coaching should feel like in 2025.
This is Clarity Coach.


Built with ❤️ for the Shipyard Competition

Built With

Share this project:

Updates