PhaseMe
Inspiration
In May, a serious accident left me facing surgery and months of bedridden recovery. Confined to a single room, I needed something meaningful to keep my mind focused, creative, and moving forward.
During that time, my partner became my greatest anchor. Her unconditional support helped me navigate vulnerability, while her fascination with celestial rhythms, cycles, and natural timing sparked an unexpected perspective: what if the timeless patterns people look for in the sky could be channeled into grounded, daily self-awareness and intentional action?
I set out to build PhaseMe—not as an esoteric fortune-teller, but as an adaptive personal companion engineered to help people reflect deeply, understand their cognitive and emotional patterns, and make higher-conviction decisions in everyday life.
What started during one of my life’s most difficult chapters became a tool dedicated to clarity, resilience, and personal growth.
What It Does
PhaseMe bridges natural cycles, behavioral psychology, and adaptive artificial intelligence into a personal operating system for self-reflection and high-agency living.
Rather than passive horoscopes or superstitious predictions, PhaseMe provides structured tools to help you calibrate your mindset, clarify intentions, and track your evolution over time.
Strategic Decision Crucible & Prediction Engine: A four-milestone calibration system (evaluating risk tolerance, execution boundaries, communication style, and energy pacing) that forecasts outcomes across complex life crossroads. By logging real-world results, PhaseMe creates a closed feedback loop that sharpens your personal decision-making playbook over time.
Manifestation & Goal Execution Engine (WOOP): Grounded in the scientifically proven WOOP framework (Wish, Outcome, Obstacle, Plan), this module turns aspirations into structured execution roadmaps. It pairs intention-setting with internal friction analysis, sealed commitment check-ins, and longitudinal reflection alerts.
Interactive Knowledge Graph & Living Memory: Powered by Google Gemini and Supabase Vector Search, PhaseMe’s VIP intelligence layer extracts evolving themes, values, habits, and life events from your reflections. Instead of restarting fresh each session, your companion maintains a visual, interactive knowledge graph of your personal context and cognitive growth.
Voice-First Subconscious Journal: Capture fleeting dreams, morning reflections, or spoken voice memos hands-free. Real-time audio transcription and thematic extraction identify recurring cognitive patterns, stressors, and subconscious symbols before they get lost to waking life.
3D Celestial Sky Compass & AR Viewport: Built on Three.js and high-precision astronomical algorithms (
astronomy-engine), PhaseMe calculates the exact real-time positions of the Moon, Sun, and planets. Users can switch between a 3D virtual sky and an augmented reality camera view to track natural cycles and scrub time forward or backward.Daily Energy Rhythm & Biorhythm Wave: Harmonizes your baseline profile with current astronomical transits to generate daily cognitive, emotional, and physical pacing wave charts—helping you decide when to push deep-work sprints and when to prioritize restorative recovery.
AI Biometric Palm Analysis: Camera-driven hand tracking using on-device MediaPipe Vision identifies palm geometry and line structures, translating traditional archetypes into introspective self-awareness metrics.
Personal Constellation (Your Universe): Every committed decision, fulfilled milestone, journal entry, and intention joins a 3D interactive galaxy canvas—transforming your lived history into a tangible, navigable universe of personal progress.
How We Built It
PhaseMe is engineered with an emphasis on local-first resilience, battery-efficient graphics, and low-latency semantic AI pipelines:
- Frontend & Architecture: Built with Vue 3, TypeScript, Tailwind CSS 4, and Pinia, running natively on iOS and Android via Capacitor 8. The UI adheres strictly to edge-to-edge device safe areas and uses hardware-accelerated shared-element view transitions for fluid navigation.
- On-Device Computer Vision & 3D Math: MediaPipe Tasks Vision runs camera-guided palm analysis directly on the device with zero cloud latency. Three.js powers the 3D celestial compass, atmospheric lighting cycles, and personal galaxy canvas.
- Offline Astronomical Precision: Ephemeris calculations (Moon phases, horizon rise/set times, eclipses, planetary coordinates) run 100% locally via astronomical coordinate algorithms, remaining fully operational even without internet access.
- Semantic Memory & Vector Routing: The VIP layer couples Gemini models with Supabase Vector embeddings and PostgreSQL Row Level Security. A custom context routing engine isolates relevant biographical nodes, reducing token bloat and LLM latency by 75–80% while preserving long-term conversational memory.
- Monetization & Infrastructure: In-app VIP subscriptions are orchestrated seamlessly across iOS and Android via RevenueCat, while AdMob Server-Side Verification (SSV) securely verifies ad-supported feature unlocks through Deno Edge Functions.
Challenges We Ran Into
- Rehabilitation & Scope Discipline: Writing high-performance shaders, vector memory pipelines, and mobile view transitions while recovering from major physical surgery demanded extreme focus. Limited daily energy forced me to eliminate fluff and focus strictly on high-impact user value.
- Context Window Drift & Contradiction Resolution: As users regularly log reflections, dreams, and decisions, raw LLM context windows quickly suffer from memory bloat and stale data. We engineered a Supersede Detection & Memory Pruning pipeline that automatically detects when past goals or relationship contexts have resolved, keeping the AI's internal model of the user clean and up to date.
- Sustaining 60 FPS on Mobile with Sensors & AR: Orchestrating Three.js scenes, real-time device orientation sensors (gyroscope + compass heading), and camera video feeds on mobile WebViews can quickly cause thermal throttling. We offloaded sensor smoothing and ephemeris calculations, bypassed expensive blur shaders, and implemented dynamic render throttling during active touch gestures.
What We Learned & Key Takeaways
- Reframing Esoteric Concepts into Grounded Productivity: We discovered that people don’t necessarily want mystical fortune-telling; they want tools that encourage regular self-reflection, reduce decision fatigue, and connect their daily habits to broader life cycles.
- Long-Term Memory Is the True Differentiator: Generative AI becomes genuinely helpful when it remembers your long-term goals, past obstacles, and emotional triggers. Giving our assistant a persistent semantic knowledge graph transformed it from an ephemeral chatbot into a trusted thought partner.
- Artistic Engineering Creates Emotional Engagement: Marrying empirical behavioral frameworks (like WOOP) with high-fidelity 3D visuals and spatial audio makes personal development feel engaging rather than chore-like.
What's Next for PhaseMe
Autonomous Goal & Habit Orchestration: Advancing from passive tracking to proactive AI agents that break down high-level intentions into micro-milestones, dynamically schedule action steps around your peak focus hours, and follow up autonomously when obstacles arise.
Proactive Decision Coaching & Ambient Synthesis: An ambient background agent that monitors your recurring patterns, decision outcomes, and energy dips. Instead of waiting for prompts, it proactively surfaces blind spots, questions unhelpful heuristics, and nudges you with contextual reflection prompts before major crossroads.
Cyclical Sprint & Energy Cadence Engine: Blending biological pacing with autonomous productivity sprints—synchronizing your calendar, deep-work blocks, and retrospective reviews to natural cycles for sustainable, burnout-free performance.
Multi-Tool Autonomous Integrations: Equipping PhaseMe’s agentic layer with direct integrations into everyday productivity ecosystems (calendars, note repositories, task managers) to transform synthesized reflections directly into scheduled commitments and actionable workflows.
Built With
- admob
- ai-agents
- android
- capacitor
- computer-vision
- deno
- gemini-api
- hnsw
- mediapipe
- natural-language-processing
- onnx
- pinia
- postgresql
- revenuecat
- supabase
- tailwindcss
- three.js
- typescript
- vector-database
- vite
- vue.js
- vueuse
- webgl
- zod
Log in or sign up for Devpost to join the conversation.