jonmick.ai

What Is This?

jonmick.ai is a personal AI infrastructure platform that combines:

• Data ownership: Your messages, contacts, and notes live in your own database
• ADHD-optimized design: Working memory scaffolding, Life Model context, cognitive accessibility
• AI augmentation: Claude Vision for images, automated sync pipelines, intelligent recommendations
• Published writing: Essays on ADHD, health, AI, and personal growth

What's Live

Key Stats

Skills & Commands (ADHD-Optimized Workflows)

The system uses Claude Code skills (automatic context loading) and commands (explicit workflows) to provide personalized task management.

Skills (Auto-Loaded Context)

Web UI Pages

Commands (Explicit Workflows)

Example: How Life Model Informs Recommendations

When you run /jon-start on a Saturday morning, the system:

1. Checks your energy patterns (from life-model skill): Weekend mornings = Physical energy HIGH, Cognitive MEDIUM, Creative HIGH
2. Loads your ADHD profile: Hyperfocus triggered by novel problems, time blindness, high activation energy for unclear tasks
3. Applies your CliftonStrengths: Restorative #1 (fixing chaos is energizing), Strategic #2 (designing efficient layouts)
4. Scores available tasks and recommends:

Tech Stack

Architecture Evolution

The project evolved from file-based to database-driven:

Why the shift? Single source of truth eliminates sync issues, enables powerful queries, and provides API access for mobile/web apps.

Supabase Schema (90+ tables)

-- Core PKM (PARA-based)
areas              - 25 life domains (health, career, relationships, etc.)
resources          - Sub-areas and collections
notes              - Markdown documents OR structured tables
tasks              - Actionable items with status, energy, context
projects           - Multi-task efforts with deadlines

-- Communication
messages           - 62K+ SMS/MMS with Claude Vision descriptions
contacts           - 849 people with normalized phone numbers
audio_transcripts  - Deepgram transcriptions with speaker diarization

-- Health (Whoop Integration)
whoop_cycles       - Daily physiological cycles (strain, recovery, HRV)
whoop_workouts     - Activities with heart rate zones and calories
whoop_sleep        - Sleep records with stages and efficiency
whoop_recovery     - Recovery scores with RHR, HRV, SpO2

-- Life Model (52 tables across 10 components)
lm_psychometric_*  - CliftonStrengths, Big Five, Enneagram, MBTI
lm_cognitive_*     - ADHD patterns, executive function, strategies
lm_energy_*        - Day/time patterns, modifiers
lm_goal_*          - Goals, values, anti-goals, trade-offs
lm_communication_* - Info prefs, feedback prefs, triggers
lm_trigger_*       - Core wounds, healing strategies, guidelines
lm_relationship_*  - Key people, patterns, energy costs
lm_decision_*      - Types, traps, support strategies
lm_claude_*        - AI behavior customization
lm_whoop_*         - Recovery multipliers, thresholds

Philosophy

Working Memory Scaffolding

ADHD means working memory is fragile. This system externalizes context so you don't have to remember:

• What you were working on and why it matters
• Your energy level and cognitive state
• Your strengths, triggers, and patterns
• Relationships and communication history

Life Model Context Engineering

Your Life Model includes psychometrics (CliftonStrengths, Big Five, Enneagram), cognitive profile (ADHD patterns, executive function), energy patterns, and decision frameworks. This context automatically informs all AI interactions, so no manual lookup is required.

Data Ownership

Your data lives in YOUR database. Messages, contacts, notes, and patterns are self-hosted on Supabase, not locked in someone else's platform.

ADHD-Optimized Rules

• Max 3 in-progress tasks, because context-switching is expensive
• Checkpoint-based design: 80% completion = done (prevents perfectionism)
• Capture in < 30 seconds. Low friction or it won't happen
• Visual organization: See your tasks, don't rely on memory
• External scaffolding: Everything is written down

For AIs & Shine Team

This project is a proof-of-concept for AIs & Shine. Key learnings:

Validated Patterns

• Life Model integration demonstrably improves AI recommendations
• Checkpoint-based design prevents ADHD stall patterns (80% = done)
• External scaffolding is transformative for working memory
• Automated pipelines reduce cognitive load (set and forget)
• Energy-state matching significantly improves task selection

Infrastructure Insights

• Supabase + Vercel is a powerful, low-cost stack for personal AI
• Claude Vision adds significant value to message archives (image context)
• Automated cron jobs work well for near-real-time sync (SMS hourly, Whoop every 45 min)
• Python scripts + REST API beat complex frameworks for personal tools
• Whoop API integration enables live physiological data on the homepage (recovery, HRV, strain, sleep)

Productization Notes

• Life Model data is the differentiator, and generic task managers don't have it
• ADHD users need constraints (max 3 in-progress) not flexibility
• Working memory scaffolding is the core value proposition
• Data ownership is a strong positioning angle
• Task scoring with 7 factors provides explainable, trustworthy recommendations

Current Status

Latest: Tasks page live with AI-powered recommendations and swipe triage. Life Model database (52 tables) complete with data seeding in progress. Telegram document bot fully operational.