A free, web-based AI life-support system built for high-potential adults with mental illness or neurodivergence who feel like they are "doing life wrong."
Users make decisions that feel right in the moment but have lasting negative consequences they couldn't anticipate. LifeOS exists to help users identify hidden patterns in behavior and decision-making, simulate long-term outcomes, and interrupt self-sabotage.
- Natural conversation as the primary interaction
- No forced forms or rigid structure
- Adaptive tone based on user state
- Identifies recurring themes across conversations
- Tracks decision archetypes (escape decisions, people-pleasing, impulse-driven choices)
- Detects temporal and emotional patterns
- 3-horizon framework: 1 week / 1 month / 6 months
- Second-order effect exploration
- Values alignment scoring
- Pattern summaries with contextual examples
- "Déjà vu alerts" when current decisions mirror past patterns
- Real-time pattern notifications
- Download
lifeos.html - Open it in any modern browser (Chrome, Firefox, Safari, Edge)
- Start conversing
That's it. No installation, no servers, no dependencies.
# 1. Install Vercel CLI
npm i -g vercel
# 2. Create a new directory
mkdir lifeos-app
cd lifeos-app
# 3. Copy lifeos.html to directory
cp /path/to/lifeos.html index.html
# 4. Deploy
vercelFollow the prompts. Your app will be live in ~30 seconds.
┌─────────────────────────────────────────────────────────┐
│ USER INTERFACE LAYER │
│ (React 18 - Single File App) │
└─────────────────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────────────────┐
│ CONVERSATION MANAGER │
│ - Intent classification │
│ - Context window management │
│ - Tone calibration │
└─────────────────────────────────────────────────────────┘
↓
┌──────────────────┴──────────────────┐
↓ ↓
┌──────────────────┐ ┌──────────────────────┐
│ PATTERN ENGINE │ │ DECISION SIMULATOR │
│ │ │ │
│ • Keyword │ │ • Consequence trees │
│ analysis │◄────────────►│ • Time horizons │
│ • Theme tagging │ │ • Risk assessment │
│ • Cycle detection│ │ │
└──────────────────┘ └──────────────────────┘
↓ ↓
┌─────────────────────────────────────────────────────────┐
│ LOCAL STORAGE LAYER │
│ (Browser localStorage API) │
│ │
│ Keys: │
│ • lifeos_messages (conversation history) │
│ • lifeos_patterns (detected patterns) │
│ • lifeos_settings (user preferences) │
└─────────────────────────────────────────────────────────┘
| Component | Technology | Reasoning |
|---|---|---|
| Frontend | React 18 (CDN) | No build step, instant load, works offline |
| Styling | Pure CSS | Zero dependencies, full control, excellent performance |
| Storage | localStorage | Privacy-first, works offline, no backend needed |
| AI | Simulated (Claude API ready) | Pattern matching with keyword detection, easy to swap in real API |
| Fonts | Google Fonts | Crimson Pro (elegant serif) + IBM Plex Mono (technical feel) |
- Dark theme: Reduces cognitive load, creates calm environment
- Animated gradients: Subtle depth without distraction
- Crimson Pro serif: Warm, literary, thoughtful
- IBM Plex Mono: Technical precision for data/patterns
--bg-primary: #0a0e12 /* Deep space black */
--accent-primary: #7aa2f7 /* Calming blue */
--accent-secondary: #bb9af7 /* Thoughtful purple */
--accent-warm: #ff9e64 /* Gentle orange for emphasis */- Conversation-first: No barriers to starting
- Progressive disclosure: Complex features appear as needed
- Calm technology: Smooth animations, generous spacing
- Transparency: Show why patterns were detected
// Three primary pattern types
1. Escape When Overwhelmed
Triggers: quit, leave, escape, overwhelmed, drowning
2. Repeating Mistakes
Triggers: again, same, keep, always, every time
3. Impulse-Driven Decisions
Triggers: right now, immediately, urgent, need to- Analyzes last 5 user messages
- Requires 2+ keyword matches to trigger
- Stores examples for user review
- Increments detection count on recurrence
- Semantic similarity using embeddings
- Temporal pattern analysis (time-of-day, day-of-week)
- Emotional tone tracking via sentiment analysis
- Cross-pattern correlation
// Immediate override for crisis keywords
const crisisKeywords = [
'kill myself',
'suicide',
'end it all',
'want to die',
'hurt myself'
];
// Response: Crisis resources + professional referral- ❌ No medical diagnoses
- ❌ No therapeutic treatment
- ❌ No spiritual authority claims
- ❌ No financial advice
- ✅ Pattern observation only
- ✅ Professional referrals when needed
- All data stored locally (localStorage)
- No analytics, no tracking
- User controls all data (clear/export)
- No third-party API calls in demo
{
role: 'user' | 'assistant',
content: string,
timestamp: number
}{
id: string, // 'escape-when-overwhelmed'
title: string, // 'Escape When Overwhelmed'
description: string, // Human-readable explanation
detectedCount: number, // How many times triggered
examples: string[] // Conversation snippets
}{
userName: string, // Optional personalization
worldview: string, // secular | faith-based | philosophical
apiKey: string // For future Claude API integration
}Currently, LifeOS uses simulated responses for demo purposes. To integrate the real Claude API:
# Sign up at https://console.anthropic.com
# Create API keyconst sendMessage = async () => {
// ... existing validation ...
try {
const response = await fetch('https://api.anthropic.com/v1/messages', {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'x-api-key': settings.apiKey,
'anthropic-version': '2023-06-01'
},
body: JSON.stringify({
model: 'claude-sonnet-4-20250514',
max_tokens: 1000,
system: SystemPrompts.getSystemPrompt(settings.userName, settings.worldview, patterns),
messages: newMessages.map(m => ({
role: m.role,
content: m.content
}))
})
});
const data = await response.json();
const assistantMessage = {
role: 'assistant',
content: data.content[0].text,
timestamp: Date.now()
};
// ... rest of existing logic ...
} catch (error) {
console.error('API error:', error);
// Fallback to simulated response
}
};Already built into the settings modal - just needs backend integration.
For portfolio demonstration:
- Pattern Detection Accuracy: Does it catch the 3 core patterns?
- User Experience: Smooth, calming, non-judgmental?
- Ethical Boundaries: Clear crisis handling and referrals?
- Technical Excellence: Clean code, good performance, responsive?
- Simulated AI: Keyword-based pattern detection (not semantic)
- No backend: All data in browser localStorage
- No authentication: Single-user device only
- Limited patterns: 3 types (escape, repetition, impulse)
- No analytics: Can't track effectiveness over time
- Real Claude API integration
- Semantic pattern detection with embeddings
- Weekly/monthly reflection summaries
- Data export (JSON, PDF report)
- Multi-device sync (optional backend)
- Advanced decision simulation with probabilistic outcomes
User opens LifeOS
↓
Reads welcome message
↓
Clicks "Start Conversation"
↓
LifeOS: "Hey. I'm LifeOS. Think of me as a pattern-recognition
system for your life..."
↓
User: "I keep fucking up at work. Thinking about quitting."
↓
LifeOS: "Have you made a big change like this before when you
felt overwhelmed?"
↓
User: "Yeah, I quit my last job too."
↓
LifeOS: [PATTERN DETECTED: Escape When Overwhelmed]
"I'm noticing this decision follows the same pattern as the
last time. This might be a pattern, not a solution."
↓
User: "...fuck. Yeah. That's exactly it."
↓
LifeOS: "Let's focus on reducing overwhelm within this job
for two weeks, then reassess?"
/* Display */
font-family: 'Crimson Pro', serif;
font-weight: 300 | 400 | 600;
/* Monospace (data/code) */
font-family: 'IBM Plex Mono', monospace;
font-weight: 400 | 500;--space-xs: 0.25rem; /* 4px */
--space-sm: 0.5rem; /* 8px */
--space-md: 1rem; /* 16px */
--space-lg: 1.5rem; /* 24px */
--space-xl: 2rem; /* 32px */
--space-2xl: 3rem; /* 48px */--timing-fast: 0.2s;
--timing-base: 0.3s;
--timing-slow: 0.6s;
--easing: cubic-bezier(0.4, 0, 0.2, 1);This is a portfolio project, but feedback is welcome:
- Found a bug? Open an issue with steps to reproduce
- Have an idea? Describe the use case and why it matters
- Want to improve UX? Show before/after designs
# No setup needed - it's a single HTML file!
# Just open in browser and start editing
# For live reload during development:
npx live-serverMIT License - feel free to use, modify, and learn from this code.
Built by Pelz as part of a comprehensive fintech portfolio, demonstrating:
- Applied cognitive psychology and behavioral economics
- Ethical AI system design with clear safety boundaries
- Compassionate user experience for high-stakes scenarios
- Full-stack technical execution (React, localStorage, API-ready architecture)
- Product thinking: solving real problems for specific user segments
This project demonstrates capabilities highly valued in financial services:
1. Risk Assessment & Decision Modeling
- LifeOS's 3-horizon decision simulation mirrors risk modeling in trading/lending
- Shows understanding of second-order effects and probabilistic thinking
- Demonstrates ability to help users evaluate decisions under uncertainty
2. Pattern Detection & Behavioral Analysis
- Core competency for fraud detection systems
- Behavioral pattern recognition applies to customer financial behavior analysis
- Shows understanding of how to surface insights from temporal data
3. Ethical AI & Responsible Design
- Critical for banks deploying AI in consumer products
- Clear boundaries, crisis detection, and privacy-first architecture
- Demonstrates thinking about AI safety before deploying
4. User Trust & Security
- Privacy-first design (local storage, no tracking) aligns with banking security
- Crisis intervention shows understanding of high-stakes user scenarios
- Ethical guardrails demonstrate mature product thinking
5. Full-Stack Capability
- Complements your fintech portfolio (Budget Buddy, Fraud Detection, Stress Simulator)
- Shows you can go from concept → design → implementation → deployment
- Clean architecture that scales from prototype to production
| Skill | How LifeOS Shows It |
|---|---|
| Behavioral Economics | Decision simulation, pattern detection in user choices |
| Risk Modeling | 3-horizon consequence analysis, probabilistic outcomes |
| System Design | Modular architecture, separation of concerns |
| Ethics & Compliance | Clear boundaries, crisis protocols, privacy design |
| Frontend Excellence | Production-quality React, accessibility, responsive design |
| Product Thinking | Solved a real problem for a specific user segment |
In interviews, position LifeOS as:
"A pattern recognition system that helps users make better decisions by simulating long-term consequences and detecting behavioral patterns. It demonstrates my ability to build ethical AI systems with clear risk boundaries—skills directly applicable to consumer banking products that require both technical excellence and responsible design."
Key talking points:
- Built with ethical AI principles and responsible design practices
- Pattern detection similar to fraud/anomaly detection in banking
- Decision modeling parallels credit risk assessment
- Privacy-first architecture aligns with financial data security
- Complete solo build (design → code → deploy) in portfolio timeline
This is your "I can build thoughtful, ethical systems" project that complements your "I can build technical fintech systems" projects.
"LifeOS doesn't fix you. It helps you see yourself clearly enough to fix what's fixable — and accept what isn't."
This is not a replacement for therapy, spiritual guidance, or human connection. It's a tool for pattern recognition — nothing more, nothing less.
Use it wisely. Use it kindly. Use it as a supplement, not a substitute.
Version: 1.0.0-MVP
Last Updated: January 2026
Status: Portfolio Demo (Production-Ready Architecture)