A CLI tool that analyzes stocks through Charlie Munger's 28 mental models using 10 parallel AI agents.
- 10 Parallel AI Agents - Each specialized in different mental models
- 28 Mental Models - Comprehensive Munger-style analysis covering:
- Core Investment Principles (Circle of Competence, Margin of Safety, Mr. Market, Intrinsic Value)
- Moats & Ownership (Economic Moats, Owner Earnings, Management Quality)
- Psychology & Behavioral (Incentive Bias, Social Proof, Confirmation Bias, Loss Aversion, etc.)
- Lollapalooza Effects & Second-Order Thinking
- Math & Probability (Inversion, Base Rates, Expected Value, Compound Interest)
- Economics & Business (Opportunity Cost, Competitive Destruction, Scarcity)
- Systems Thinking (Feedback Loops, Critical Mass, Scale Economics)
- Decision Filters (Too Hard Pile, Checklist)
- Code Execution - Agents can run Python for DCF models, Monte Carlo simulations, statistical analysis
- Web Search - Optional Exa AI integration for real-time market research
- Finviz Data - Automatic scraping of fundamental and technical data
- Charlie's Verdict - Final synthesis in Munger's voice with actionable rating
# Clone the repo
git clone <repo-url>
cd lattice
# Install dependencies
bun install
# Build
bun run buildRun the interactive setup to configure your API keys:
lattice setupThis will guide you through entering your API keys and save them to ~/.lattice/config.json.
Alternatively, set environment variables:
# Required
export ANTHROPIC_API_KEY="sk-ant-..."
# Optional (enables web search)
export EXA_API_KEY="exa-..."Or manually create the config file:
{
"anthropicApiKey": "sk-ant-...",
"exaApiKey": "exa-..."
}# Analyze a stock
lattice analyze AAPL
# Watch mode - live dashboard showing all agents working
lattice watch AAPL
# Verbose mode - show detailed analysis per mental model
lattice analyze AAPL --verbose
lattice watch AAPL --verboseAnalyses are automatically saved and can be retrieved later:
# List all saved analyses
lattice history
# View a saved analysis by ticker (shows most recent)
lattice view AAPL
# View a saved analysis by ID
lattice view abc123
# View with verbose mode
lattice view AAPL --verbose
# Clear all saved analyses
lattice history --clearAnalyses are saved to ~/.lattice/history/.
# Mock analysis results for UI testing
lattice mock
# Demo the live agent dashboard with simulated activity
lattice mock-watchEnable verbose logging to troubleshoot issues:
# Logs written to /tmp/lattice-debug.log
LATTICE_VERBOSE=1 bun analyze AAPL
# Custom log file
LATTICE_DEBUG_FILE=/path/to/debug.log LATTICE_VERBOSE=1 bun analyze AAPLThe analysis provides:
- Signal Summary - Bullish/bearish/neutral counts across all mental models
- Charlie's Verdict - STRONG BUY | BUY | HOLD | SELL | STRONG SELL | TOO HARD
- Key Numbers - Intrinsic value estimate, margin of safety, projected CAGR (when computed)
- Category Insights - One-line summaries for each mental model category
- Red Flags - Specific concerns identified during analysis
- What Would Make This Better - Conditions that would improve the investment case
╔══════════════════════════════════════════════════════════════════════════╗
║ LATTICE AAPL Apple Inc ║
║ Sector: Technology / Consumer Electronics Market Cap: 3678.01B ║
║ $250.24 ▼ -2.07% ║
╚══════════════════════════════════════════════════════════════════════════╝
Signal Distribution
● 3 bullish ● 15 bearish ● 14 neutral ● 0 n/a
[?] CHARLIE'S VERDICT: TOO HARD
Well, here we have one of the finest businesses ever created - Apple makes
more money than God and has built switching costs that would make John D.
Rockefeller envious. The problem is simple: you're being asked to pay $250
for something worth maybe $149, and that's if China doesn't blow up in
their face and the AI story works out...
Key Numbers
├─ Current Price: $250.24
├─ Intrinsic Value Estimate: $149 (PEG 1.5x, historical P/E 20-24x)
├─ Margin of Safety: -68%
└─ 5-Year CAGR Projection: 3.5%
Category Insights
▸ Core: Exceptional business, but paying $250 for $149 of value
▸ Psychology: Confirmation bias ignoring China collapse
▸ Math: Expected 5-year return barely exceeds risk-free rate
▸ Economics: Winner-take-most in US offset by China collapse
▸ Systems: Ecosystem flywheel decelerating measurably
▸ Filters: Passes quality but fails valuation checklist
Red Flags
• P/E of 33.5x for 10% growth = PEG of 2.6 (68% overvalued)
• China revenue down 17% annually, lost #1 to #3 market position
• Insider ownership 0.1% while authorizing $110B buybacks
10 agents • 28 mental models • 338s
A full analysis takes approximately 4-6 minutes depending on API response times. The 10 agents run in parallel, each potentially using code execution for quantitative analysis.
CLI Input (ticker)
→ Finviz Scraper (cheerio + fetch)
→ 10 Parallel Agent API Calls (Promise.all)
→ Agentic Loop (handle tool calls)
→ Aggregate Results
→ Final Synthesis (Charlie Munger voice)
→ Ink Render Output
| Agent | Mental Models |
|---|---|
| 1 | Circle of Competence, Margin of Safety, Mr. Market, Intrinsic Value |
| 2 | Economic Moats, Owner Earnings, Management Quality |
| 3 | Incentive-Caused Bias, Social Proof, Availability Bias |
| 4 | Confirmation Bias, Commitment & Consistency, Loss Aversion |
| 5 | Lollapalooza Effects, Second-Order Thinking |
| 6 | Inversion, Base Rates |
| 7 | Expected Value, Compound Interest |
| 8 | Opportunity Cost, Sunk Cost Fallacy, Competitive Destruction, Scarcity |
| 9 | Feedback Loops, Critical Mass, Scale Economics |
| 10 | Too Hard Pile, Checklist |
- Runtime: Bun
- Language: TypeScript
- AI: Anthropic Claude (Haiku 4.5 for agents, Sonnet 4.5 for synthesis)
- CLI UI: Ink (React for CLI)
- Web Scraping: Cheerio
- Search: Exa AI (optional)
lattice/
├── src/
│ ├── index.ts # Entry point
│ ├── cli/
│ │ ├── index.tsx # CLI commands
│ │ ├── commands/analyze.tsx # Analyze command
│ │ └── components/ # Ink UI components
│ ├── agents/
│ │ ├── orchestrator.ts # Parallel execution
│ │ ├── agentConfigs/ # 10 agent configurations
│ │ └── tools/exaSearch.ts # Exa search handler
│ ├── data/
│ │ ├── finvizScraper.ts # Finviz scraper
│ │ └── types.ts # TypeScript interfaces
│ ├── models/mentalModels.ts # 28 mental model definitions
│ ├── synthesis/finalAnalysis.ts # Charlie Munger synthesis
│ └── utils/
│ ├── config.ts # Config loader
│ └── formatting.ts # Utilities
├── bin/lattice.js
├── package.json
└── tsconfig.json
MIT