📐 base10 — Decision Intelligence Engine

base10 is a premium, structurally rigorous decision intelligence framework designed to map complex, high-stakes decision problems (e.g., startup founder transitions, career shifts, capital investments) into probability-weighted, 5-level consequence trees.

Built on custom Expected Value (EV) mathematical recursion engines and Risk-Adjusted Expected Value (RAE) variance models, base10 exposes the full structural anatomy of your choices, highlighting irreversible gateway traps and stochastic luck vectors before you exhaust your capital.

---

📖 Table of Contents

• Design Philosophy & Aesthetics
• System Architecture Flow
• Mathematical Formulation
  • 1. Expected Value (EV) Recursion
  • 2. Risk-Adjusted Expected Value (RAE)
  • 3. Consequence Alpha-Pruning
• Local Setup & Deployment
• Security & Sandboxing Policy
• License

---

🎨 Design Philosophy & Aesthetics

1. Ink-Drawn Sketch Aesthetic: Light mode warm parchment theme (#FAFAF5), ink-drawn grid lines, raw drafting crosshair styling, and structured cards using JetBrains Mono for metrics and EB Garamond for textual narratives.
2. Deterministic Control vs. Stochastic Luck: Decomposes every decision problem into deterministic actions (choices you control) and stochastic events (probabilistic outcomes dependent on market chance).
3. Information Saturation: Dynamic multi-round elicitation structure maps and convergence score meters verify that all structural variables are locked before processing.
4. Secure Key Sandboxing: Strict key safety. Anthropic API keys are loaded solely into browser sessionStorage and never persist on databases, local disk, or servers. Safe offline "Demo Mode" is available with pre-baked high-fidelity simulations.

---

⚙️ System Architecture Flow

The following diagram illustrates how the client-side engine maps user inputs, parses schemas, processes structural trees, and computes mathematical matrices:

graph TD
    Key["API Key Setup - Local sessionStorage"] --> Select[Select Decision Problem]
    Select --> Config[Configure Utility Weights & Probabilities]
    Config --> Tree["Generate Consequence Tree: 5 levels deep"]
    Tree --> LLM["Optional: Elicit nodes using Anthropic API"]
    Tree --> Model["EV Recursion & Risk-Aversion Variance RAE Model"]
    Model --> AlphaPruning[Alpha-Pruning of Sub-Optimal Branches]
    AlphaPruning --> UI[Interactive Tree Visualizer & Risk Meters]

    classDef default fill:#fafaf5,stroke:#2b2b2b,stroke-width:1.5px,color:#2b2b2b;
    classDef process fill:#f0eedd,stroke:#2b2b2b,stroke-width:2px,color:#2b2b2b;
    class Tree,Model,AlphaPruning process;

Loading

---

📐 Mathematical Formulation

1. Expected Value (EV) Recursion

A bottom-up recursive traversal computes the expected value of every decision node:

$$\text{EV}(n) = \text{utility}(n) + \sum_{c \in \text{children}} \text{prob}(c) \times \text{EV}(c)$$

2. Risk-Adjusted Expected Value (RAE)

To punish high-variance extreme downside scenarios, base10 incorporates standard deviation weighting ($\sigma$):

$$\text{RAE}(n) = \text{EV}(n) - \lambda \times \sigma(n)$$

Where:

• $\lambda$ represents the user's risk-aversion coefficient.
• $\sigma(n)$ represents the standard deviation of outcomes propagating from node $n$.

3. Consequence Alpha-Pruning

Branches whose local expected value falls below the optimal action's threshold are pruned using:

$$\text{abs}(\text{EV}(n)) < \max(\text{abs}(\text{EV})) \times \alpha$$

---

🚀 Local Setup & Deployment

Run Development Server

npm install
npm run dev

Build Production Bundle

npm run build

Deploy to GitHub Pages

git push -u origin master
npm run deploy

---

🔒 Security & Sandboxing Policy

No tracking, no cookies, no backend database. Your API key never leaves your local browser memory space. Closing the tab immediately wipes all operational keys from sessionStorage.

---

📄 License

This repository is licensed under the CC0 1.0 Universal (CC0 1.0) Public Domain Dedication.