Self-Building AI Operating System

Eight core primitives. Infinite autonomous intelligence.

import { PlannerAgent, CodeGenerator } from '@lumenos/lumen-g-kernel';

// LumenOS: AI that builds and improves itself
const planner = new PlannerAgent();
const plan = await planner.createPlan({
  goal: 'Build a REST API for user authentication',
  constraints: { time: '1 week' }
});

// All outputs cryptographically signed & verifiable
for (const step of plan.plan) {
  console.log(`Step ${step.step}: ${step.purpose}`);
  // Each step has a verified signature proving provenance
}

The 8 Core Primitives

1

Box Kernel

Cryptographic signing, schema enforcement, and verification for all outputs

2

Schema Factory

Create JSON schemas from natural language descriptions

3

Program Factory

Generate and improve code in TypeScript, JavaScript, and Python

4

Environment Interface

Generate safe shell commands with security analysis

5

Reasoner

Structured meta-reasoning with tradeoff analysis and confidence scoring

6

Memory

Conversational context with automatic summarization and persistence

7

Planner

Executive function that sequences primitives to achieve goals

8

Box Registry

Self-model maintaining system knowledge and relationships

Why Lumen-G?

🔐

Cryptographically Verified

Every output is signed with Bitcoin Secp256k1 ECDSA, proving authenticity and provenance

🔄

Self-Building

Eight primitives recursively compose to build new agents, tools, and capabilities

🎯

Goal-Directed

Planner creates multi-step plans, evaluates quality, and adapts to constraints

🧠

Self-Aware

Registry maintains complete self-model with metrics, relationships, and evolution tracking

Multi-Provider

Works with OpenAI GPT-4 and local Llama models via unified interface

📡

REST API

15+ endpoints expose all primitives for remote access and integration

8
Core Primitives
6K+
Lines of Code
15+
API Endpoints
Possibilities

From Eight Primitives, Infinite Possibility

Drop this kernel in an empty environment, and it can bootstrap a city of intelligent agents.