Open Source

L-Zero

Let AI Build Software For You

L-Zero is an instruction set designed for AI agents to generate executable programs. Just download the runtime, feed your AI the documentation, describe what you want, and watch it build.

How It Works

Three simple steps. No coding required.

1

Download L-Zero Runtime

Get the pre-built binaries for your platform. Includes the VM, assembler, and plugins for database, HTTP, file I/O, and more.

2

Give Your AI the README.md

README.md is the complete language specification. Paste it into your AI's context (Claude, GPT, etc.) so it understands how to write L-Zero programs.

3

Describe What You Want

"Build a blog with a SQLite database and HTTP server." The AI generates a .asm file, you compile and run it. Done.

Your Prompt
+
README.md
AI Agent
.asm
L-Zero VM
Product
# Example: Ask your AI to build a web server
# AI generates this .asm file:

SETS 1, "8080"
TEXEC 0x8000, 1, 255        # HTTP_INIT on port 8080

SETS 10, "GET /|Hello, World!"
TEXEC 0x8001, 10, 255       # Register route

SETS 20, "10"
TEXEC 0x8002, 20, 255       # Serve 10 requests

HALT

Why L-Zero?

Traditional programming languages are designed for humans to read and write. But when AI generates code, human readability doesn't matter. What matters is determinism, safety, and execution.

AI-Native Format

Assembly syntax minimizes tokens. One instruction per line. No syntax errors from missing brackets or commas. AI generates valid code the first time.

Type-Safe by Design

56 instructions, each with a strict schema. If it compiles, it runs. No runtime type errors. No undefined behavior.

Self-Describing

README.md is both documentation and specification. AI reads it once and knows everything: ISA, tools, plugins, examples.

Full-Stack Capabilities

Database (SQLite), HTTP server, file I/O, JSON, regex - all included. Build complete applications, not just scripts.

Semantic Governance

Vector operations and drift detection let AI agents self-monitor. When behavior drifts, execution pauses for human review.

AOT Compilation

Compile to native C code for maximum performance. Same program runs on VM during development, native binary in production.

The Principle

L-Zero flips the traditional model. Instead of humans writing code that machines execute, AI writes code that machines execute. The human's job is to describe intent.

Machine Determinism

Every instruction has exactly one meaning. No ambiguity. No interpretation. Perfect reproducibility.

Token Efficiency

Assembly format uses minimal tokens. More capability per context window. Lower API costs.

Sandboxed Execution

256 registers, managed heap, plugin isolation. AI-generated code can't escape its sandbox.

Extensible Tools

Core ISA (56 instructions) + unlimited plugins. Add capabilities without changing the language.

L-Zero is not a language for humans to learn. It's a language for AI to speak. You speak requirements. AI speaks L-Zero. The machine executes.

Get Started

1. Download

Grab the latest release for your platform from GitHub Releases. Includes all binaries and plugins.

Download

2. Get README.md

Copy README.md into your AI's context. This is the complete language reference - ISA, tools, examples, everything.

View README

# After downloading, install:
./install.sh              # User install to ~/.l0
./install.sh --system     # System install to /usr/local/l0

# Then run any .asm file your AI generates:
l0asm my_app.asm > my_app.l0
l0vm my_app.l0