I’m building NAEOS: an engineering runtime for AI coding agents

I started building NAEOS after seeing a gap in modern AI coding tools:

AI is getting better at writing code, but not necessarily at preserving engineering intent.

Agents can modify repositories, write tests, and execute complex tasks, but they still need to understand architecture, dependencies, policies, and system constraints.

This led me to a simple question:

What if software was specified first, and code became an output of that specification?

That became NAEOS.

The core model is:

Specification → NEIR → Validation → Execution Plan → Artifacts

NEIR (NAEOS Engineering Intermediate Representation) provides a structured model between engineering intent and implementation.

I don't want NAEOS to be another AI coding agent. I see it as the layer above agents: defining engineering intent and providing the context and constraints that different agents can execute.

NAEOS also explores executable governance, where architecture and engineering policies can be validated by machines rather than existing only as documentation.

The project is open source because I want the architecture challenged.

I'm particularly interested in feedback from engineers working on compilers, developer platforms, infrastructure, programming languages, and AI agents.

I'm also looking for technical contributors, potential co-founders, and strategic partners who want to explore this direction with me.

If the idea is wrong, I'd rather discover that now.

GitHub: https://github.com/NAEOS-foundation/naeos

Specify Once. Build Anywhere.