Hi everyone,
Today we're launching NAEOS on Product Hunt, and I wanted to share the story behind it.
The Problem Nobody Talks About
Every engineering team has a spec document. Maybe it's a Google Doc, a Confluence page, or a YAML file in the repo. It captures the intent — the modules, the services, the architecture decisions.
But here's the dirty secret: the spec and the code drift apart within weeks.
Nobody updates the spec because "the code is the source of truth." But then:
New engineers have to reverse-engineer the system from code
AI assistants generate code that doesn't match your architecture
Governance and compliance become afterthoughts
Documentation rots silently
The Idea: Make the Spec the Source of Truth
NAEOS inverts this. Instead of treating code as the source of truth, we make the specification the single source of truth — and build tooling so good that keeping the spec accurate is the path of least resistance.
You describe your system once in YAML/JSON:
project: my-app
modules:
- name: api path: ./api
- name: auth path: ./auth services:
- name: rest-api kind: http port: 8080 architecture: pattern: microservices generation: languages: [go, typescript] Then NAEOS:
Parses the spec with variable interpolation, cross-references, and multi-file composition
Normalizes it to canonical form with defaults and type validation
Resolves dependencies and detects circular references
Builds NEIR — a complete internal engineering model of your system
Validates structural correctness and governance policies
Schedules parallel execution based on the dependency graph
Generates production code in Go, TypeScript, Python, Java, and Rust
Reviews generated artifacts against quality rules
Writes everything to disk with a machine-readable manifest
The AI Angle
Here's where it gets interesting for the AI era. The NEIR model captures your entire architecture — modules, services, dependencies, patterns, constraints. NAEOS compiles this into instruction sets for 6 AI tools:
GitHub Copilot → copilot-instructions.md
Claude Code → CLAUDE.md
Cursor → .cursorrules
Gemini CLI → .gemini/CONFIG.md
Codex → AGENTS.md
OpenCode → AGENTS.md
Your AI assistants don't just see individual files — they see the entire architecture, dependency graph, and design intent. Plus an MCP server exposes the model to AI agents programmatically.
What's in v3.1.0
This release focuses on developer experience and platform hardening:
Pipeline caching — naeos run --cache-dir caches stage results keyed by NEIR hash. Rebuilds with unchanged inputs skip regeneration.
Run-level profiling — --profile captures per-stage timing, --pprof starts a live pprof server for heap and CPU inspection.
Architecture patterns — monolithic, microservices, and serverless are now first-class in the NEIR model.
WASM plugin hardening — plugins run in isolated memory with metadata validation.
The Numbers
67 CLI commands — run, validate, compile, context, test, docgen, mcp, marketplace, diff, watch, migrate, and more
5 languages — Go, TypeScript, Python, Java, Rust
6 AI tools — Copilot, Claude Code, Cursor, Gemini CLI, Codex, OpenCode
86.9% test coverage with race detector, fuzz testing, and benchmark gates
424 test files across the codebase
10 GitHub Actions workflows for CI/CD
Try It
curl -fsSL https://naeos.dev/install.sh | sh
naeos create
cd my-app
naeos run --input-file spec.yaml
Links
Website: https://naeos.dev
Docs: https://docs.naeos.dev
GitHub: https://github.com/NAEOS-foundation/naeos
Product Hunt: https://www.producthunt.com/products/naeos?utm_source=other&utm_medium=social
Discord: https://discord.gg/WnUWmm7XMv
Whitepaper: https://naeos.dev/whitepaper
Help Us Out
If this resonates, an upvote on Product Hunt helps enormously: https://www.producthunt.com/products/naeos?utm_source=other&utm_medium=social
And if you've tried the quick start, a comment about your experience (good or bad) helps more than the upvote.
Thanks for readin
Top comments (0)