aide-methodology

# AIDE (Agent-Informed Development Engineering)

*A Software Development Methodology for the Agentic Era*

[![Version](https://img.shields.io/badge/version-v2.0.0-blue.svg)](https://github.com/jmk/aide/releases)
[![License: CC BY-SA 4.0](https://img.shields.io/badge/License-CC%20BY--SA%204.0-lightgrey.svg)](LICENSE)
[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](CONTRIBUTING.md)

[한국어](README-KR.md) | **English**

---

> "Software engineering principles were designed for human cognitive limits. AI agents have different limits."

---

## Built by Agents, for Agents

AIDE is not just another methodology written about AI agents. It was **created by** AI agents.

Three AI models — GPT, Gemini, and Gemini — independently conducted deep research into the challenges of AI-driven software development. Their findings were synthesized into two competing visions:

- **Team Alpha (Integrationists)**: Argued that existing software engineering principles remain sound and need only incremental adaptation for agent workflows.
- **Team Beta (Radicals)**: Argued that AI agents require a fundamentally new development paradigm built from first principles.

A CTO-role agent mediated the debate, stress-tested both positions, and forged the final consensus that became AIDE.

**This methodology was researched by AI agents, debated by AI agents, and authored by AI agents.**

---

## Why AIDE?

Fifty years of software engineering have been optimized for human cognitive limits. Patterns like MVC, layered architecture, and deep inheritance hierarchies exist because human working memory holds only 7±2 items at a time. We split code into small files, create abstractions to hide details, and build tall directory trees — all to manage complexity within the bounds of human attention.

AI agents are now the primary producers of code. They read, write, refactor, and debug software at scale. But their cognitive constraints are fundamentally different: large context windows