Stop Starting AI Projects with Code. Start with Intent.

Stop Starting AI Projects with Code. Start with Intent.

Over the past year, I’ve written extensively about Intent-Driven Engineering—why intent matters, why enterprise context matters, and why AI-assisted software development needs a better operating model.

The most common question I receive isn’t about Claude Code, GitHub Copilot, MCP servers, or prompt engineering.

It’s much simpler.

“Where do I actually start?”

For a while, the answer was spread across dozens of articles covering repository organization, intent files, AI skills, agents, architecture, business rules, hooks, and governance.

I realized something important.

Every successful software movement eventually creates a canonical starting point.

React had Create React App.

Spring has Spring Initializr.

Docker has official starter examples.

Modern AI-native software development needs something similar.

Introducing the Intent Drive Starter Repository

To make that first step easier, I’ve published the Intent Drive Starter repository.

Repository:

https://github.com/kendallmark3/intent-drive-starter

The objective isn’t to promote a particular AI coding assistant.

Instead, the repository demonstrates how to organize a project so that any capable AI assistant can understand it from day one.

Whether you’re using Claude Code, Codex, Cursor, Gemini CLI, or future AI development tools, the principles remain the same.

The Repository Starts With Understanding

Traditional repositories begin with source code.

Intent-Driven Engineering begins with knowledge.

Instead of placing all of the intelligence inside prompts, the repository captures that knowledge as durable project assets.

The starter repository organizes:

• Project intent

• Current feature intent

• Architecture

• Business rules

• Domain glossary

• Integrations

• Security guidance

• AI skills

• Specialized agents

• Development hooks

• Templates

• Source code

• Tests

• CI/CD workflows

Every folder has a purpose.

Every document provides context.

Every AI assistant starts from the same understanding.

Why Intent Comes First

Most AI coding conversations begin with a prompt.

That’s useful for solving today’s problem.

It isn’t a great way to build software that will be maintained for years.

Intent-Driven Engineering shifts the primary artifact from the prompt to the repository itself.

Business goals become permanent documentation.

Architectural decisions become shared knowledge.

Business rules become reusable context.

Feature intent becomes the implementation contract.

The result is software that remains aligned with business objectives instead of drifting over time.

Greenfield Projects Start Here

If I were starting a brand-new project today, I wouldn’t create the src folder first.

I’d clone the starter repository.

I’d define the project intent.

I’d create the first feature intent.

I’d document the architecture.

Only then would I ask an AI assistant to begin implementation.

That sequence changes the quality of everything that follows.

Built to Evolve

This repository isn’t intended to be a static template.

As new patterns emerge, it will continue to evolve with additional examples, reusable skills, specialized agents, governance patterns, templates, and best practices learned from real-world enterprise delivery.

The goal is to provide a practical foundation that teams can adopt immediately while remaining flexible enough to grow alongside the rapidly changing AI ecosystem.

The Bigger Vision

Intent-Driven Engineering isn’t about replacing developers.

It isn’t about replacing Agile.

It isn’t about replacing software architecture.

It’s about giving both humans and AI assistants a shared understanding before implementation begins.

That’s why I believe repositories of the future won’t be organized around source code alone.

They’ll be organized around intent.

If you’re beginning a new AI-enabled software project, start with intent.

Clone the repository.

Replace the sample intent with your own business objective.

Build your project on understanding instead of prompts.

Repository:

https://github.com/kendallmark3/intent-drive-starter

Because the future of AI-native software development won’t be defined by who writes code the fastest.

It will be defined by who captures intent the best.

:::