AI First Dev Overview

PreviousNext

Overview of the AI First Dev approach

Overview

AI First Dev is a development methodology that leverages AI agents throughout the entire software development lifecycle, from initial design to final documentation.

This documentation is part of the repository (apps/zooly-docs/content/docs/). Developers should keep it up to date, as AI agents use it for context and decision-making.

Workflow

1. Design Phase

Start each major development task with a design plan:

  1. Create a design document in the docs/ directory
  2. Provide a rough sketch of your design concept
  3. Ask the AI agent to expand and formalize the design
  4. Review the design using multiple AI agents/models to ensure completeness and clarity
  1. Create an implementation plan by asking the AI to break down the work into phases
    • You can add this to the design document or create a separate plan document in docs/
  2. Review the plan manually or with AI assistance until it's ready for implementation

2. Implementation Phase

For each implementation phase:

  1. Implement using a quick model (e.g., Composer) for initial code generation
  2. Review and refine using stronger models (e.g., Opus or Codex), potentially multiple times with different models
  3. Update documentation by asking the AI to reflect changes in the design/plan documents

3. Testing Phase

  1. Manually test the implementation and verify it works as expected
  2. Fix issues using AI assistance
  3. Update documentation to reflect any changes discovered during testing

4. Documentation Phase

Create final documentation in apps/zooly-docs/content/docs/ that includes everything necessary for future iterations and maintenance, whether performed by humans or AI agents.

Documentation guidelines:

  • Final docs should reference code files, not include code snippets
  • Temporary design/plan docs can include code snippets for quick reference during development
  • Avoid code in final docs because it won't be maintained and will confuse AI agents

Using Existing Documentation as Context

During development, include relevant documentation in the AI context to help it understand the current task and make better decisions.

You can add documentation to the AI context in two ways:

  1. Drag or mention the document file directly in the chat
  2. Use the copy icon located at the top right of the document file to copy its content to the chat