Skip to content

Specifications

Overview

Specifications are the foundation of reliable AI systems. This section teaches you how to write clear, actionable specifications that reduce cognitive friction and help AI models understand exactly what you need.

Rather than vague requirements that force models to guess, you'll learn a complete framework: MUST constraints, SHOULD guidelines, CONTEXT for planning, INTENT for understanding why, and VERIFICATION protocols to confirm success.

What You'll Learn

Core Framework

The Specifications framework has four interconnected layers and Verification Protocols:

MUST Constraints - Hard boundaries and non-negotiable requirements
SHOULD Guidelines - Flexible preferences that adapt to context
CONTEXT - Planning information about environment, users, and constraints
INTENT - The "why" behind your requirements and trade-offs
VERIFICATION - How to confirm the model met your needs

Complete Coverage

Why Specifications Matter

Vague requirements create cognitive friction. When models must guess what you mean, they:

  • Waste context on wrong paths
  • Make incorrect assumptions
  • Produce unreliable results
  • Require multiple iterations

Good specifications eliminate guessing. They give models the clarity needed to deliver what you actually want, the first time.

Getting Started

New to Specifications?

Start here:
Foundations → MUST Constraints → SHOULD Guidelines

Work through the core framework sequentially to build a solid understanding.

Need Something Specific?

Key Principle

All five work together. MUST defines boundaries. SHOULD provides flexibility. CONTEXT enables planning. INTENT reveals why. VERIFICATION PROTOCOLS confirm success.

Missing any layer increases cognitive friction and reduces reliability.

Ready to begin? Start with Foundations →