Only GitHub Copilot

Exclusively Anthropic

Only open-source LLMs

ClaudeClaude, Gemini CLI, Copilot, Cursor, and others

Specifications become executable artifacts that drive the entire implementation process.

Human developers focus solely on syntax while the AI handles all product requirements.

AI agents are given full autonomy to write code without any prior documentation or oversight.

Coding occurs before documentation to ensure the spec is accurate to the final product.

Code serves specifications, acting as the concrete realization of a pre-defined contract.

Specifications are generated from the final source code to ensure they remain up-to-date.

Specifications are optional guidelines that AI agents can ignore to optimize performance.

Code and specifications are developed independently to allow for creative exploration.

It allows the project to be compiled faster by optimizing the compiler flags.

It ensures the AI follows the existing project's coding standards and architectural rules.

It automatically fixes all existing bugs in the legacy code through machine learning.

It provides a legal license that forbids AI from training on the project's source code.

Writing the initial boilerplate code and setting up the project's folder structure.

Defining the 'what' and 'why' through user stories and functional requirements.

Deciding which frontend framework or database to use for the project feature.

Breaking down technical implementation into small, isolated git commits.

To track the history of all changes made to the codebase for auditing purposes.

To establish non-negotiable project principles that guide all subsequent technical decisions.

To list the contact information and roles of all project stakeholders.

To store the encrypted API keys and environment variables for the development team.

Brainstorming initial user stories and feature motivations with stakeholders.

Manually reviewing the final pull request before it is merged into the main branch.

Writing the individual helper functions and unit tests for the application.

Defining the architecture, technical stack, and data contracts for the feature.

It is a performance-tuning method that Spec Kit automates using machine learning.

It is high-uncertainty AI prompting that Spec Kit replaces with structured specifications.

It is a way to write code using audio cues that Spec Kit translates into documentation.

It is a collaborative pair-programming style that Spec Kit formalizes into a UI.

A breakdown of requirements into small, reviewable, and testable chunks of work.

A set of interview questions for hiring engineers with specific expertise in the stack.

A deployment script that pushes the local code to a cloud-based staging environment.

A comprehensive set of marketing materials and user guides for the feature launch.