The article delves into the two primary approaches for creating a GraphQL service: schema-first and code-first, discussing their advantages and disadvantages. In the schema-first method, the focus is on defining the schema using the Schema Definition Language (SDL) before implementing the code, which facilitates clear communication and collaboration across teams but can lead to code duplication and a lack of dynamism. Conversely, the code-first approach begins with coding the resolvers, creating the schema as an artifact, and allows for a dynamic and unified source of truth that can accommodate complex and expandable schemas but might be less readable. Different implementers advocate for each method based on their specific benefits, with schema-first being likened to test-driven development and promoting modularity, while code-first is praised for reducing reliance on tooling and supporting dynamic features. The author expresses a preference for the code-first approach due to its flexibility and capacity for innovation, predicting it will gain more popularity in the future while acknowledging that the choice between the two should depend on the specific needs of a project.