GitHub details how to build initial commands and remote control in the Copilot app

The experience of opening a new artificial intelligence tool can be like standing in front of a blank page, where formulating the desired question becomes a task in itself. Kayla Cinnamon, an advocate for artificial intelligence developer tools at GitHub, explains that starting does not require writing a perfect command from the first moment. Text commands are just a description of what the user wants to achieve, it is enough to start with the available information, provide the application with the appropriate context, and then gradually modify the request as work progresses within the GitHub Copilot application.

To accomplish modifications, it is necessary to link intelligent agent sessions to a code repository on GitHub or a local folder on the computer. This connection gives the GitHub Copilot application direct access to the required codes and files to accomplish the task. Developers can choose a previous project from the application's main screen or add a new local folder, where providing the session with programming context is a fundamental step that precedes writing the first command.

Formulating programming commands does not require learning a special sentence structure or searching for the best way to express, but rather describing the desired change in direct natural language is enough.The article gives an example of this by requesting the addition of a sorting option for games by funding to a game list. The application examines the project, determines the relevant parts in the codebase, and then implements the desired change. Formulating commands is an iterative process that does not require predicting all details in advance, the key is to start with the available environment and then continuously improve.

The GitHub Copilot application allows users to choose the artificial intelligence model that manages the required task. The available models vary in their strengths, some are characterized by the ability to think logically, while other models specialize in processing simple tasks more quickly. The default model is a suitable choice for starting without needing to understand all the technical differences in advance, with the possibility of switching between models while working when the task requires more complex inference or when there is a need to change the approach to achieve the desired results.

Writing on the keyboard is no longer the only way to enter commands, as the application includes an integrated voice input feature. This feature benefits developers when thinking aloud about solving problems or when needing to explain long requests. Speech is converted to written text inside the command box, allowing the user to review and modify it before actual submission, thus thinking and experimenting without strict commitment to the first formulation.

Sessions provide additional flexibility by allowing customization of how work is managed, where the designated intelligent agent can be changed from the session title and agents prepared for different types of programming tasks can be chosen.The remote control feature also allows sessions to be run over the web, rather than being limited to the local device only.This option allows the developer to start the programming task, then close the laptop, and return to the session later from another device without losing progress or work location.

The article emphasizes the importance of starting with small changes within a known project, describing what is required in clear language, and then following up on the results and modifying requests as needed. Switching between models and adjusting the session operation mode are auxiliary tools that become smoother with continued use, where the most important thing is to start directly and assign the first task to the application.