Evolving GitHub Copilot’s next edit suggestions through custom model training

Next Edit Suggestions (NES) is a custom model integrated into GitHub Copilot to predict logical code edits, enhancing developer productivity within Visual Studio Code. Since its debut, NES has undergone significant updates, driven by challenges in predicting edits, which demand understanding developer intent without explicit prompts and maintaining a balance between suggestion speed and quality. The development of NES required a shift from relying on pull request data to collecting real-time editing behavior, facilitating the creation of a more accurate dataset. Utilizing supervised fine-tuning and reinforcement learning, NES achieved improved model quality by addressing the limitations of traditional training methods and leveraging unlabeled data. The continuous refinement process, influenced by developer feedback, has led to releases that reduce latency, improve suggestion quality, and adjust model eagerness based on user interactions. Future enhancements aim to offer cross-file suggestions, faster responses, and smarter edits by anticipating context and dependencies. The development of NES exemplifies an "AI-native" approach, emphasizing end-to-end integration and collaboration between model training, UX design, and the VS Code team to prioritize the developer experience.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.