Exploring the Future of Software Engineering with GitHub Next

As we embark on a journey of exploration and innovation at GitHub Next, we're excited to share our latest findings with AI-powered software development - from code completions to task completions, and beyond.

  • 1. GitHub Next is a team of researchers, senior developers, and code-focused builders who work on exploring the future of software engineering.
  • 2. The team consists of about 20 people and reports directly to the CEO.
  • 3. Their goal is to explore new technologies and trends in software engineering and pass their learnings on to product and development teams.
  • 4. One of the areas they are focusing on is the impact of AI on software development, with Andrew Ng stating that AI is like electricity and has the potential to transform the field.
  • 5. The team's approach is to rapidly prototype and experiment with different ideas, then release them as tech previews or incorporate them into existing products based on user feedback.
  • 6. One example of a successful exploration is GitHub Copilot, an AI-powered code completion tool.
  • 7. The team is currently working on two active explorations: Next Edit Suggestions in the Co-pilot Workspace and Task Completions SL.
  • 8. Next Edit Suggestions provides suggestions for editing and updating existing code, rather than just completing new code.
  • 9. The idea behind this exploration is that while code completion is useful, most of the time developers are editing or updating existing code.
  • 10. The team is still experimenting with different aspects of this exploration, such as using natural language to specify program functionality and improving code completions with runtime information.
  • 11. Task Completions SL aims to provide AI support for all aspects of the "inner loop" of software development, including getting started on a task, trusting the output of the AI, problem-solving, and
  • 12. The team has gathered feedback from developers and found that the biggest difficulty they face is getting started on a task and trusting the output of the AI.
  • 13. Developers also want to retain control and view the AI as a thought partner or sparring partner for problem-solving.
  • 14. The Co-pilot Workspace addresses these pain points by simplifying the process of getting started on a task, providing a built-in runtime for quickly verifying code, an environment for iteration, a
  • 15. The team is also working on other improvements, such as faster file completions and synthesizing terminal commands in a natural language format for mobile use cases.
  • 16. Other active explorations include rethinking developer learning with AI and the impact of AI-generated code on code reviews.
  • 17. The team plans to continue exploring these and other areas, with some explorations resulting in tech previews or being killed if they are not successful.
  • 18. The goal is to jointly explore the future of AI in software development and enjoy its benefits as soon as possible.

Source: AI Engineer via YouTube

❓ What do you think? What are your thoughts on the ideas shared in this video? Feel free to share your thoughts in the comments!