Lessons Learned from Building GitHub Copilot(s)

Title: Lessons Learned from Building GitHub Copilot(s)
Date: October 2, 2024
Duration: 1HR

SPEAKER
Eddie Aftandilian
Principal Researcher, GitHub Next

MODERATOR
Rahul Pandita
Staff Reseacher, GitHub Next

Registration Link

GitHub For Dummies, 2nd Edition (free Percipio book for ACM Members with Skills Bundle)
Hands-on GitHub Actions: Implement CI/CD with GitHub Action Workflows for Your Applications (free Percipio book for ACM Members with Skills Bundle)
Complete Git Guide: Understand and Master Git and GitHub (free O’Reilly course for ACM Members with Skills Bundle)
Learning GitHub Actions (free O’Reilly book for ACM Members with Skills Bundle)
GitHub Fundamentals (free Pluralsight course for ACM Members with Skills Bundle)
Measuring GitHub Copilot’s Impact on Productivity

1 Like

I have a question for the talk: could you please discuss the long-term vision for GitHub Copilot? How do you foresee its features and capabilities evolving as AI continues to develop into the future?

1 Like

Please include following in your talk…

  • Privacy of Devs and the source code
  • IP rights and how that works
  • How the CoPilot monitors the source code (changes in multiple files)
    • Is it constant communication with server
    • cost
    • usefulness vs cost
  • Agents in CoPilot
    • Tester, Typo checker, etc…
    • Algo Improver, etc
    • Design helper
1 Like

1.) Do you see the human pair programming paradigm of “Driver-Navigator” being an effective mental model for working with Github Copilot? What key differences should we keep in mind if we’re coming from a human pair programming background?

2.) Similar question when working with the “RISEN” or “ROSE” prompt structures that many use with other chat interfaces: how do you think these approaches should be applied to using a more specialized tool like Github Copilot?

3.) How do you plan to keep up with advances in Open AI’s models? What are the main challenges in integrating new model versions with Github Copilot?

Hello
Please can we consider having a Copilot for GHClassroom Codespaces. One that is tuned to not provide answers but support students in formulating their own responses through supportive pedagogy.

In recent years, AI developer tools have progressed from research projects to critical parts of the developer toolset.

can you justify the “critical” claim?