[idan]

Hello! GitHub Next here, happy to answer questions and unpack how we think about AI tools for developers (spoiler: it's less about codegen and more about helping with the rest of the dev cycle — building an understanding of how the system works, clearly specifying how it should change, etc)

[roboyoshi]

Interesting stuff! I've been trying ollama/gpt + continue.dev and copilot in VSCode for a bit now and the chat-style assistant is a huge plus. Especially in DevOps (my main work) the codegen is rather unhelpful, but the ability to let LLMs explain some sections and help in rubber-ducking is huge.

I see that a good output requires a good input (prompt). How does copilot workspace determine a good input for the prompt? I see that in the github repo there is already a bunch of "Tips and Tricks" to get better results. What is your experience so far? Should we change our way of creating issues (user-stories / bug-reports, change-requests) to a format that is better understood by AI/Copilot? (half-joking, half-serious).

[idan]

Well, that's basically the heart of Copilot Workspace! The whole UX is structured to make it easy for the human to steer.

- Alter the "current" bullet points in the spec to correct the AI's understanding of the system today - Alter the "proposed" bullet points in the spec to correct the AI's understanding of what SHOULD be - Alter files/bullets in the plan in order to correct the AI's understanding of how to go from current to proposed.

That said, I think there's definitely a future where we might want to explore how we nudge humans into better issue-writing habits! A well-specified issue is as important to other humans as it is to AI. And "well-specified" is not about "more", it's about clarity. Having the right level of detail, clear articulation of what success means, etc.

[rjindael]

If it is more about code planning, how much different is it than simply telling ChatGPT the overall structure of your code and asking it to give you a rudimentary plan on what to do next? Would it be able to actually execute steps of the plan by generating code, then creating PRs for it? I feel like this is a great tool for our team since my understanding of the announcement is that it is more or less functionally equivalent to hiring another programmer on your team (or, if not that, at the very least having a really useful assistant.) Kudos to the GitHub team and I have immensely enjoyed using Copilot thus far to increase my productivity :-)