Copilot Workspace Is GitHub's Take On AI-Powered Software Engineering

An anonymous reader quotes a report from TechCrunch: Ahead of its annual GitHub Universe conference in San Francisco early this fall, GitHub announced Copilot Workspace, a dev environment that taps what GitHub describes as "Copilot-powered agents" to help developers brainstorm, plan, build, test and run code in natural language. Jonathan Carter, head of GitHub Next, GitHub's software R&D team, pitches Workspace as somewhat of an evolution of GitHub's AI-powered coding assistant Copilot into a more general tool, building on recently introduced capabilities like Copilot Chat, which lets developers ask questions about code in natural language. "Through research, we found that, for many tasks, the biggest point of friction for developers was in getting started, and in particular knowing how to approach a [coding] problem, knowing which files to edit and knowing how to consider multiple solutions and their trade-offs," Carter said. "So we wanted to build an AI assistant that could meet developers at the inception of an idea or task, reduce the activation energy needed to begin and then collaborate with them on making the necessary edits across the entire corebase."

Given a GitHub repo or a specific bug within a repo, Workspace -- underpinned by OpenAI's GPT-4 Turbo model -- can build a plan to (attempt to) squash the bug or implement a new feature, drawing on an understanding of the repo's comments, issue replies and larger codebase. Developers get suggested code for the bug fix or new feature, along with a list of the things they need to validate and test that code, plus controls to edit, save, refactor or undo it. The suggested code can be run directly in Workspace and shared among team members via an external link. Those team members, once in Workspace, can refine and tinker with the code as they see fit.

Perhaps the most obvious way to launch Workspace is from the new "Open in Workspace" button to the left of issues and pull requests in GitHub repos. Clicking on it opens a field to describe the software engineering task to be completed in natural language, like, "Add documentation for the changes in this pull request," which, once submitted, gets added to a list of "sessions" within the new dedicated Workspace view. Workspace executes requests systematically step by step, creating a specification, generating a plan and then implementing that plan. Developers can dive into any of these steps to get a granular view of the suggested code and changes and delete, re-run or re-order the steps as necessary. "Since developers spend a lot of their time working on [coding issues], we believe we can help empower developers every day through a 'thought partnership' with AI," Carter said. "You can think of Copilot Workspace as a companion experience and dev environment that complements existing tools and workflows and enables simplifying a class of developer tasks ... We believe there's a lot of value that can be delivered in an AI-native developer environment that isn't constrained by existing workflows."

There is no "AI powered Software Engineering"

[gweihir]

There is even cheaper production of crappy, insecure and unreliable code. And that is not desirable at all.

Re:

[nightflameauto]

There is even cheaper production of crappy, insecure and unreliable code. And that is not desirable at all.

Techies often lose sight of what all these AI "assistants" are actually meant to do. It's not meant to help folks who have a grip on what they're doing. It's meant to give the management teams of large companies a warm fuzzy in the pocketbook. Why hire a competent software engineer when you can hire a half-competent newb and hand them an "AI" to do the heavy lifting. I look forward to a *LOT* more crapware being shoveled out the door in the near future. Hopefully we wake up before it becomes so entrenched a

Re:

[Nrrqshrr]

If anything, am optimistic about what these "AI" tools can do for the exact same reasons. We see this over and over again in video games and software development, where a small team or, hell even a solo-dev, can do what big companies refuse to touch with a stick.
In skilled hands, it will allow one good guy to do what he couldn't manage to pitch in a corporate office. And hopefully in the bad hands of HR drones and terrible product managers, this will cause the collapse of many dinosaurs that have been coast

Re:

[gweihir]

And hopefully in the bad hands of HR drones and terrible product managers, this will cause the collapse of many dinosaurs that have been coasting on past successes and haven't done anything good for humanity for the last 10 years.
But maybe am just being too optimistic.

Well, that would at least be one good outcome. MS came close last year to losing it all, for example. A gigantic house of cards is all they have.

Re:

[gweihir]

I completely agree. To all your points.

Re:

[Tony Isaac]

Managers at large companies are quite skeptical of AI, mostly because of it's steep price. They think it's a shiny new toy to get them to cough up more money.

As a developer, I love using GitHub Copilot. It allows me to write code significantly faster than I could if I had to go to Stack Overflow all the time, to look up code snippets and figure out which one works in my case, and then adapt it to my context.

I can't speak for what the assistants are "meant" to do. But I can confirm that Copilot is a tremendo

Re:

[Tony Isaac]

If someone uses AI to write code, and doesn't inspect the code for quality and security, they are doing it wrong. In that regard, you are right.

But what I find AI tools good for, is getting the ball rolling. For example, I recently needed to transform some XML into HTML in a we page. I wasn't very clear about how specifically to make that happen in JavaScript, so I asked GitHub Copilot to write the transform function for me. Sure, I could have looked it up in the documentation, but Copilot did it a lot fast

Get used to it

[twdorris]

Companies and, indeed, developers themselves are still figuring this stuff out. What works, what doesn't, which subset of our collective wet dream feature set actually makes sense and/or is actually needed? Nobody knows for sure.

What I do know, however, is that there are changes, probably BIG changes IMO, coming for all things development, management, knowledge related. Code, engineering, medical practice, legal practice, tax prep, I dunno....everything.

This statement from TFS is pretty fair in my experi

Re:

[AuMatar]

I'd say its absolutely useless for that job. SO you have something you don't know how to do, and don't know how to evaluate, and you have a program that sometimes kind of can do things, but its trained on random data with no actual understanding of what it's doing, and you're trusting it to work? I'll take the result of a google search over that any day of the week. At least then I'll get several options and can compare them to see what feels right. About the only people who could use this are so techni

Re:

[twdorris]

Yes, I'm trusting it 100%. That's totally what I said. Moron.

The early adopters have left slashdot

[Tony Isaac]

It used to be that slashdot was full of people who were eager to get their hands on the latest whiz-bang technology. Those days seem to be long gone.