Yeah, when you had multiple agents working on the same machine, branch isolation was no longer sufficient.
A repository folder can only be on one branch at a time.
A worktree is basically equivalent to a cp -R + git branch, which allows this new workflow to occur.
I loved this particular historical insight as to why `git worktree` was added in 2015:
Before worktrees, kernel devs faced a major inconvenience when switching contexts, e.g., stopping feature work to fix an urgent bug on a release branch.
Running git stash and switching branches alters timestamps on thousands of files.
This forces `make` to perform a full re-compile, which can take up to an hour on large kernels.
I use Conductor's spotlight (and a Pi extension I derived from it) but it's not perfect. Once a migration executes, I'm pretty much bound to it but at least once, I have just backed up the database before switching and then restored that.
Hah, I have a prototype of the same idea on my backburner! Excited to see this, though I don't understand some of their design choices. Will need to check out more closely.
I've toyed around with worktrees but haven't found them very useful beyond that. I generally find it much easier to carefully prompt an agent so $TASK1 does not interfere with $TASK2
I set up multiple work trees in one vscode workspace last year and wrote in the agents.md how to merge branches - but I spend about a third of the time helping agents integrate and merge. I remember wishing the tooling would catch up
i have some fun experiments i'm doing with full virtualization middle ware of all sys calls for agents tools/shell commands/io, still far from daily driver, but allows me to do a very rich overlay / virtual file system tom foolery in place
The reason these tools exist is not because of non-professional developers, but quite the opposite.
A lot more professionals are now working on more projects simultaneously- something that was not practical just a year ago.
Though, while this is nice, considering that all of the action is happening on the same device, I am worried this is going to increase supply chain risks. Before, a developer would work on clearly designated projects for practical reasons. Now, the same developer can work across many projects that are quite different - for example, the marketing site and the backend - and because of an obscure and unimportant component on the marketing site, there can be an impact on backend systems.
If I'm interpreting this correctly, GitHub will use their existing actions infrastructure to run versions of the code in isolated worktrees. I think this could be a very beneficial process.
What I've done on my end is created a project where I have a remote Linux workstation. I can create multiple worktrees for each repo in that workstation, and then my agent can push PRs to GitHub and use the actions infrastructure to see if the integration tests that it writes for itself are successful without needing to run those integration tests on the local environment. It's expensive in terms of runner hours, but the automaticity of it is incredible.
> Looks good, but after pricing change I have already used 26% this month with very light usage.
Forced me to switch to OpenCode Go (not Zen), which isn't quite as nice as Copilot Chat deeply integrated into VSCode / Intelli but the models (DeepSeek v4 Pro, GLM 5.1, & Qwen 3.7 Max) are able replacements for Claude Sonnet 4.x / GPT 5.x while being way cheaper (especially DeepSeek).
I rather like Ace better because the key problem right now is teams not working together and shipping the wrong things. When AI can generate the code, then it feels like product should be bringing the functional vocabulary and grammar while the engineering team provides the technical grammar to build the right thing.
This app is just another "let me talk to product, copy their convo, go off and build this in isolation with an agent" which I think is directionally wrong.
The "rooms" or "streams" should be multi-player instead of product looking at it at the end saying "no, go fix that" and dev copies text from one source and pastes into another.
It's weird. I still remember 2008, when GitHub's claim to fame was that it was "the easiest (and prettiest) way to participate in the collaborative development of software."
Now they want to end that collaboration, and turn it into automation. Many C-suite executives right now are smiling bigly. Meanwhile, we're leading the exodus. Turns out, we still want the easiest and prettiest way to participate in the collaborative development of software, and GitHub ain't it!
Unrelated to the feature itself, but remember a few months ago when someone posted Github's beta feature for stacked PRs, and a ton of people slammed them for releasing a seemingly vibe-coded site? To quote Mitchell Hashimoto, "One of the most requested GitHub features in years and the website looks like it was designed by someone 9 years into a 2 year community college program."[1]
When opening the posted link, my first thought was "imagine if the stacked PRs site had the same amount of effort put into it as the Github Copilot App site". They clearly have other preview features on this site already, so maybe I'm just confused on why stacked PRs got some b-grade announcement site. The obvious answer is "copilot", but I'm still curious.
Target market for stacked PRs are ICs who don't have much decision making power and let's be real do not care too much about the look and feel of a "launch site" for the feature. It's also something few if anyone is making a purchasing decision over.
Target market for copilot includes people with actual purchasing power and also many new users where this is an actual make or break feature. So this is worth the investment into design while stacked PRs is questionable. I actually question why they bothered with anything more than a blog post at all for stacked PRs (looking at the post it doesn't seem like too too much more than a blog post though).
I predict next “Windows Copilot App” (not to be confused with any of the others), Microsoft’s AI that controls your computer (like OpenAI Operator and Claude Computer)
I know it has the same functionality, but it also looks like the Codex app which looks like Cursor Agents! Are they sharing some VS Code primitive here?
It's kind of interesting that everyone is going for the desktop app format now.
These desktop agentic coding tools are a large UX step up from the CLIs, but I still think the future is going to be remote development as the coding agents start running for hours at a time. Building a desktop app seems short-sighted as it would just lock them out of the remote option completely.
Doesn't lock you out at all. Codex already had a companion app for mobile so you can send prompts to your desktop app while you go about your business. The infrastructure is there. Server might move from your desktop to cloud at some point but not much changes. Still needs somewhere to run.
A repository folder can only be on one branch at a time.
A worktree is basically equivalent to a cp -R + git branch, which allows this new workflow to occur.
I loved this particular historical insight as to why `git worktree` was added in 2015:
Before worktrees, kernel devs faced a major inconvenience when switching contexts, e.g., stopping feature work to fix an urgent bug on a release branch.
Running git stash and switching branches alters timestamps on thousands of files.
This forces `make` to perform a full re-compile, which can take up to an hour on large kernels.
But!
The reason these tools exist is not because of non-professional developers, but quite the opposite.
A lot more professionals are now working on more projects simultaneously- something that was not practical just a year ago.
Though, while this is nice, considering that all of the action is happening on the same device, I am worried this is going to increase supply chain risks. Before, a developer would work on clearly designated projects for practical reasons. Now, the same developer can work across many projects that are quite different - for example, the marketing site and the backend - and because of an obscure and unimportant component on the marketing site, there can be an impact on backend systems.
I wrote more about this here: https://chatbotkit.com/reflections/everyone-is-a-vip-now
What I've done on my end is created a project where I have a remote Linux workstation. I can create multiple worktrees for each repo in that workstation, and then my agent can push PRs to GitHub and use the actions infrastructure to see if the integration tests that it writes for itself are successful without needing to run those integration tests on the local environment. It's expensive in terms of runner hours, but the automaticity of it is incredible.
Last month I used Copilot heavily, much much more than I usually do, but did not manage to use more than 58%.
Forced me to switch to OpenCode Go (not Zen), which isn't quite as nice as Copilot Chat deeply integrated into VSCode / Intelli but the models (DeepSeek v4 Pro, GLM 5.1, & Qwen 3.7 Max) are able replacements for Claude Sonnet 4.x / GPT 5.x while being way cheaper (especially DeepSeek).
I swear I did few tests and it's rare you need more than medium on mundane job work.
Edit: This short talk – https://maggieappleton.com/zero-alignment
the copilot app looks more like just a straight clone of the codex UI (and the new antigravity UI, which is also a straight clone of the codex UI)
Particularly the left sidebar and conversation view look near identically structured.
This app is just another "let me talk to product, copy their convo, go off and build this in isolation with an agent" which I think is directionally wrong.
The "rooms" or "streams" should be multi-player instead of product looking at it at the end saying "no, go fix that" and dev copies text from one source and pastes into another.
Now they want to end that collaboration, and turn it into automation. Many C-suite executives right now are smiling bigly. Meanwhile, we're leading the exodus. Turns out, we still want the easiest and prettiest way to participate in the collaborative development of software, and GitHub ain't it!
Go here and you’ll be redirected with a query string including a customer parameter. That looks like trouble.
https://github.com/settings/billing/usage
[1]: https://github.blog/news-insights/company-news/an-update-on-...
When opening the posted link, my first thought was "imagine if the stacked PRs site had the same amount of effort put into it as the Github Copilot App site". They clearly have other preview features on this site already, so maybe I'm just confused on why stacked PRs got some b-grade announcement site. The obvious answer is "copilot", but I'm still curious.
[1] https://x.com/mitchellh/status/2043788123008868600
Target market for copilot includes people with actual purchasing power and also many new users where this is an actual make or break feature. So this is worth the investment into design while stacked PRs is questionable. I actually question why they bothered with anything more than a blog post at all for stacked PRs (looking at the post it doesn't seem like too too much more than a blog post though).
I predict next “Windows Copilot App” (not to be confused with any of the others), Microsoft’s AI that controls your computer (like OpenAI Operator and Claude Computer)
Other than fewer features.
These desktop agentic coding tools are a large UX step up from the CLIs, but I still think the future is going to be remote development as the coding agents start running for hours at a time. Building a desktop app seems short-sighted as it would just lock them out of the remote option completely.
There's support in VS Code and Jetbrains IDEs. You can access your agent sessions on the web.
(I work at GitHub, but not on Copilot)
Plus the whole naming confused people
I still talk to co-workers who think claude code == agents and copilot is just VS autocomplete
And after they will accuse the growth and all to be responsible for their stability issues...