Transcript

00:00:00With the rise of coding CLIs, there are a ton of different tools out there that help manage your AI agents to write your code for you.
00:00:06In today's video, I'm going to show you some of the main ones and how they compare.
00:00:10What I use day-to-day myself right now is Conductor, but this is quite flexible and I'm often jumping around between tools.
00:00:16Some other tools that are starting to become popular include SuperSet, C-Mux, the Codex app is actually quite nice,
00:00:23and then a lot of people are just writing in regular terminals like Ghosty, Warp, or Iterm.
00:00:27And of course, people are still using tools like Cursor or VS Code.
00:00:31I'll focus on Conductor, SuperSet, and C-Mux in this video because these are quite new tools that not that many people have tried.
00:00:36Everyone knows how to use VS Code or Cursor, so I won't go into them too deeply.
00:00:40So first up is Conductor. This is actually what you see on my screen right now.
00:00:44So you can see I have a whole bunch of different chats running in parallel.
00:00:48This is real work that I'm doing right now for Inbox Zero.
00:00:50Each of these chats down the side is a different work tree.
00:00:53If you're not familiar with work trees, they became very popular due to AI agents running in parallel.
00:00:58And what it does is it takes your core project and basically clones it into a new folder in your computer.
00:01:03So here you can see I've got the Inbox Zero repo and I've just got lots of copies of it on the computer.
00:01:08This one's called Belgrade, Algiers.
00:01:10Each one has a different name, and that is basically what you're seeing over here in my sidebar.
00:01:15Each one is basically its own branch that I can work on.
00:01:18It's got its own PR as well.
00:01:20So the changes over here aren't impacting any of the changes over here.
00:01:25It's as if you have seven different employee workers working for you, different AI agents,
00:01:30and each one is changing different files and can work in parallel without stepping on each other's feet.
00:01:35Now to show you around how Conductor works, it's just a regular chat.
00:01:38So here you can see I've got one task in progress, so I'm telling it the AI to continue its work.
00:01:43And you can see once the task is actually ready and complete for me, I can go and check the PR that it's attached to.
00:01:49I can also click here to merge it.
00:01:51I can click a review button over here, which is going to just spawn up a new agent in a new tab to go and review the changes in this work tree or branch.
00:01:59I can also see the different files.
00:02:01So here is the .env example file.
00:02:03And over here on the right, I have a regular terminal.
00:02:07Now, because of how work trees work, they're copying and pasting your main project into a new location.
00:02:13And that often means you need to do a new setup for that folder.
00:02:17So something you'll see in all these projects that we're covering today is there's a setup script and a run script.
00:02:24A setup script basically means when I hit plus to start working on a new feature in a new work tree, it will go and do something like a PMPM install or even run the project on a new port.
00:02:35It's up to you how you use those.
00:02:36But typically what you'll see is the setup script is maybe doing a PMPM install and then the run script is actually running the app and then you'll be able to visit it on localhost 3000 or some other port.
00:02:46You could have every single work tree running on a different port for you.
00:02:50And this is necessary because each work tree has different changes.
00:02:54So if you run the project in one folder, it's obviously not going to show you the changes in another folder on your computer.
00:03:00Now, another thing which is really nice is you can see the checks.
00:03:04So your build or your GitHub comments.
00:03:07So here you can see I've got lots of checks running and you can see I've got some comments from Kubik and Baz, which are AI code review tools.
00:03:15And if I click on one, you can see here's a comment that it's given me as part of its review.
00:03:19If I want, I can even add all these comments to the chat over here and have the AI fix it up automatically by hitting send.
00:03:25The next step I want to show you is super set.
00:03:28Here you can see it's a very similar idea.
00:03:30You have all your different agents running down the side.
00:03:32Frankly, I think a ton of different projects have been inspired by Conductor, which is why you see them all converging on a similar interface.
00:03:39You can see the ability to commit on the right.
00:03:41You have different files of the project.
00:03:44And what's interesting here is there's more of a first class experience for the terminal.
00:03:49So if you like to just run regular Claude code or codex, you can do that.
00:03:54Here you can see a regular run that's happened.
00:03:56They do have their own chat interface as well.
00:03:58So if you prefer that, I think this is quite new.
00:04:00Conductor pushes you to using their custom chat interface with super set.
00:04:04You have that flexibility.
00:04:06Now, the reason you might want to use the real version of code code or codex is that it will have all the latest features.
00:04:13What I mean by that is there could be an update to codex and there's a new slash command.
00:04:18You do a slash here in Conductor and it just doesn't appear because they haven't updated Conductor.
00:04:22It is largely a wrapper around code code and codex, but it's not one to one because you're using their chat interface.
00:04:29But over here, it's a similar idea using the chat interface from super set, but you can also use the CLI as well directly.
00:04:36Something else that you'll find in super set is that you can also work off of the main branch.
00:04:40Now, surprisingly, that's a little bit annoying to do in Conductor.
00:04:43You can't work off of main in Conductor as far as I'm aware.
00:04:47You have to open up the project in something like cursor or vs code.
00:04:52And by the way, while I'm here, Conductor has this exact same thing.
00:04:54But here you can see I can open up the project from super set directly in some other editor like zed or whatever I choose to open.
00:05:01Exact same idea over here.
00:05:03I can open it in finder or ghosty or whatever I want.
00:05:06Another cool part of super set is that you can open up browser tab so he can see inbox zero and this is running directly within super set.
00:05:15This can be helpful if you're doing front end code.
00:05:18I should mention as well that both these apps, they sync with linear and GitHub.
00:05:22So you can create a new work tree based off of a linear issue, for example.
00:05:27And here are some other features that super set has.
00:05:29I think these will be really interesting when cloud workspaces and the mobile app launch.
00:05:34I assume Conductor is going to do something similar as well.
00:05:36These apps change very quickly, as you can imagine.
00:05:39And what we're seeing here today will be quite a different story in three months time.
00:05:42If I show you how we actually use super set, you can see I can choose between codex or cloud code or frankly, lots of other options like open code.
00:05:50I guess this is another advantage of super set over Conductor.
00:05:54With Conductor, you're limited to only codex and cloud code.
00:05:57Here I have lots of other options that I can use.
00:05:59So let's go and make that change, create a workspace.
00:06:03You can see the workspace is setting up, doing a few things.
00:06:06It's opened up codex over here and you can see at the same time it's doing a PMPM install over here in another pane.
00:06:15Now that it's done, I'm going to ask it to create a PR.
00:06:19By the way, if you note down the side, our branch has been renamed.
00:06:22We can open a preview of it in the browser, which is nice.
00:06:26And you can see the PR is ready.
00:06:27It's linked it for me over here.
00:06:29But I can also see down the side that we're at PR 1915 and you can see the checks are running very, very similar experience to Conductor.
00:06:37I can also click here and see the changes or the commits.
00:06:40If I want to see the diff, I can click in here and hopefully I can see the diff down here and you can see the CI checks are also running.
00:06:48This is happening in things like GitHub actions and other bots I have connected.
00:06:52You can see the Vercel bot and socket security are already finished and the AI code reviewers are still running as well as some end to end tests.
00:07:00Something to be aware of when using Superset and Conductor.
00:07:03You can do search files, but I don't believe there's a way to search for specific words.
00:07:07So often you are opening up the code in another editor.
00:07:10Let's say I wanted to find this in the project.
00:07:12I don't think there's a way for me to do this in Conductor or Superset today.
00:07:17It's not that important, but there are a whole bunch of times I will be opening up another editor to more deeply see what's going on.
00:07:24But for the most part, day to day, I'm just in Conductor managing my different agents down the sidebar.
00:07:29And by the way, here you can see this one.
00:07:31This change is now being created.
00:07:33I can click create PR could also click review and it just continues with the process.
00:07:37So it has a lot of nice built in buttons that make the workflow just much easier to use.
00:07:42Cmux is the next tool we're going to look at.
00:07:44But before that, let's take a quick look at what Ghosty looks like.
00:07:46Ghosty is just another terminal.
00:07:48It's cool. You should download it.
00:07:49It's very performant. It's built in Zig.
00:07:51Here you can see Codex running in Ghosty.
00:07:53The exact same thing. This is Claude code running in iTerm2.
00:07:56And here you see we have warp running.
00:08:00Also a cool terminal. This one does stand out a bit because you can use AI with your terminal.
00:08:04And it does a whole bunch of other things like agentic coding.
00:08:07I haven't used it a ton for that, so I'm not going to go into it.
00:08:10But it is like a very nice terminal to use via natural language instead of having to type commands.
00:08:16And here you can see what warp looks like in practice.
00:08:18But jumping over to Cmux, we can see that Cmux is a native Mac OS app built on Ghosty.
00:08:24Vertical tabs, notification rings when agents need attention, split panes and socket API for automation.
00:08:30This is what it looks like. You can see lots of tasks running down the side.
00:08:33You can also see lots of panes over here running.
00:08:37And you can even see a browser tab running.
00:08:39And this is what Cmux looks like in practice when you first open it up.
00:08:43It's just a Ghosty terminal which is very performant, which is great, but then you can start to do other things.
00:08:48So this starts to work more like Tmux. If you're not familiar with Tmux, this is the idea behind Tmux.
00:08:53Here you can see Tmux running in my terminal.
00:08:55I can do things like different panes within Tmux itself. It's a terminal within a terminal.
00:09:00Also known as a terminal multiplexer.
00:09:03And here you can see I've actually gone and run Tmux within Cmux, but we're going to go and leave that.
00:09:08And here we have Cmux running. And you can see I can keep splitting it into different windows.
00:09:13And so this is really nice. If I want to run four different codecs in parallel, I can do that.
00:09:18And I don't need to have four different windows that I'm managing.
00:09:21Of course, I can do another terminal over here and I can jump back and forth between the terminals.
00:09:27And one of the really cool things about Cmux is that it's a CLI that you can call.
00:09:32So, for example, here Cmux new split left and I can get it to create another tab for me.
00:09:37And you can imagine if you're using Clawed Code, it could, for example, spin up multiple different agents in lots of different tabs for you.
00:09:44And you can also get Clawed Code to use the Cmux CLI to control other things.
00:09:49So, for example, here you can see we're on PR 1905.
00:09:52And so I was able to have that set up by just telling Clawed Code, "Hey, I want to see the PR in the sidebar over here."
00:09:59You can also get it to handle things like notifications for you automatically and customize those.
00:10:04So, for example, when Clawed Code or codecs is finished with a task, they can call the Cmux CLI to send you a notification to tell you which tab it's in and that it's done.
00:10:14And you can also use it for things like running the browser and browser automation.
00:10:18So, for example, here I've done Cmux browser open get inbox zero dot com.
00:10:21And you can see it's popped up in a new tab over here.
00:10:24Take a look at the docs for more about what it can do.
00:10:27Here you can see the browser automation CLI.
00:10:30Here you can see how it manages notifications for you.
00:10:33Cmux notify. Let's try this one out.
00:10:36And you can see we've got a new notification. That's great.
00:10:39Now, there are some differences between Cmux and Cmux.
00:10:42Tmux is really great for, let's say, remote session work.
00:10:46If your session gets disconnected from your server, so Tmux will keep running.
00:10:50Cmux is Mac only, so it's not going to work on your server, probably.
00:10:54But this also means if you're running it on your computer and then you want to continue work on your phone, it won't help you there.
00:11:00Now, I personally use a tool called Echo on my phone. It's really cheap, costs three dollars.
00:11:05But basically what it allows you to do is SSH into your computer.
00:11:08And if you already have a Tmux Claude code session running, you can just continue it directly for your phone.
00:11:14You can have your computer and your phone on the same tail scale network to get that set up.
00:11:18And here you can see what that experience looks like. And it's actually quite nice.
00:11:21I don't use it a ton, but the app itself is super simple, super polished.
00:11:25So I recommend taking a look at that as well.
00:11:27That's the end of the video. I hope this gives you some good ideas on
00:11:31the different tools that exist out there and how you can use them.
00:11:34Frankly, they all do almost the same thing.
00:11:37Ultimately, you're just chatting with your AI via a slightly different interface.
00:11:41To me, the most important thing is can I run multiple agents in parallel?
00:11:45The way I found that easiest to do is using work trees.
00:11:48So I should actually mention Cmux doesn't have anything native around work trees.
00:11:52So you do need to look around for scripts to set that up.
00:11:55But if you're using something like Claude, you could just hit Claude work tree
00:11:58and it will set things up in a new work tree for you.
00:12:01Work trees are built into Git, so literally every app you're using can use them.
00:12:05But something like Conductor or Superset where you just hit plus
00:12:09and you get a new work tree and everything set up for you automatically,
00:12:11you don't have to worry about that. So that's very nice.
00:12:14If you're using something like Cmux, there are scripts online that can help you do the same thing.
00:12:17And frankly, there's a few lines of code to help you do the same thing.
00:12:20I hope you enjoyed the video and found it interesting.
00:12:23Be sure to subscribe for future videos. And if you want to do me a favor,
00:12:26check out GitInboxZero.com. Start the repo on GitHub. It's fully open source.
00:12:31It's an AI email assistant that helps you manage your email and helps you get to inbox zero fast.

Key Takeaway

Modern AI coding workflows are shifting toward parallel execution via Git worktrees, managed by specialized wrappers like Conductor or multiplexers like cmux that facilitate multi-agent collaboration.

Highlights

Comparison of modern AI-driven coding interfaces including Conductor, Superset, and cmux.

The central role of Git worktrees in enabling multiple AI agents to work in parallel without file conflicts.

Overview of setup and run scripts required to manage environment dependencies across different local project clones.

Detailed look at cmux as a native macOS terminal multiplexer built on Ghosty with unique automation APIs.

Alternative mobile workflows using SSH and the Echo app to continue AI agent sessions on the go.

The trend of AI coding tools converging toward a sidebar-based multi-agent interface.

Timeline

Introduction to AI Coding CLIs and Agents

The speaker introduces the rapidly evolving landscape of command-line interfaces designed to manage AI coding agents. While popular tools like Cursor and VS Code are well-known, this video focuses on newer, more flexible alternatives such as Conductor, Superset, and cmux. The core benefit of these tools is their ability to automate the writing of code through specialized agentic workflows. The speaker notes that they personally use Conductor daily but often rotate between different tools depending on the specific task. This section sets the stage for a deep dive into how these platforms optimize developer productivity.

Deep Dive into Conductor and Git Worktrees

Conductor is showcased as a multi-chat interface where each sidebar entry represents a distinct Git worktree. Worktrees allow the project to be cloned into separate folders, enabling multiple AI agents to work on different branches simultaneously without interfering with one another. The interface includes built-in features for checking pull requests, merging code, and initiating AI-driven code reviews. A critical component discussed is the use of setup and run scripts to handle environment configurations like 'pnpm install' for each new worktree. The speaker demonstrates how AI comments from tools like Kubic and Baz can be directly imported into the chat to trigger automatic fixes.

Exploring Superset and Flexible Agent Options

Superset is presented as a flexible alternative that offers a similar parallel agent interface but with a more first-class terminal experience. Unlike Conductor, Superset allows users to work directly off the main branch and provides the freedom to use various CLI tools like Claude Code, Codex, or OpenCode. It features an integrated browser for frontend previews and syncs natively with project management tools like Linear and GitHub. The speaker highlights that Superset serves as a more direct wrapper, ensuring users have access to the latest slash commands and features of the underlying AI models. This section emphasizes the importance of choosing a tool that doesn't lag behind the rapid updates of AI CLIs.

cmux: The Native macOS Terminal Multiplexer

The video shifts to cmux, a native macOS application built on the Ghosty terminal that functions as a terminal multiplexer. It allows developers to split panes and manage multiple terminal sessions, such as running four Codex instances in parallel within a single window. A standout feature of cmux is its socket API, which enables AI agents to programmatically control the terminal, open browser tabs, or send system notifications. The speaker demonstrates using the 'cmux notify' command to alert the user when an agent has completed a long-running task. This tool is particularly suited for power users who prefer a highly customizable, scriptable environment over a locked-in GUI.

Remote Workflows and Final Comparisons

The final section covers the limitations of cmux, such as its lack of native worktree management and its restriction to macOS. To solve the problem of remote access, the speaker recommends the Echo app, which allows users to SSH into their machines and continue Tmux sessions from a mobile phone. They reiterate that while these tools have different interfaces, they all serve the primary goal of facilitating natural language interaction with code. The speaker concludes by encouraging viewers to prioritize tools that support parallel agent execution and mentions their own project, Inbox Zero. Ultimately, the choice of tool depends on whether the developer prefers a streamlined GUI or a flexible, scriptable terminal.

Community Posts

View all posts