2026/01/11
I was sad and unsurprised to get an email from Terragon announcing that they were sunsetting the product.
I was a happy user for the two or so weeks in which it was first the only and then the most polished web interface for passing a frontier model to your codebase.
It was clear then that their moat was thin, and that the model providers themselves would soon cannibalize their UX. I don’t blame them for shutting down; it, like many other of these tools, was a bit of a side project for the creators that exploded in popularity rather than a deliberate years-long bet.
I bring this up as a way of transitioning to talking about Conductor, which is a tool that I’ve been using for the…
2026/01/11
I was sad and unsurprised to get an email from Terragon announcing that they were sunsetting the product.
I was a happy user for the two or so weeks in which it was first the only and then the most polished web interface for passing a frontier model to your codebase.
It was clear then that their moat was thin, and that the model providers themselves would soon cannibalize their UX. I don’t blame them for shutting down; it, like many other of these tools, was a bit of a side project for the creators that exploded in popularity rather than a deliberate years-long bet.
I bring this up as a way of transitioning to talking about Conductor, which is a tool that I’ve been using for the past few days and enjoying a fair bit.
Conductor is essentially a lot of very nice duct tape around Claude Code and git worktrees to let you run multiple longer tasks in parallel without worrying about conflicts. (There are some other niceties: for instance, I really like their Spotlight abstraction, which lets you pull in a worktree into your main branch so you don’t have to figure out spinning up dev servers and things of that nature.)
But Claude Code Desktop already has worktree support: the app itself is a barely usable interface at the moment, but largely for tractable reasons, and it’s a question of weeks rather than months until they slurp up the most obvious innovations Conductor et al have developed.
It’s tempting to think that this is a forever cycle: smaller apps develop innovations in UX, larger apps copy them and then compete them out of existence. 1This dynamic made even odder by the fact that the smaller apps delegate 95% of the “real” functionality to the larger apps. I think this is true in a generalist sense, but there might be more moat on the margins.
Per Stack Overflow, the most successful non-VS IDEs fall into one of two buckets:
- Legacy (Notepad++, Vim, Nano)
- Single-language (IDEA, PyCharm, Android Studio, Jupyter Lab)
We’re still in the very, very early days of innovation here. It might be premature to niche down! And yet I could also see this being exactly the right time to build a lot of institutional knowledge around “how to build the best harness for writing Python code with an LLM”, rather than all code.
About the Author
I’m Justin Duke — a software engineer, writer, and founder. I currently work as the CEO of Buttondown, the best way to start and grow your newsletter, and as a partner at Third South Capital.
Colophon
You can view a markdown version of this post here.