So this product is a thin wrapper over a remote virtual machine, a new name for remote virtual machines, and an exhortation to not look inside or reason about remote virtual machines?
I feel like up until recently this would been pitched as “the missing UX/DX for cloud compute”, or something. But instead now it’s marketed to this new trend where one should aggressively not understand the tower of abstractions one is standing on. Odd.
I haven't paid super close attention to Orbs, but I also build an agent runtime + orchestration layer on top of remote VMs (https://github.com/gofixpoint/amika). I do think it's smart marketing to name the concept. And for coding agents, there's a lot of annoying fiddly stuff you have to do to get the experience to work well:
1. auto-refresh base VM snapshots with latest delta of git commit changes
2. extend the VM lifetime when it receives an interaction (chat, HTTP request, etc.) and sleep it when inactive
3. oauth refresh of MCP servers, coding agent subscriptions, etc.
4. support inter-VM and communication for agents and exec commands
5. keeping track of agent work across VMs
That said, I think it's important to not hide the VM and computer from the agents. The user should have total control to put 1 or more agents on a VM, to make VMs ephemeral or permanent, to manage the network access of VMs, etc. It's basically a networking + VM scheduling problem, for which there's a ton of prior art since the 90s (or earlier).
A few examples:
Many remote agents products have a "one agent per VM" restriction, but it's better to have N agents across M VMs. For example, I have agents work on a singular VM in parallel worktrees and then stack the worktrees at the end on that VM for the final change.
Just give me OpenSSH with normal SSH key management. Lets you manage VM access the same way we've done it since the 90s.
Connect VMs in a network topology: we build our VM/agent scheduling + remote control product inside itself, so we often test interactions between user, control plane, and sandbox(es). If you can directly control VM creation and lifecycle, it's easy to spin this structure up and let agents work across it.
I should be able to use whichever coding agents I want, connected to the VM. Whether that's OpenCode or Pi Web in the browser, or loading the VM into Cursor or Codex app, or using the TUI or sending messages to agent(s) via API.
Hey! Author of the post here. Sounds like I failed my mission with the post then :)
Yes, like many things, Orbs are "wrappers" around VMs, but the point is that some wrappers enable new ways to hold and use and reuse the thing they wrap, and even give you a different perspective on it.
That's what happened with our team once we had Orbs working. We knew beforehand all the things they're made of (fast to start sandboxes, scale to zero, live streaming updates on all clients, one machine per conversation, etc) but what really surprised us was how it changed our workflow by removing friction we didn't even know was there. The friction of creating new checkouts and worktrees and managing local resources -- it sounds trite, of course, but man, once you can stop thinking about it, it's so much better.
So, yes, maybe they're wrappers but hey, some people say the tortilla is what makes it a burrito, you know?
I get it. You need to market. But it helps us all if you do your homework on terminology and prior art. For example, I assume you aren't referring to an object request broker, a music streaming service, an older open router system for blocking spam, etc.
> 1. For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software
Obviously I'm not claiming that this will be anywhere near as successful as the iPod or Dropbox. It might not succeed at all. But I would have thought we'd be past this sort of dismissal by now.
I think very soon big LLM providers (OpenAI et. al.) will provide a service like this. With the LLMs becoming faster, the agentic task completion bottleneck will soon move to the tool calls (both execution time and round-trip latency), so it makes sense to have a server host the project very close to the actual LLM doing the inference in order to minimize latency.
Also, they can literally train the LLM inside the VM they create so it becomes an expert in whatever tools are there. And since they control the entire sandbox, less permission issues are likely.
I think they will become extremely vertical integrated sooner or later because that will be the most efficient and effective way to do this. I'm guessing anyone building this kind of thing now has only one hope: acquisition.
Anthropic has Claude Code Web -- sorry, looks like the official name is "Claude Code on the web" -- which I think is basically this? I've been using it for a while and I really like it.
Nothing to do with latency, I think, it's all about convenience and safety. I don't have to worry about running an agent on the computer that knows all my admin API tokens and SSH keys, and the agent isn't interrupted if I close my laptop.
The one thing that might be useful is the ability to run on a beefier machine when you need it, e.g. make sure it has direct access to a GPU. But then you're back to thinking about the host VM, not fully abstracting away from it.
They've basically come up with what exe.dev has been doing for awhile now?
Except, exe.dev is more about, you have a VM and can do LLM work on it. Orbs seems to be minimizing the VM machinery more and making it about the "agentic experience."
It sounds like great engineering, but everyone in the coding agent space is building the same thing. And big labs like Anthropic and OpenAI have more money to burn. Cloud agents that run in VMs are being offered from everyone from VPS and infra companies to the frontier labs.
We all seem to be building this kind of Orbs or whatever you want to call a harness with some user interface.
I am at version 4 and I have a “open in vscode” button on each task/orb/instance just in case the UI is not enough and you want to corner the agent really close
To give another perspective from the generally negative comments here, I've found Orbs to be just fantastic! The "it's just a VM, everybody has that" argument misses the point, the real value lies in just how easy it is to use and the surrounding functionality. Portals, multiplayer, local machine syncing, Puck (something like a meta assistant), referencing threads/Orbs from each other, etc. all just work as you'd hope and expect.
It's the best $20/month I'm spending on any dev tools, the Amp coding agent in itself is worth that much (I actually came for that and stayed for the orbs). I can only encourage anyone to give it a try and maybe you'll be pleasantly surprised :)
With ParaSpace I can use a real terminal and my normal tooling to develop locally on fast booting isolated containers. I even get to bring my own dotfiles.
For you not to waste your precious time - "orb" is just a virtual machine. And the thing requires a phone number, which it tells you at the END of the registration.
I have new marketing jingle for that company: "market your AI slop with non-AI slop".
I am on the Amp team. I just turned off SMS verification. Not needed anymore. Was required for ~5% of signups for anti-abuse, but I think we're good without it now.
I feel like up until recently this would been pitched as “the missing UX/DX for cloud compute”, or something. But instead now it’s marketed to this new trend where one should aggressively not understand the tower of abstractions one is standing on. Odd.
A few examples:
Many remote agents products have a "one agent per VM" restriction, but it's better to have N agents across M VMs. For example, I have agents work on a singular VM in parallel worktrees and then stack the worktrees at the end on that VM for the final change.
Just give me OpenSSH with normal SSH key management. Lets you manage VM access the same way we've done it since the 90s.
Connect VMs in a network topology: we build our VM/agent scheduling + remote control product inside itself, so we often test interactions between user, control plane, and sandbox(es). If you can directly control VM creation and lifecycle, it's easy to spin this structure up and let agents work across it.
I should be able to use whichever coding agents I want, connected to the VM. Whether that's OpenCode or Pi Web in the browser, or loading the VM into Cursor or Codex app, or using the TUI or sending messages to agent(s) via API.
Yes, like many things, Orbs are "wrappers" around VMs, but the point is that some wrappers enable new ways to hold and use and reuse the thing they wrap, and even give you a different perspective on it.
That's what happened with our team once we had Orbs working. We knew beforehand all the things they're made of (fast to start sandboxes, scale to zero, live streaming updates on all clients, one machine per conversation, etc) but what really surprised us was how it changed our workflow by removing friction we didn't even know was there. The friction of creating new checkouts and worktrees and managing local resources -- it sounds trite, of course, but man, once you can stop thinking about it, it's so much better.
So, yes, maybe they're wrappers but hey, some people say the tortilla is what makes it a burrito, you know?
Also most proprietary software, also most web-based software, also most APIs... also what my plumber does when I call him over!
(Also most hardware?)
[1] https://orbstack.dev/
> No wireless. Less space than a nomad. Lame.
https://news.ycombinator.com/item?id=8863
> I have a few qualms with this app:
> 1. For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software
Obviously I'm not claiming that this will be anywhere near as successful as the iPod or Dropbox. It might not succeed at all. But I would have thought we'd be past this sort of dismissal by now.
I think they will become extremely vertical integrated sooner or later because that will be the most efficient and effective way to do this. I'm guessing anyone building this kind of thing now has only one hope: acquisition.
https://chatgpt.com/codex/cloud
From my experience, Orbs have been more consistent, along with Cursor's agents
Nothing to do with latency, I think, it's all about convenience and safety. I don't have to worry about running an agent on the computer that knows all my admin API tokens and SSH keys, and the agent isn't interrupted if I close my laptop.
The one thing that might be useful is the ability to run on a beefier machine when you need it, e.g. make sure it has direct access to a GPU. But then you're back to thinking about the host VM, not fully abstracting away from it.
Except, exe.dev is more about, you have a VM and can do LLM work on it. Orbs seems to be minimizing the VM machinery more and making it about the "agentic experience."
I am at version 4 and I have a “open in vscode” button on each task/orb/instance just in case the UI is not enough and you want to corner the agent really close
It's the best $20/month I'm spending on any dev tools, the Amp coding agent in itself is worth that much (I actually came for that and stayed for the orbs). I can only encourage anyone to give it a try and maybe you'll be pleasantly surprised :)
This is why I made ParaSpace (https://paraspace.dev/)
With ParaSpace I can use a real terminal and my normal tooling to develop locally on fast booting isolated containers. I even get to bring my own dotfiles.
- "Cloud Agents" on cursor
- "Cloud Coding Agents" on Hoplite.sh
- "Sessions" in Devin cloud (they don't make a big deal of them being remote)
Amp is just trying to pitch an increasingly-commodity experience as a special sauce.
>requires a phone number…at the END of the registration
I do not appreciate that at all. Phone number, understandable, but it better be all the way up front.
Also Grok Bot is incidentally just a wrapper for cursor cloud sessions/vm's
They’re just names to be used within the context of their parent ecosystem. If it were a general purpose tool, I’d tend to agree.