exe.dev

@davidcrawshaw @ssh_exe_dev Took Shelley around for my long weekend and wow it's a great harness. Having a native iOS app was cherry on top! Kudos.

I really like @ssh_exe_dev

While we validate a kernel change for januscape (CVE-2026-53359), we have disable nested KVM across our fleet for individuals and teams without dedicated hardware. We will restore nested KVM in a rolling release as soon as we can. Apologies for the inconvenience.

細部にこだわる必要のない自分用のツールなら出先のスマホからでもサクッと作れる。いい時代だ。 Thank you @ssh_exe_dev

at opencode we've started developing with agents on exe.dev boxes, which automatically serves http. I can just ask agents to write HTML files and see them they could be static artifacts, live thoughts, or little apps I underestimated how cool and fun this is

now, in 2026, I spin up dedicated dev VMs (over top of @ssh_exe_dev) for each distinct piece of work, forked off of a base VM I've already configured to my liking. exact amount of RAM/disk/compute needed for that work, scaled up/down as needed. isolated space for my agents.

@ssh_exe_dev works perfectly for creating a site with QIP! I’m impressed that it was able to just read my docs and create a fork and run Zig! I can then use natural language to change the site content and theme.

@brianmichel @davidcrawshaw As you are the very first person to ask... how about we send you one? DM and we'll figure it out.

@davidcrawshaw Damn, how do I get one? Love the orange high viz

@davidcrawshaw I love @ssh_exe_dev

My agents now have many computers. I wrote about my experience with @ssh_exe_dev and its iOS app. // exe.dev: Give Your Agents Many Computers – YPWU ypwu.net/2026/06/24/exe…

I finally found a good way to run agents unmonitored. A shocking number of people run `claude code` or similar without a sandbox. They either `--dangerously-skip-permissions` or tediously approve each request, which quickly becomes mindless and almost as bad as YOLOing. This opens you up to mistakes (e.g., `rm -rf`), prompt injection, and supply chain attacks. The right way is to run agents in their own VMs, but that's annoying to set up correctly, so most people don't do it and just hope nothing goes wrong. Enter: @ssh_exe_dev. You can have a VM running in 30 seconds. And it makes all the stuff that's normally hard or annoying (secure web access, auth, API secrets, sending/receiving email, etc) super easy. (This is not an ad! I'm just a paying customer, who came across @ssh_exe_dev because it was founded by @davidcrawshaw , a @RecurseCenter alum.)

Moved all servers and services to exe.dev. Having access to LLM on server via backend panel has been great. Cut down server setup time by half. Great job @ssh_exe_dev 🎉

Build an app on your phone. Create a commit in a dev VM. Run a data analysis connected with an integration to your data lake. Do developer things on the go. Read more here: blog.exe.dev/ios

Announcing the exe․dev iOS app

@jamiepinheiro @BachelderDan You can use it directly from codex on your VMs! We thought about configuring this automatically for you but didn't want to break anyone's custom configurations. Here's how to do it: #use-with-codex" target="_blank" rel="nofollow noopener">exe.dev/docs/integrati…

@ssh_exe_dev @BachelderDan I guess you’d need to use Shelly then, right?

I used @ssh_exe_dev for this project One of my happy-surprises was how nice it is being able to run codex on the same “computer” that the production server is running on - I get an agent that can answer data questions for ~free No funky cloud abstraction MCPS/auth to configure

@BachelderDan @jamiepinheiro Have you tried our ChatGPT integration? exe.dev/integrations

@jamiepinheiro @ssh_exe_dev No, they don't seem tied to the system in any way so you can just push them around. As long as you're using codex or another harness that understands it needs to use the refresh token periodically or are pushing the auth.json frequently enough it will just work.

New day, new integration: bring your ChatGPT subscription to exe. Power Shelley and the LLM gateway with it for all your VMs.

Under the hood, this is about careful management of the global rtnl_mutex in the host linux kernel. Rapid creation of new namespaces on busy machines requires carefully managing the contention of this lock. (We have some more work in the oven in this space!)

Before this, business customers had access to compute nodes that could predictably boot VMs in under half a second. Now everyone gets sub-second starts.

Yesterday we deployed a network namespace scaling project we have been cooking up. This is designed to keep compute nodes behaving predictably under load. As a bonus, it means even multi-GB systemd-based images boot in under a second, on any of our machines.