AI Coding Agents Move Into Isolated Workspaces
AI coding agents are moving off developer laptops and into the enterprise. That shift raises hard questions for platform teams. Where do agents run? What can they reach? Who controls the data, and who pays for the tokens? Many organizations are also rethinking whether AI workloads belong in a hyperscaler or on infrastructure they control.
Coder Developer Relations Engineer Dave Ahr examines that challenge with Alan Shimel at KubeCon 2026 in Salt Lake City. Ahr came to Coder after years at HashiCorp and VMware.
From Remote Workspaces to Agents
Ahr explains that Coder began as a way to build workspaces developers could connect to and shut down when done. Adoption was slow at first, since developers rarely want to give up their laptops. Then large customers asked Coder to put agents inside those workspaces.
Coder is self-hosted rather than SaaS. That appeals to regulated industries and government agencies that need to control their own data. Ahr sees the same pull toward local control in how companies think about IP, AI and token spend.
Hardware Isolation on AWS
AWS released Nitro Sandboxes during KubeCon week. These hardware-isolated microVMs give each workload its own boundary. Ahr spent the week building a joint demo with AWS on how Coder fits.
Coder builds its workspaces from Terraform templates. That lets teams orchestrate the full environment, including network policies and pod identity. Ahr says the result is a workspace that can be locked down as tightly as needed. It works for human users, automated processes and agents alike.
Platform Teams Own the Bubble
Ahr argues the problem sits with platform engineering, not security or developers. He sees two shifts underway. First, organizations need isolated workspaces that put AI in a bubble. Second, networking is becoming a priority again.
Teams now have to consider where agents reach and how large the blast radius could be. Data loss can happen without any malicious intent. The goal is to contain AI coding agents without slowing adoption. Ahr also points to non-developers building apps of their own, a group many companies struggle to support.
Transcript
Hey everyone, welcome back here. We're going to wrap up our day one coverage of KubeCon with our next guest. It's been a tremendous day, but I got to tell you, the next two days are packed with people as well.
So catch what you can today. Be sure to join us tomorrow. I think we start around, well, we'll start around noon Eastern Time with the gang live here, and we'll go from there.
My next guest is Dave Ahr. Uh-huh. A-R-G- A-H-R.
A-H- A-H-R. And Dave is a DevRel engineer- Correct ... with Coder.
Yes, sir. And he's joining us today here. Dave, welcome to Techstrong.
Thanks for being here. Thanks for having me. Appreciate it.
This is good. My pleasure. So Dave, DevRel engineer.
Yeah. We've gotten a lot of DevRel people here. Sure.
Tell us your story. Well, before Coder, I was at HashiCorp for a couple of years. Before that, I was at VMware for about eight years.
I spent two years in sales engineering, solutions engineering. Sure. I spent about six years in consultancy.
So I did everything that VMware did, I did it. I had to go put it in there. And before that, I was a full stack developer.
Really? We got acquired by VMware. That's how I came to be at VMware.
" And that's how I got here. So I started as an SE, and then I moved to technical product marketing manager, and now I moved into this DevRel role. Love it.
So I've got two hats. You know what that's like, man, right? Let's talk about Coder a little bit.
Yeah. Tell us the Coder story. It's an interesting story because Coder got started in 2017 from a couple of young kids that were building some AppSec, doing some gaming, and they had decided, "You know what?
" Right? "We have to load all this stuff on there. " And so they built out Coder for that reason, and it was okay.
It trudged along a little bit because let's face it, developers don't want to give up their laptops, right? They're not compelled to do that. But about two years ago, we had some very large customers that came to us and said, "Hey, you already figured out infrastructure.
" ... " So we did, and sales just hockey sticked. Love it.
Especially when we opened up the office in London with GDPRs and all that, they had to have a self-hosted solution because remember, we're not SaaS, we're a self-hosted- Right ... managed solution. So we're very attractive to highly regulated industries, government agencies, because they host it, they control the data.
And so- I think you got another thing, good tailwind. Yeah. Not to say headwind.
Yeah. Tailwind that's helping you too is, I think with this whole AI thing, people are reevaluating, "Do I want to host it in a cloud or a hypervisor, a hyperscaler? Do I want to open the closet, the server closet again?
Do I want to host it locally? Do I want to control my data? I want to control my IP.
I want to control my AI. " Exactly. I want it local.
That's what we do. It's funny, the pendulum swings. Oh, it's hilarious.
Because I was in sales first, and that was a hard sell. But now it's becoming really easy. " Now you got the tech.
We got to have it, right. 100%. I think that's something that's happening for sure.
Yeah. So Coder, for a relatively smaller company, has a huge footprint here at KubeCon. Everything from sponsoring the Wi-Fi network.
Oh, yeah. Your booth's right there. Actually, from where they're looking, that's the Coder booth right over- That's my homie right there ...
you can see it on our screen. Those are my peeps. Well, it's probably blurry because they got us focused on you and I, but they can see it's back there, right?
Yeah. That being said, what's going on at KubeCon for you guys? Well, today has been crazy.
First, we had the tech field day this morning. Yes. And that was funny because we found out about it on Friday, and so we're scrambling.
We don't even know what it is. We haven't even been a part of that. And so, my colleague Josh and I, we went up there, and we didn't know what to expect.
We didn't know if they were going to hang us in tortures. We didn't know. No, they don't.
No, great group. Super great group. And so we got through that, and then AWS just released their Nitro Sandboxes.
Yes. " And so I did it with Kata, and I turned it all in, and Rodrigo, my new friend and colleague at AWS, he said, "Hey, what about these Nitro Sandboxes? " So they dropped it on Monday.
So I'm on the plane, and I'm working on this thing on the plane for three hours. Then I'm at the hotel. I was supposed to be at the EKS Roundtable, and I said- Jesus Christ ...
"I can't make it, Rodrigo, I'm sorry. " And so I just spent a lot of time at the AWS booth because these Nitro Sandboxes fit perfectly with what we do. These are hardware isolated- Yeah ...
small microVMs that we can drop a workspace in- Yeah ... with an agent. And then what we really do, and it's funny because I was thinking about this last week, and then I come over here, and I start seeing this, what they call the substrate.
We're the substrate orchestrator, and it's like, we've been doing that for years. Yeah. And so that's really what it was, was now, with our Terraform, we use Terraform for our templates.
That's what we build our workspaces off of. Got it. And Terraform is so widely known, it's widely used.
It has providers- It's the standard ... for everything. Yep.
So now we can orchestrate that entire substrate on AWS to include network policies. We can include pod identity, all these things to where you can lock that thing down just as tight as you want to. And so they were excited because we're just...
AWS doesn't care how you consume, just we want you to consume. And we're providing that avenue to say, "We got it. " Right?
For not only just users, but for processes, for AI agents, we do Claude. And so it's great. And we're self-hosted, so you can run it in your VPC, right?
I love it, man. That's good stuff. It's fun.
It's been a fun day. I'm tired. I'm tired as hell, but- You and me both.
It's been a long... We got two more days of this. I know.
Two more days. I know. I was talking to someone earlier in one of the interviews, if you didn't have this cloud native stack, we wouldn't have the AI that we have today.
I agree. And so, I don't know whether it's karma or just things happen for a reason, but these things just keep building. And you've probably seen it over the course of your career, right?
Sure. Your VMware days to Google to... Things just keep building.
Yeah. Building and building. And what's the old saying?
The days are long, but the years are short. Yeah. Right?
" Oh, especially now. You look over months, you look over weeks now. Weeks now.
It's like next week something else is coming out I got to go catch up with. Exactly. It's crazy.
It is. But for me, I use AI for everything. Me too.
So for me, it's cut down my learning curve 70%. Agreed. These Nitro Sandboxes, I didn't know anything about them until they dropped yesterday, and I was like, "Look, talk to me like I'm a child.
" Explain it. " So it's like, I don't have to go wade through web pages and pages and pages of docs. Just give me the TLDR and I'm done.
You know what? I was talking to someone yesterday at this breakfast we did. " Yeah.
So everyone's going to have their own models. Crazy. Well, and we're seeing that.
So we're seeing that now. So what we're seeing from our customers is they're starting to look at how do we... Well, there's a couple of things, right?
First of all, what I call a citizen developer. And a lot of people wince at me when I say that, but we're talking about the developers that are non-developers. And I've heard some of the largest companies say, "We don't know what to do with our non-developers.
" And so we're seeing that side of it, which Coder fits in perfectly because we're doing it for our own non-developers. We're eating our own dog food for that, right? Everyone is.
Yeah. I mean, you don't want to slow them down. I love it.
I call it AI for all. Let's get it, right? It's a great equalizer and it's the greatest learning tool we've seen, man.
Oh, it's phenomenal. It's phenomenal. Hey, for people who want to get more information on Coder, what do they do?
com. com- And it's a great way to learn ... com/learn.
We have curriculum out there. It's not just about selling Coder. We're actually putting content out there that shows our AI journey and how we're using Coder and what we're learning along the way.
It's fantastic. We have a Coder community, and we have a Discord channel. You can join that.
It's very active. I just had a couple of young guys at All Things Open a few weeks ago. They were like, "Dude, we're on your Discord channel.
" And we're like, "Yeah, we love it. We're excited. " So a lot of companies I know are using Discord.
It's quite a great community. Well, what they told me was there's a lot of companies that do Discord just because they want to be able to advertise. Cool.
But we're actually doing community first. We have a whole community team- Oh, no, I know ... our DevRel team, because we're open source first.
Everything's open source for us. So we really hit that hard. And I was just walking home.
I walked to the hotel from a restaurant last night. " Oh, no. " And I just got- You got the jelly bean?
I just got one of our really nice camouflage hats that I've been trying to get my hands on, and I couldn't help it. I had to give it to the kid. You gave it to a little kid.
I had to. " Tell him I said to give you another one. Oh, they will.
They always do. Thank you. Thank you for coming on here.
We appreciate it. Oh, absolutely. It was my pleasure.
This has been a joy. I wanted to ask you just before we end, I got one more kind of question I got to go over with you. Sure.
Go for it. So AI coding agents are moving off developer laptops and into the enterprise. Platform teams, platform engineering, now on the hook.
Where do agents run? Where can they reach? And who's paying?
Coder development helps there, right? Oh, Coder. Yeah, absolutely.
So really, the way we talk about our solution is it sits with the platform engineer. It doesn't sit with security. It doesn't sit with the developer.
I'm sorry, platform engineers, you're just going to have to accept it. That's your job. Yep.
That's what you do. What we're seeing is a couple of different ways things are starting to really kind of shift, right? Number one, these isolated workspaces.
We got to get AI in a bubble. And number two, you're seeing this shift back to network first. Yeah.
Right? Networking kind of took a backseat, but now we're realizing networking is essential, right? Now we've got to go back and really think about where these agents are reaching.
Yeah. Right? What kind of blast radius are we looking at here?
We don't want to lose our data. We don't want to exfiltrate our data. And it's not even to be malicious.
It's just happening, right? So that's kind of where we've really seen a hockey stick effect, is that it's like we weren't considered before, but now it's becoming a necessity. We've got to get these things isolated, but we don't want to slow down AI adoption.
No. You want to know the truth? I don't know if you can.
No, you can't. This thing is- No, you can't ... it's kind of its own perpetual motion machine.
Oh, I tell everybody, I was like, "Let's just let him take over. " I'm cool with it. They're all good.
I'm really cool with that. At this stage of the game, I'll take the seat, man. Thank you.
Hey, that's going to wrap up day one here at KubeCon. I hope you've enjoyed our Techstrong TV coverage. We'll be back, as I said, tomorrow, noon Eastern Time with Techstrong Gang and a full day after that.
Until then, this is Alan Shimel. Thanks for watching.