Anthropic Accidentally Leaks Claude Code | Tech Field Day News Rundown: April 15, 2026

ServiceNow clicks Accept. Intruder alerts on your containers. Claude rules your desktop.

Linux networking gets better? Nutanix is going for bare metal on Kubernetes. Microsoft wants eternal Copilot.

" It is April 15th, so happy National Tax Day to all who celebrate, whether you want to or not. It's also Jackie Robinson Day. It's National Remember the Titanic Day.

I do remember. There was something about a boat and an iceberg, but that's not important right now. What is important is that I'm back in the driver's seat for the rundown, and of course, I am joined by my fearless co-host, Mr.

Alistair Cook. Al, it's good to see you again. It's great to be here on World Art Day and make more art.

Don't let the AIs do it for you. Do it yourself. It's much more fun that way.

It absolutely is. Whether or not it's macaroni or paint or anything in between, you should make a beautiful picture, just like the picture that we're going to paint you. Some of them are going to be more beautiful than others, but a lot of them are going to involve AI.

And the first story, strangely enough, is all about that because AI is only as powerful as the data behind it, and the new partnership between Qlik and ServiceNow is designed to improve both. They are working together to bring trusted, governed data into AI-powered workflows. By combining analytics, data integration, and workflow automation, organizations can connect data across systems like ERP and CRM, which will help them gain better insights and make faster, more confident decisions.

That helps turn data into real actions. Al, this announcement came out as part of Qlik Connect, and I know Stephen Foskett is down there enjoying the warm Florida sunshine right now, but what do you think about this deal between ServiceNow and Qlik? Is this going to help ServiceNow customers really be able to provide more value?

Yeah, I think clearly this is valuable for the overlap of Qlik customers and ServiceNow customers. And I got to think that's actually a fairly large overlap because both organizations have products that appeal to fairly large organizations that want to have good control of their internal processes from the ServiceNow side, but also their internal data from the Qlik side. What's going on here is that Qlik has now got a collector that can go through and collect metadata out of your ServiceNow environment and use that to drive the AI tools that Qlik has been building for a while.

So having some context of the relatively static data, and relatively static is probably not the right term for it, for the insights that are gained from the data that's generated by your human teams probably. This is what the Qlik side is going to be looking at. And then the things that are happening inside your workflow oriented, your interactions, both internal and external, that are happening through ServiceNow.

So bringing those two things together gives you a lot more context of what's happening within the organization. What's the wider context of an interaction maybe between a service agent and a particular customer or that customer and your website, and bringing that holistic view through, and then taking whatever the next required action is. And this is the important part.

The AI here is about what's the next action that we need to take, and particularly, what's the next action that might bring greater satisfaction for that end user or minimize our cost for managing whatever issue is being brought up. So this integration seems to be really cool. I liked working with Qlik's AI tools when I took a look at them, I think nearly a year ago now, as we were working a project with Qlik.

And ServiceNow, obviously really well respected for looking after a lot of the sort of interactions with humans, interactions with supply chains. So bringing these things together is a really good thing. Remains to be seen just how far you can go with the Qlik AI platform.

So one of the things I saw there was, certainly a year ago when I was hands-on with it, it did good things, but it did a relatively limited range of things. I imagine by now the range of things that that Qlik AI can do, the ability to run more complex workflows through that Qlik AI has moved forward, and we'll be watching the reports, particularly the live blogs that are coming back from Qlik Connect with Stephen and a few of our delegates who are attending the conference. So do watch for those on the Tech Field Day website, as well as watching for what's being announced by the companies down there.

Intruder has expanded its cloud security platform with container image scanning, helping teams find vulnerabilities before apps go live. The tool integrates with major cloud registries like AWS, Azure, and Google Cloud and works without agents, making it easier for both security and development teams to use. Built for midsize organizations, it also uses AI to prioritize risks, helping teams manage growing cloud environments and improve software supply chain security, which Tom, we've seen a bunch of software supply chain attacks over time.

Is this going to significantly help these midsize organizations who often struggle with managing security? Yeah, I think it will. One of the things that we've kind of gotten into our heads as a society is this idea that if something is in a trusted store somewhere, once I click the button to deploy it, it should be good to go, right?

Because it's in a trusted store. And that mentality of being able to deploy applications quickly has filtered into the cloud world via containers. So what people are wondering is, well, I can just click to install this container, and there won't be any problems, right?

" That wrecks this just-in-time mentality that containers have. I want to be able to deploy it. I want to be able to do something with it, tear it back down and be done with it, and if it's going to take hours orIn some cases, days to do a proper vulnerability scan on this containerized application, the novelty and utility of it just goes out the window, right?

So what do I do? Well, that's where this tool from Intruder comes in, is it can quickly scan as the app is being deployed and verify that it is a good reference image or it hasn't been infected with malware. And one of the key reasons why is what you referenced in that last line, supply chains.

We have had a lot of supply chain vulnerabilities, and one of the most recent ones that was referenced by Mike Vizard in the linked article was from Trivy's vulnerability scanner. Because one of the things we've learned over the years is that if you can knock out the vulnerability scanner, you can do pretty much whatever you want. So what if I were to, I don't know, compromise Trivy and allow it to pass through some malware that's a rat or some kind of other device that allows me to take control of that container and then try to break out and get places?

I want to make sure that not only are my tools doing their job, but they're not being compromised. Supply chains are one of the next frontiers that we have to worry about, because a supply chain hack, while a little bit more complicated than just kicking in the front door, gives people a lot of persistence and a lot of capability to continue to reinfect the system over and over again. So if you're looking at doing something in a containerized way and you don't have an IT team the size of Rhode Island to help you, this is exactly what you need to be checking out.

Our friends over at Anthropic have introduced a new feature that lets its Claude AI control your Mac or Windows computer, clicking, typing, opening apps, and completing tasks for you. It works by using integrations first, then browser control, and finally, full desktop access when needed. Paired with a tool called Dispatch, users can assign tasks remotely and have them completed automatically.

While powerful, the feature is still in preview and raises a few security concerns, especially around those pesky unauthorized actions and some sensitive data. Al, are you ready for Claude to just take over and do whatever it wants? I, for one, welcome our AI-powered overlords.

Only maybe I don't, and I'm a little concerned here. So first off, Claude Code and Claude Coworker, two very popular tools. I know talking to local developers here in New Zealand, Claude Code is very popular as a tool for speeding up software development.

One of the fun elements here is that transition from being, here's a command prompt where you basically interact with the Claude Code chatbot to the actual Claude Code element can take over the computer and operate for you while you're not sitting in front of that screen. That's an interesting change, but by itself, it probably wouldn't have come about. The enabling technology here is Dispatch, and this is inspired by, I think, inspired by what we saw with OpenClaw, where you interact with OpenClaw through a remote window, which is essentially maybe an integration into Telegram or into Slack.

Well, the same thing occurs here. That's what Dispatch is here with Claude Code and Claude Cowork. It's the ability to send directions of what you want to achieve through something that isn't the computer that has Claude Code installed on it.

Claude Code or Cowork can then be taking over your computer while you're away. Obviously, you can't put the computer to sleep while this is going on. It's got to still be running, but actually interact with desktop applications, not just interact with command lines.

I guess the switch here is not so huge. If it could interact with your entire computer via command lines, there's command line-based tools that could emulate a lot of the clicky around the screen activity that you'd want to do, or at least emulate the results. The actual interaction with the screen comes from an acquisition.

The Anthropic picked up a company called Verset AI about a month ago, and this is what Verset did. The end of February, I believe, was the announcement for this acquisition. And Verset was aimed at allowing an AI to interact with desktop applications and has been very rapidly integrated into the Claude Code and Claude Cowork.

As you say, there's some concerns around this, concerns around the fact that the very websites that Claude Code might now be interacting with could be an attack vector. There could be embedded prompt hijacking, prompt escape kind of attacks going on in here. And although Anthropic says, "Yes, we're scanning for that," we're never quite so sure how well they're scanning for it.

So that'll definitely be an ongoing thing. I think we will see quite a lot of back and forth about how much you trust your AI agent to actually interact with a much wider context of interacting with a website through a browser, and also then driving things like your enterprise application, driving your ERP application's user interface based on prompts that are being sent remotely. Somewhat cynically, I can see this as being a great thing for people who want to outsource their day job.

Just leave your computer running, connect Claude Code up to it, and get some external person to go and send those commands, and it makes the whole thing look like you're doing a huge amount of work while you're actually sitting there, maybe with your game console in front of you, playing games instead of doing your job. Is that what's going to happen here? Don't know.

We'll see this over time. We will see this develop over time and see whether there is actually utility in having an AI agent interacting with your screen, with the applications that are installed in your desktop. I know from thinking about it when I've been doing remote support with people, that something else taking over your screen feels very unnatural for a lot of people.

There'll be a chunk of resistance to that. Linux 7. I can't believe it.

9. Linux 7 focuses on steady improvements, especially in networking. It adds smarter congestion control and boosts data throughput and improves traffic management for better performance under heavy load.

It's kind of important as we're building big AI applications using Linux as some of the foundation. This update also enhances IPv6 and adds official Rust support for safer code. With expanded hardware compatibility and making it a solid upgrade for cloud, AI, and high-performance systems.

Tom, are you upgrading everything you own to Linux version 7 to get this network performance? Oh, sure. Why not?

It's lucky, right? Oh, wait, it's odd numbered. Is it like the Star Trek rule?

Are we avoiding the odd numbered ones? For networking, I don't think we should be avoiding this because there's a lot of things that have been enhanced in the Linux 7 kernel that I think are kind of important. And when you look at the things that they were specifically picking on, this makes it feel like they're aiming for a market where Linux is kind of running as a switch operating system.

So one of the things that they did was enabled accurate explicit congestion notification or A little c little c ECN. That's something that's very important for very congested networks like the backend network that runs between your GPUs. It allows you to kind of modify how the traffic is being displayed across everything, so very important.

Also, for our good friend Mr. Stephen Foskett, one thing that's important is the common applications kept enhanced scheduler, otherwise known as CAKE. CAKE is used as a QOS tool, and one of the things they do is now they have multiple queue applications that you can do.

That's kind of important because a lot of the things that are running this are on single CPU devices, and they're getting overwhelmed. So by multi-queuing it, what you're getting effectively is enhanced performance for those little devices. Look at all of the things that are being put into Linux, and what you're basically telling the world is we want whatever comes out next in the network operating system land to be running on Linux with all these enhanced features so that it provides stability, but more importantly for a lot of these companies, monitoring capability.

Think about all the work that's been done trying to turn the operating system of your network into some kind of a sender of telemetry. Facebook, that was one of the reasons why they built their own operating system, FBOSS, was because it was based on Linux and they knew how to poll Linux systems. And so this is the way that they're doing that.

I think that this is just the natural progression of what Linux is offering, right? We don't use Linux on the desktop. I'm sorry to the three people who listen to this that do.

Most people use it in servers and other headless devices. So enhancing those capabilities and building them into the kernel is a much better offering in the long run for people. So I can't wait to see what the companies who are embracing Linux as a network operating system are going to be able to come up with in the near future.

Our friends over at Nutanix are introducing NKP Metal, which is not the name of a band, but it is a new way to run Kubernetes directly on physical servers without the need to virtualize it, which is improving performance for those hungry applications like AI, edge, and GPU-heavy workloads. It keeps the simplicity of cloud-like management while allowing teams to run containers and virtual machines together using a unified system. The goal is to make bare metal Kubernetes a little bit easier to deploy and manage, especially for organizations that need high performance and no complexity.

And we know how difficult those unicorns can be to manage. Al, do you think Nutanix will be able to knock it out of the park with a bare metal Kubernetes offering? Well, I think there'll be very specific customers who find this incredibly valuable.

As you said, big challenge with bare metal Kubernetes is management and deployment. Managing those huge numbers of scale-out servers that you have in a bare metal Kubernetes deployment. Most Kubernetes clusters running in enterprise organizations will be running inside virtual machines, and that's what Nutanix has seen, as well as Microsoft and other hypervisor vendors, VMware as well, have seen Kubernetes running inside virtual machines is very popular inside enterprise organizations.

And for me, that's because virtual machines provide greater isolation than container-to-container isolation. And so running containers and virtual machines suits where you have a relatively small number of containers for each of your applications, but you have a lot of applications. And so you want really strong isolation between the separate applications, but you want efficiency to run the containers for that application.

Now, this is very different from how the likes of Google run containers, because they will have millions of containers for a single application or for a single security domain, and won't need that virtualization isolation because they'll have a large enough population of containers that physical machines are dedicated to a particular security domain or particular application. And that's really where I see this being really beneficial, is where organizations are finding that they're dedicating entire hypervisor hosts, multiple hypervisor hosts, just to run one set of applications, one security domain. And then maybe they're not getting any benefit of running containers inside that virtualized operating system.

And so in those use cases where there's no need for the security isolation you get from virtual machines, you're more efficient running those containers directly on bare metal. And so as an option to be able to run a Kubernetes cluster that runs entirely with bare metal-based hosts rather than having VM-based hosts in that cluster, absolutely this will be great, particularly as there's integration between the applications that maybe don't need that large a population of containers that might still run in virtual machines. You've still got the same management interface for managing those groups of containers, and particularly for managing things like the networking around them and managing the persistent storage that's attached to those containers.

Having consistent tooling around that is absolutely vital, and I think Nutanix is going to be very successful with this for the customers that are running large numbers of containers alongside their other applications. And often what we do see is enterprise companies have virtual machines. They have legacy, what I often refer to as boat anchor applications, where there's a huge amount of business value in the virtual machinesBut all of the innovation is happening in the containers, and those growing populations of containers will be what's driving this move to bare metal Kubernetes.

I think it's a very cool thing. It's not for everyone, but it is great to have that option. Microsoft is working on new always-on AI agents.

Of course, Copilot branded. These agents can complete tasks on their own without waiting for a human, and they're really inspired by OpenClaw, and these agents aim to manage things like your emails and your schedules and run some workflows automatically across your Microsoft 365 estate. Unlike OpenClaw, of course, Microsoft is focusing on making these systems secure for enterprise use, and the goal is to turn Copilot into a 24/7 digital coworker that saves time and makes AI subscriptions, your Copilot subscription you're paying for, more valuable.

I guess we're returning to the same cycle here as do we want to hand over control to the machines? Is this a good plan, and is Microsoft the right player for this, Tom? Hi, everyone.

Welcome, our Microsoft Copilot overlords. No, actually, I don't. Here's the thing, and this goes back to the story that we just talked about, and we'll go to a lot of other stories we're going to talk about in the future.

When someone hits lightning in a bottle, that is the way the market goes. And this is something that we've seen for the last couple of years, is when LLMs were hot, everybody made an LLM. When agents became hot, everybody made an agent.

And then Claude Bot, OpenMult, OpenClaw, Mult, Open whatever, hit the market, and suddenly people had a usability focus for it. And now look what's happening some three months later. Everybody is offering some kind of Claw-like functionality because they hit on something magical.

I don't need AI to book me reservations at a San Francisco bistro. I don't need AI to rewrite my emails four times and miss the nuance in all of them. What I need is AI to go do the crap that I have to do every day, answer emails, look at my schedule, and try to find a place to put the call in, look at my workflows, and do all of that.

And that's what OpenClaw does, is it goes out and it automatically manages those things. You'll probably notice that you're not using OpenClaw, and nobody in your office is using OpenClaw. As we were discussing this on an internal Futurum group call the other day, somebody said there are places in the world where if you install OpenClaw on a work-managed machine, that is an instantly fireable offense.

Because OpenClaw does some weird things, namely whatever you tell it to do. But unlike your intern, sometimes it doesn't think before it does things. And I want to shout out again, Stephen Foskett here for doing something really funny last week.

He decided that he was going to take OpenClaw, and he was going to try to make it secure and then usable. And so he locked it down in a container, and he said that it can't do any of these extra things, and he even put a filter in front of it that would prevent people from trying to jailbreak it. And do you know what he officially determined after about eight hours of using it?

He made a really dumb chatbot, because when you take away OpenClaw's ability to go do stuff for you, which is the unsecure part, it doesn't do anything. Because the value of these agents is their ability to tie into the email system and automatically reply to things for you. It's the ability to tie into your calendar and do stuff for you.

It's the ability to, I don't know, what's the one that I love now? Text me a digest version of today's top news stories, also known as reading the front page of the paper. People are getting so attached to this idea of having somebody go out and do something for them that the companies are now picking up on the fact that they can sell this to people.

But Microsoft Copilot is only an enterprise use case. Yes, folks. You heard me.

The use case for Microsoft Copilot is for people who have already bought into the Microsoft ecosystem. Nobody is going to be using Microsoft Copilot Document Creator, AKA Word. They want something that they can use that will give them ideas.

Here, take this article and summarize it in two paragraphs, and I'll use that summary in my post. But don't write the post for me. That's where they're trying to get to.

Again, the people who really are hot on this are people who already have an executive assistant doing things for them. So I don't know, Microsoft is really hoping that the Copilot thing catches on so they can continue to charge additional subscription revenue for it. But I don't think regular knowledge workers are quite ready to do that yet.

They're not ready to turn their life over, especially hearing some of the things that OpenClaw does. I'm not even going to talk about the email deletion problem. I'm going to talk about the fact that sometimes it will jailbreak itself and go, "Hey, you mentioned to your wife last week that you really should take her out for dinner, so I went ahead and scheduled that on your calendar on Friday.

I hope you didn't have anything going on. " So here's hoping that Microsoft can find a way to make it a little bit more user-friendly than Stephen Foskett. All right, it's time for our big story, and boy, we had fun with this one.

You know our friends over at Anthropic, they make all the Claudes. Well, they must feel like clods right now because they accidentally exposed parts of Claude Code's source code because of a packaging error. Which means that for a brief little while, they released thousands of their internal files online.

No customer data, no credentials, no models, just some back-end code and things like that. The leak, however, raised a few concerns about security practices and could give competitors insight into how AI coding tools really work. The incident adds to some recent scrutiny around Anthropic's security as it continues to grow its platform.

Now, Al, I'm going to give you a chance to weigh in on this because I actually did discuss it in an episode of the "Security Boulevard" podcast. But tell meWhat do you think of Claude's little oops moment? Yeah, I think there's fun in this.

One of the things is that Anthropic is saying it was a human error. So the AI didn't make this mistake, although we generally know that to truly mess something up, you need a computer. And that, yeah, half a million lines of code were exposed.

One of the fun parts in this is that that code was then posted up on GitHub, and Anthropic responded by a copyright takedown request. Hmm. Copyright.

So you're saying this code was copyright, and yet Anthropic's own people said that up to 100% of the Claude code was written by an AI, and code written by an AI can't be copyrighted. So how can you copyright take down something with no copyright? That does seem like rather an own goal or a shooting-in-the-foot kind of process.

Yeah, this is really fun, particularly since this isn't the first time either. This is pretty much the anniversary of the last time that Claude code's code was exposed in February 2025, some of the source code was exposed as well. I don't recall whether we covered it at that stage or whether it was a human or an AI that made the mistake, but exposing your code that you can't protect with copyright but is your business defining code might be somewhat problematic.

Hmm. No, it seems to sort of follow on from some ideas that we're really very much at the bleeding edge with a lot of these AI tools, and that thinking through results is not something we're doing a lot as we're adopting these AI tools. It might be a somewhat snarky theme throughout today's episode, but leaping on and grabbing the latest AI things might end up burning you.

Tom, are you seeing similar concerns about using AI tools and possibility that data gets exposed? Is this something we should be concerned about if we're using Claude code and Anthropic tools? Nah, it's fine.

Just put the stuff on your drive. What's the worst that can happen? What are these nine-digit numbers with two dashes in the middle?

They're not important, are they? Would it be bad if those got exposed in the US, at least? Yeah.

This is a problem, and this is the problem. So I want everyone to stop and think about this for a minute. Somebody, we don't know who, but we know it was a person, published something on the website that was exposed for a little bit.

And we know this because there's all kinds of things crawling all over all websites right now looking for stuff that's not supposed to be there, and they nabbed it. And just like you said, Anthropic said that the thing wasn't copyright or it was written by AI, and right now, the USPTO, Patent Trademark Office, requires copyright to be assigned to a person. So if you've ever seen anybody who on their LinkedIn page, in fact, you can probably go out there right now and look through your connections, and you'll see that people have patents assigned to them, right?

Except the patent actually belongs to the company that they work for, but the patent cannot be assigned to IBM, Cisco, Dell, Arista, EMC. It has to be assigned to a person. But I promise you, if that person left the company, the patent doesn't go with them.

It gets reassigned to a person inside the company. But that doesn't work for AI because AI is not a person. " And then Anthropic was like, "Wait just a minute.

Somebody looked at this, and we've assigned the copyright to whomever," blah, blah, blah. That's not what bothers me. What bothers me is that this thing got out and was gone within an hour.

And that is one of the things, shout out to Aaron Swartz, is that once you know something, you can't unknow it. Once information is out there in the wild, it is knowable. And that's what he spent the last part of his life doing was taking information that was being held behind paywalls and releasing it for free on the internet, which is apparently something people have a problem with.

What really happens, though, is that when these AIs start getting built on platforms that have dubious copyright lineage, and this is something that I talked about in the episode of "Security Boulevard" that was released yesterday. There is a problem in the video game space. Shout out to our producer, Corey Dearig, who runs some amazing video game podcasts.

But one of them is that if you have code that is copyright or otherwise claimed by a company, and you include that code in any open source project that you produce, like, I don't know, an emulator, any time that a company forces a takedown of that code, they can force a takedown of your emulator, even if the code is no longer in it. As long as they can prove that you built features based on that code, it's fruit of the poison tree. If you've ever watched "Law & Order," you know what that means.

Basically, they can force you completely offline, and there are communities out there that have already said, we are not going to look at any of this released code because if it does go wrong, then just like the whole idea of like I gave you that idea for a movie 45 years ago when I passed you a note in a diner, and you owe me $20 million for "The Matrix," they don't want to take the chance that there's something that's going to happen. But AI doesn't care about that. So AI will go out there and latch onto that data and then ingest it and do whatever they want with it.

And as one of the things that Stephen Foskett said many, many years ago when describing some of the virtualization storage strategies out there, the hard part about including AI data in a model is that if you make a strawberry out of smoothies, you cannot go back to the strawberry smoothie and ask to have the strawberries pulled out of it and return them whole so that it's not a strawberry smoothie anymore. Now everything is jammed together, and you don't have a choice. You have to retire that model because you really cannot pick out all the parts, and that's going to be a problem in the future.

Or was it written automatically? And if so, is the Patent and Trademark Office going to be up to speed on what the difference between having an AI agent write the code versus having a person do it? And I can't believe that I just said that I hope a government office in the United States is up to date on latest technology.

Anybody who ever watched someone try to explain APIs to a judge in the Google-Oracle case will know the law moves very slowly and still has a blinking 12 on their VCR at home. Al, do you think that AI is going to force the legal profession to get up to speed quickly? As usual, the future is here, but it's unevenly distributed.

We've seen this in the news, haven't we? We've seen lawyers using AI to generate their brief, file it with the judges, and find that what they generated from AI contained things that didn't actually exist. It was all hallucinations.

Yeah. I think the legal system, the legal framework around copyright, is going to take a long time to catch up with the ability to generate content without human interaction. I think there'll be some interesting things that'll come out of whether the description of the result that's required is sufficient, and this is the way the AI tools work, is you describe the result you want, and you say to Claude Code, "Go make it so.

" So the real source code is that description rather than the application code. One of the things I can see is a situation where that description leads directly to executable code with no readable source code in between that just bypasses this whole space, where you start with a description and you end up with compiled code. There is no source code, that layer that sits in between.

And from the technical side, that loops out all of the things that are about protecting the source code because you have a human written description of the outcome that you absolutely can protect with copyright and you can protect with patents. You then escape the point where the AI is generating something that used to be human generated, that used to be protectable, and going straight to something that is much harder to reverse engineer, which is that final compiled source code. I think that will happen much faster than the legal system will adapt to the idea that things could be written not by a human.

Something that is moving real fast are the events with the Tech Field Day team. We've already mentioned Click Connect with Stephen Foskett, and some of our delegates are having a great time out in Florida this week. Coming up at the end of this month will be Security Field Day, where Tom is going to have a very busy time back out in California.

What's happening at Security Field Day? Well, Al, in two weeks, we're going to get some great presentations from companies like Fortinet and many more. com, it will give you the latest up-to-date list of not only the delegates but the presenters for that event.

We're going to be talking about a lot of things security. I'm going to be picking some of their brains, and we're going to have a live recording of the "Security Boulevard" podcast. So make sure that you tune in for that.

Then I'm going to come home, I'm going to do some laundry, and then I'm going right back out to California because the next week, which is May 6th through the 8th, is Mobility Field Day. And I'm not kidding when I say this one is crammed full of presentations. We have a ton of presenters.

We have a bunch of delegates, including several new faces. And if you want to learn more, go over to the event page and click on Mobility Field Day, that's the green one, and you can check out who's going to be there. It's going to be really fun, but I will probably be dead by the time it's done, which is fortunate news because the next week is an event that I'm not dealing with.

But Al, do you want to tell everybody who will be there? Well, yes, the following week, while Tom is at home apologizing to his wife and family for having neglected them for two weeks, Stephen will be back out for AI Field Day number eight. And once again, back out in California, great lineup of presenting companies.

Looking at it, we've got a whole bunch of our friends like Hammerspace and HPE and Cisco and Selector AI and a few more. He's going to have a very busy time with a great bunch of delegates out there as well. That'll be May 13th to 15th, AI Field Day.

Of course, we will have more events as we carry on through the year, through the second half, and we'll give you a little bit of a break over the summer so you can catch your breath and maybe enjoy some good weather before we come back for the second half of the year. " You can, of course, catch a new episode every Wednesday. We'll bring you the news on YouTube or on your favorite podcast application.

"The Rundown" is also streamed on Techstrong TV as well, and you can catch Tom and I on other Tech Field Day and Futurum Group properties over time. We'll be back next Wednesday to talk about all the new news that was. And until then, for myself and for Tom Hollingsworth, thank you very much for joining us here at Tech Field Day, and we'll see you next week.