Organize projects with workspaces and build agents with skills

your Neo typer jumping around on Facebook having fun. Oh, yeah. Hey, everybody. How the heck are you doing this fine day? Dustin and Steve here from the AI Academy. Now, you're all not friends with Dustin on Facebook. I happen to be inner circle friend of Dustin's on Instagram or Facebook, I don't know which one it is. But Dustin has been oversharing his image editing as he makes himself look tremendously heroic, leaping from building to building in superhero type stuff. Yeah, pretty good.

I have had quite a few requests of how are you doing that, and how do I do it for myself? Maybe we'll dedicate another... I think we should- Yeah, we could. But I think the thing is, you know what makes it work is your freaking hair. It's the long hair. The long hair does help. It does that Neo type feel. Anyways. Well, he didn't have long hair, he had short hair.

But anyways, it's very, very cool. Hey, so what do we got planned for today, my friend? Well, so last week, for anyone who's paying attention, OpenAI had their developer day. They made some announcements, they announced some new products. Amongst those new products are a thing called Dots, which are always-on agents, Spaces, which are collaborative environments in which you can share with other members of your team,

and Pages, which is essentially a document editor. And so today I wanted to dive into those because it is a reminder that Magai has been right all along. But essentially what OpenAI has did is they've finally given people what Magai has always called workspaces.

In the beginning we called them brands, but we quickly realized they should be called workspaces because that's what they are. They are ways to silo off your different projects. Now, ChatGPT has had projects already, and Claude has had projects, and they've had them for a couple of years now, I think. Again, Magai was still first in this concept, but we realized really quickly that you need a way to separate your different things, your personal life, your work life, your side hustle, your whatever else you're working on, right?

They shouldn't bleed into one another in many scenarios. And so OpenAI introduced this as if it was a brand new product. And part of this today is not so much pragmatic. We're going to dive into how do we use this type of tool, this workspace element, this editor. But how do we also recognize when a big company is just reselling us the same exact product, and they've just repackaged it in a way that sounds new and exciting, and is making all the headlines because they have mass psychology

tactics at play for them. So Spaces, all they are is workspaces. If you're familiar with Magai, you are familiar with the idea of workspaces, and so I'll open up my Magai here. We'll create a new chat here. Let me flip over to light mode so it's easier. I'll zoom in a little bit.

But in Magai, in your sidebar, under your account tab... No, I don't want to sign out. At the top of your menu here is listed the current workspace that you're working in. This dictates everything that you see, all of your chats, all of your media. This is assigned to your workspace, and when I change my workspace to, say, The Board workspace, where I do all my The Board work, everything changes. The chats list, the media, the scheduled tasks, all of that is

isolated as a way of siloing off your different projects. And so if I go to my personal workspace, again, you'll see completely different listings here, different chats, different media. All right? And we've always done it that way. And there are many, many benefits.

Obviously, the first of which is having that clarity, that organization of not bleeding things over into one another. But also from a team aspect, if you work on a team, if you share projects amongst your teammates, you can add people to your workspace. Just like here in Magai, I have some team members, and that allows you to have this collaborative space where you can then share chats.

So if I wanted to share this chat with one of my team members, I can say, oh, well, I want Vilni and Paul to be able to see this chat in their feed. And furthermore, I want Precious to be able to actually edit and contribute to this chat. All I have to do is set those permissions up, hit save, and the next time those team members log in, they'll see those chats inside their feed.

So it's not just siloing off your projects, but also being able to collaborate and share content across to your teammates, to your virtual assistant, what have you, your boss. I'm curious how many of our... We've got 65 people on with us right now. Excuse me. I'm curious, how many of you work in a collaborative environment where you collaborate? Just in chat, tell us, do you collaborate documents, or do you collaborate AI currently, if you do one or the other?

I know that I, with my team, collaborate documents all the time, but we have not started collaborating our AI agents or our AI spaces. And I'm just curious how many people are doing it because it sounds awesome. Yeah. Yeah, and I would assume the large majority of people are either solo or at most, maybe they have a virtual assistant that they share things with.

I share a lot of things with my virtual assistants because oftentimes either they will begin something and I will be the last person to see it and edit it, or I will start something and have them finish it. So it goes both ways a lot of the times for me personally. And I have a team of 12 now.

Most of us are just collaborating back and forth in Slack, just sending messages, but every now and then we'll have an AI project that we'll work on together. So for example, if I go up here to apps that I've created, and let me see. Let's see, add Battlegrounds. I think this might have been something that I was working on.

Yeah, so we were thinking of trying to come up with a way to build sort of a prompt library that people can vote on and up vote and surface really good prompts, like a new type of library. And I wanted to share this with my team, and instead of having to screenshot it or instead of having to use the public share link, which I think I did at first, I can also just share it directly with them and give them access to this and just add them as an editor, and they can continue that conversation, making refinements

as needed. So, that's one example of collaborating with my developers on stuff. And that's been really helpful in a lot of cases. Or if I had a marketing document that I wanted my marketing person to review, or vice versa, she wanted me to review it, that would be another great example here. But workspaces are great, and that's how OpenAI is pitching them. They're really pitching them as collaborative spaces.

They're not so much leaning into the individual, siloed-off projects type thing, because again, they already have a feature called Projects. And on top of that, Spaces is limited to business customers. And I believe you have to have the $100 a month plan in order to access Spaces. I'd have to double-check that just to make sure, because they announce a lot of things, and a lot of those things are paywalled behind the $100 a month plans.

Now, see, the other benefit here of spaces that, apart from just siloing things off, apart from the collaborative nature that we've always built into workspaces, is the ability to have customized context. And we've talked about this many times in our AI academy sessions, but it's extremely helpful to set up custom context per workspace. And for those who are new, essentially what this means is you can put information into the

custom context box here for your workspace that is project-specific, that you want to make sure that every time you start a chat in this workspace, the AI already knows these things. These could be business details. What is your business that this workspace is built around? What are your products, services, operation hours?

Anything that's critical information to your business that, again, every single conversation you have would be enhanced by or would be worth having in its context. For regulated industries, if you are in a regulated industry, it would be really important for you to include any of those regulatory details in this custom context. So again, you start a conversation with the AI about something, and maybe you forget that you have to run this through any regulatory things.

If you've put it in your custom context, the AI will already know it and will already know to check it against that. So custom context comes in extremely-- It's highly valuable to have that in there so that you don't have to repeat yourself every time. On top of that, you can upload custom knowledge files. So if you have standard operating procedures, if you have-- Again, this could be a good place for any regulatory documents.

Any information that is documentized or something you want the AI to constantly be able to reference in order to better serve your needs when you start every single chat. Think of it as pre-filling the context without having to repeat yourself every time in every new chat that you start.

So things like union handbooks or- Yes ... employee handbooks. Contracts, if you have contracts, I mean. Mm-hmm. Putting them in. All of those sorts of things are just terrific to have. Yeah. There's just so many places. If you work for a community center, they have a thing of events, right? And all of the descriptions of all of the activities that are available at community centers, the Zumba classes and all that sort of stuff.

Input that, just save that as a document, and then you've always got access to it. Yeah. There are so many different baseline kind of institutional knowledge things that should be incorporated in your workspaces, in your workspaces. Absolutely.

Sorry, yeah. You can drop them in here, and unfortunately, I don't have any knowledge files dropped here. But you drop them in here, and then anytime you need to add or remove to them, you can always come back and add or remove knowledge files from your workspace. Now, a recent addition that we've had is this idea of memory and the shared memory across chat. And if you look below the knowledge files area inside of the workspace editor, you'll see the memories item here, I guess. And you can click on it, and you can actually see the list of

memories that the AI has created while you've been working with it. And you can look through them, you can modify, delete them. Are these like summaries for other chats to know what went on in the different chat? Essentially, yes. It's summarizing certain things that you've done, or maybe you had a preference that came out as you were working with the AI.

It'll take note of that preference and kind of store it in its memory. It's not necessarily the whole of a conversation. It might just be a snippet that the AI goes, "Oh, this would be helpful to know going forward, so I'm going to store this in my memory." And these little snippets help continue to improve the AI as you work more and more inside this workspace. And the reason that they're limited to this workspace is because oftentimes you may have certain preferences, and again, don't want things to bleed over into other projects.

So memories are now a product kept inside of your workspace and limited to each workspace that you have. And so when you switch workspaces, you'll have a completely different set of memories based on the work you're doing in that workspace. So this, you're probably uniquely equipped today to deal with this particular question, but Melody's asking, if you're creating these knowledge bases, how does AI deal with proprietary documents and company-based privacy issues?

That's a great question. So Magai, from day one, has been privacy-focused. None of the content that you add into Magai, whether that's through chat or whether that's through knowledge upload or whether that's through custom context or anything. None of that is used to train models. Every single one of our vendors has to agree to our privacy and security policy that they cannot store your information in such a way to use it to train their next model. So all of your proprietary information, anything

you put in Magai is always protected by that clause. So no model can use your information and use it as training data. Even the utilities that we use, it's all stored in Magai-owned servers and Magai-owned storage stacks, without using any of the technical terms.

All the vector databases are under our control and owned by us, not visible to anyone else. Except for the instance where OpenAI has to process the request and then send it back, and then they are required to delete that information once they are legally within their ability to do so. Some states and governments require all things to be sort of recorded or stored for a set amount of time. I think it's 30 days, is just for your security and so forth. But yeah, nothing is ever used to train

your data. So your proprietary information and stuff is safe inside of Magai. Julie's asking, would you suggest putting branding info into a custom context or a knowledge file, or does it not really matter? I would absolutely suggest that. So I'll give you our Magai workspace as an example. So some of the things that I've put in here, let me expand this real quick.

So in here I put things like what makes us different. I put in things like our core features, and then I put in our plans. And then I have a whole section here that says Magai brand style guide. And I go so far as to actually put a URL in there where the AI can find our logo. This is just a public link where our logo is stored, and that way if it ever needs to find our logo and put it into something that it outputs, it knows where to find it. Colors I've listed, typography.

This is like a very minimalist version of branding guidelines, and I suggest everyone do something similar. If you have a strong brand and you want to make sure that it's followed and consistent, absolutely put that in your custom context for the workspace where you do that kind of work. One thing I've noticed is I did a brand kit as far as graphics, and as opposed to just being a text document, it created a beautiful PDF-

Mm-hmm ... with all my branding information. It just makes me look so like an agency had created it for me. Yeah. I've done that several times over. For every project now, I basically do a brand style guide, and I'll show you the I go a little bit over the top in that respect. Really?

I think I do. Say it's not true. You be the judge. We be judging. Our current magi.co/brand. So we have an entire webpage dedicated to our brand, mostly for press entities when people are doing press on us, they want easy places to grab the logos. But we have what to do, what not to do, colors breakdown.

Our gradient is very specific in how it's used, typography, voice. And to take it even further, we have an internal one. Oh, I think I typed in the wrong thing. There we go. So we have design system, that's what I'm looking for. One design system for all of Magi. So I had the AI literally go through it and make the finest tuning of every piece of what we build visually. So it has rules for how it follows it.

It even has components like these are what buttons look like, and these are what forms look like. Brand assets, of course, which is the same as what we showed previously. So I got a little crazy on exactly how things are styled and how things look. And you can get as crazy as you want to, I guess, with that.

But when it comes to your custom context, you don't want to put everything in there. You don't want to just overload that context. You want to keep it efficient when it comes to your custom context that you put in here. I would say keep it to 1,500 words or less. Short, simple, to the point. These are just bullet points mostly.

Because what's essentially happening here is anything you put in the custom context is going to be added at the system prompt level. So it's going to be inserted into the conversation before you start it, so that every time a new request is sent, that information is going to be there. And so all of that is adding to your context length.

And we had a discussion a few weeks back about not having too long conversations, because the longer those conversations go on, more expensive they become, the faster it consumes your usage, right? So you don't want to overload the context unnecessarily. You want to keep it as concise as possible. And you'll find that the AI, it's gotten to a point where it can do a lot with a little. You don't need to overexplain things.

In fact, I read a really interesting article on how one developer has instructed his AI to speak like a caveman, and the reason he's having it speak like a caveman is because it reduces the amount of words or tokens that are output by 60%. So he reduces his usage by more than half just by having the AI speak like a caveman and using less words.

Just returning to this context. Again, it comes down to vernacular in a lot of cases. What you're talking about as far as limiting the amount of context that you put in, correct me if I'm wrong, but in Claude, that's the system instructions. Claude, for example, separates the system instructions from the context.

So the system instructions you want to keep lean, but the context, they want you to upload all the related files. They are two different things, right? All of the context is not loaded till they have to look for it, right? In that case. There are several layers of how AI puts together context. There is the obvious layer, which is the input-output messages, right? You send a message, "Hello," that is an input message, and you get an output, right? That's one layer of context.

Another layer of context is what you said, the system prompt, and that's your agent instructions, right? We've talked a lot about these instructions that these agents have. That's another level where context is added that helps guide the AI. This is more specifically about behavioral context, right? And that's another layer. The custom context for your workspace is also layered into that system prompt, that system message.

So that's kind of all the same sort of context input method. Now, another method that was recently popularized is this skills method. Mm. Yeah. So skills are sort of another layer on that, whereas it's not inserted into the system instructions. Every time you send a message, the AI sees its skills and makes a decision, which one of these skills do I need to reference or read before I respond? So it's going to go through and it's going to pick which of these skills are relevant to the current discussion, the current request, and it's going

to read those ones before responding. And so it's sort of adding its own additional layer of context for any given request. Now, another layer is that knowledge layer. So agents can also have knowledge files uploaded, just like workspaces can have knowledge files uploaded. Knowledge files are a different type of context that's added, but it's also selective in the way that skills are selective.

So if you upload 15 documents here, it's not going to read all 15 every time. It's going to take your request, try to figure out which of these knowledge files, and specifically which part of these knowledge files, because it doesn't read the whole file every time necessarily. It'll say, "Which of these knowledge files is most relevant for me to read before I respond to help me respond better?" So that's another layer.

And then the last layer would be very similar, but if you have any documents attached, like a canvas element here, that would be another layer. It would read whatever content is there. And I guess the other layer there would be file attachments, if you attach any files. So there are lots of layers that the context is added at this point in how AI is capable of reading context.

Great. Just catching up on a few other questions. William's asking, "As a member of nonprofits, does it make sense to have a workspace for each of them?" I think so. Again, like the shared memory, just the nature of things bleeding over into one another.

A lot of times you don't want... The example I always gave in having different workspaces is, I'm a Marvel fan, but I also own an AI company and talk to enterprises, and so if I had all my Marvel fandom chats that are just for fun in the same workspace as my professional business communications with enterprise C-level executives, I don't want those kind of influencing one another, right?

I don't want to have the AI write a response to an email, clean up my language to sound more professional, and then throw in a Captain America reference, right? It just wouldn't be appropriate. It would be me fighting against myself, in terms of context and what's influencing the AI. So we separate everything out that is worth separating, if that makes sense. That you don't want to bleed over into one another.

Superb. Okay, let me go back here. You can continue. Yeah. So the other thing that they announced was Pages, and again, this is really just a repositioning of something they've always had, but maybe a slight upgrade as well. So Pages are essentially a document editor that you can use in ChatGPT now. You've always had a document editor for the last several years in ChatGPT, just like Magi has a document editor. So you can say something like, "Create a report

document on the migration of whales." In the canvas. So we'll just tell it in canvas. So you can have the AI build a document for you. Should be fairly quick. So here we go, document pops out, and it works similarly in ChatGPT.

And once it's finished writing here, we have full editing capabilities. We can get into this document, we can write, we can highlight items, we can bold, italic, underline, do all the document things that we want to do with this. You can say insert or generate a photo for the top and insert it.

And the AI can do that. It put images in there, it put charts in there. And so essentially that's all it is. They've taken their previous document editor, repackaged it, and maybe upgraded a bit. There are some cool interactive things you can do with the collaboration aspect. Multiple people can be working in that document at the same time. You can kind of see who's in there, kind of like Google Docs.

If you've ever shared a Google Doc with coworkers and all of you are working in it one time, you can kind of see the people working in there. Oh, now it wants me to pick, so nano banana. Sure. But that's essentially all it is. And all of these tools that are, all these repackaged products that these AI companies are coming out with, they essentially boil down to the same exact things.

You have an AI model who is given tools in which to work, and it is given an interface in which to present the work that it's doing. And we'll see if it adds this photo of whales here. There it is, right at the top of the document. And there we go.

So you have your document editor. Next question. Oh, sorry, do you want to keep going? No. Yeah. Let's jump in. Quick question from Lee Vishlof. Lee is asking, "Do you have a flowchart that shows all the layers available and the sequence and priority?" I think he's talking about the different contexts, the different building blocks for context. Could be kind of a pyramid, wouldn't it, actually?

It would be a wild... It would kind of be all over the place. Again, going back to this concept of it's all just tools. It's a model that has tools. And the way that the models work is not something the average person needs to know. You don't need to know what system prompts are, what system instructions are. You just need to know you give an agent instructions, and it follows them. The mechanics of it aren't really necessary to know or understand.

It's just knowing what tools you have access to is the helpful part. And one of the techniques that we've used almost since the beginning of AI is using AI to help make sure that we have all of the different areas covered. Mm-hmm. So asking AI, I'm setting up this project, what do you need from me in order to do your job most effectively? And then AI will slot that information in or tell you where to slot it in, depending on which tool you use, because as we know, the nomenclature is so different with all of the different tools, but they're the same

basic silos of information. So I would just collaborate with my AI tool of choice. If it's Magai, just ask it, making sure that all of the different bases are covered. Is that good advice? Yeah, absolutely. Good. I like giving good advice. Continue on, sir. Yeah. So, the overall lesson here today is hopefully you're using workspaces. And if you have not learned to silo off your different projects, hopefully you understand that there is great value in that, especially if you have an environment

where memories are being created and it's learning from your chats, and you don't want those things bleeding over and negatively influencing things that they shouldn't. And next time the big AI company comes out with a brand-new sparkly feature, just realize that it's just a model that has some tools, and they're presenting it in an interface. I think we also talked about this a little bit last week.

They introduced Dots as their thing, their new always-on agent. Similarly, Grok came out with Bots several weeks ago, and now there's this app circulating called Instinct, which is your AI assistant that text messages you and kind of always on. Then we have Meta's Muse, which is their always-on agent. These are just models that have been given tools, and they're not always on 24/7. They're not sitting there running their computer nonstop.

They have a schedule. So it's just a scheduled task that schedules its next task automatically. You give it a job, and it does a thing, and then what it does is it says, "Okay, what's the next logical step here, and how long should I schedule that next step out to wait for the user to do something?" So it's all just a model on a schedule, given some tools like connections to your Gmail or connections to your whatever tools it needs to accomplish that job.

So it's really all just the same product, all the same workings, just maybe presented in a different package. Should we run quickly through what the different platforms call workspaces? Yeah. So again, in ChatGPT and Claude, historically, it has been mostly called projects. And if you open up your ChatGPT, you will see you have on the left-hand side the ability to create

sort of folders. That's one way to look at them. Great. Connecting my bank account to ChatGPT. No, thanks. No, thank you, Sam Altman. So projects are essentially like creating folders for different projects that you're working on. And they live in your sidebar here, and they're kind of all jumbled in the same sidebar. Same with Claude.

They allow you to create these folders as projects. Whereas in Magai, your workspace is what you are in, and your sidebar only contains things in that project. And you can have folders within that workspace. But it's all part of the same workspace. And then to change it, you just click here, click on the workspace, and choose the one you want or create a new workspace with a new name, new logo, and any context or knowledge files that you need.

So that's the same across all things. There are other apps out there. I think Gemini also calls them projects, if I remember correctly. But same concept, siloing things off. And then everybody's coming out with their always-on agent. Always-on. But it's all the same thing. They're all the same product.

Just like back when OpenAI came up with GPTs, it's the same exact concept that we had when we called them personas at first. Gemini called them gems. Claude never really took on the idea of agents or personas or custom instruction GPTs. They invented what we now know as skills, and those skills were just shared across the one generic agent that kind of helped guide and mold it into specific one-off or specific tasks, I guess.

And then you just kind of accumulate skills over time, which can kind of be a disaster. Which is why it's helpful to have agents where you can sort of group the skills yourself. And so my blog writer, for example, only needs these skills and doesn't have to sort through 15 other skills to kind of get the one it needs. You can build skills for each individual agent.

So every company has sort of different approaches to these ways of working with AI. But ultimately, you want to find the one that makes the most sense and is the easiest to use and doesn't change every other week. That's probably the best idea. Okay. Oh, just a minute. I'm in the middle of a reply here. Let me go back to the thread here.

Let me see here. Sorry. Lisa, "Many of the AI results include LaTeX math embedded in the markdown text. I don't see a clean way yet of printing this output in one step to get an easily readable format. What am I missing?"

I'm not familiar with LaTeX. I don't use it in a lot of what I do. But you might be able to just ask the AI how to do that. In most AI tools, Magai included, you might just be able to have it just say, "Can you create a canvas with this LaTeX output formatted correctly in a way that is easy to read?" Something like that. Just prompt it into the canvas, I guess.

So Kate's asking, "If I create a workspace in Magai and share it with my colleagues, would they also have to have a subscription to Magai?" Yeah. In order to use Magai, you need a Magai subscription. Mm-hmm. Julie's asking, "How do I create a skill? How do I think about what is needed for a skill? I'm a bit lost in this, to be honest."

Yeah, no, fair question. So think of a skill as just, again, creating an agent the same way you create an agent. The beauty of it, the way that it works in Magai, is you don't have to know how to do it. So you can literally just say, "I need an agent to help me with book editing."

And say something like... Obviously, only you know your job and the job that you need done. But for the purposes of this demonstration, I'm going to be vague. But as with all good prompting techniques, you want to be as specific as possible. Define the job that you need done in as much detail as you can muster. We'll just say, "Editing my fantasy superhero

short stories. Build an agent that will have all the skills I need it to have to do a great job as my editor." Magai is just going to do all that work for you.

You just have to articulate what you need, articulate what you want, and Magai will help you build it. So it's going to be very elaborate here. I think it's mostly thinking. Yeah, this is just all, it's planning it out. "Build this proprietary dedicated editor agent, strong editorial engine, plus a full set of role-specific editing skills layered the way a real editor works.

Let me create the agent first and then attach every skill it needs." So it created our agent, Red Pen, and I can switch to that agent right now by clicking that button. But probably wouldn't be helpful because it's still creating the agent, and so it's creating these skills. Developmental edit, scene craft, dialogue and voice, line edit prose, copy edit, proofread. Man, it's really going to town on this, isn't it?

Many skills it just created. Eight skills, and then it explains itself like AI always likes to do, and some suggested next steps. We can test it, we can blah, blah, blah, blah. But what I can do is I can actually switch to the agent. I can just click on any one of these skills, and we'll switch to the agent here. Here's our Red Pen, and it'll be probably at the bottom here. I can open that up now, and you can see it's created the description for me.

It's created the instructions for me. I didn't have to go create those agent instructions. And here are each of the individual skills that it produced. I can hit the settings button and open those up and see what those skills are if I want to. I can make edits to them. But I mean, it did all the work for me. You don't have to really know, how do I do this?

Much like in the early days of AI, people would ask, "How do I get the AI to summarize a video for me?" Just ask it. Just literally just ask it. Well, how do I get it to do this thing right here? Well, exactly how you explained it to me, explain it to the AI, and it can probably do that for you. You don't need magic words.

You don't need magic knowledge anymore. It's really kind of remarkable. It's still kind of magical in the way that it can just do the thing that you want it to do just by asking it in normal human language. Especially now that we've built this automatic agent and skill creation in Magai, it really makes the whole process so much smoother. Now, you can even have it create its own profile picture.

So I can say, "Create a profile picture for Red Pen and set it for me." And it can do that. Oh, and of course, it wants me to pick a model. Of course, I want...

We'll do Nano Banana Too Light. So it's going to go ahead and generate its photo, and it's going to actually set the photo for the agent without me having to do any of that work. Interesting. Excellent. And there we see Red Pen now has a profile picture.

There you go. Awesome. Let's call it a day. Just a programming note, for those of you that have joined us for Webinar Wednesday in the past, today's Wednesday, we have a Webinar Wednesday today. Great topic today. How often do you have a phone number appear in your texting on your iPhone or on your... Yeah, usually it's iPhone.

But in messages on your computer, it's got their name, and you're saying, "I thought my accounts were syncing, my contacts." Does that happen to you, Dustin? All the time. I have hopefully worked out all the reasons and discovered the issue, and if you wanted to learn it, today on Webinar Wednesday, there's the link to it. We'll be doing that in about an hour.

It's such a rabbit hole. Yeah. I have that problem constantly, so I'm into it today. Yeah. It comes down... And it's self-inflicted for the most part because we have so many devices and so many accounts. Mm-hmm. And we store all the accounts on the cloud, and we've had so many devices. You think of all the places that your address book has lived over the years.

Hmm. It's an issue. All right. That's it. We will be back... Are we back next week? I don't know. Let me just check. I don't think I've scheduled next week's yet, have I? Yes, we have. We're back next week. All right. It's just after Canadian Thanksgiving. To all of my fellow Canadians, Happy Thanksgiving. We have our Thanksgiving on Monday, and I am heading up to the lake tomorrow, so.

Well, Happy Thanksgiving, Steve. It's my traditional time to close up the cabin, get a last bit of fishing in, and so we're going to have a big family gathering up at the lake, so. Nice. Happy Thanksgiving to one and all. And thank you so much for all of your efforts today, sir. All right, folks, till next week, have fun storming the castle.