0:02 Hermes is the number one AI agent in the
0:05 world, but when you give it an agentic
0:06 operating system, you unlock
0:09 capabilities that 99% of people don't
0:11 even know exist. And in this video, I'm
0:14 going to show you exactly how to build a
0:17 visual intelligence layer for Hermes
0:19 agent and Claude code, which will make
0:21 using Hermes 10 times easier, meaning
0:23 you don't waste your time, and you will
0:25 get light years ahead of your
0:26 competitors. And if you're new, I'm
0:28 Jack. I built and sold my last tech
0:30 startup with a gazillion customers. And
0:32 now I'm running my own AI businesses,
0:34 and I share here the stuff that actually
0:36 works. So, if you haven't already, grab
0:38 that beautiful coffee, and let's dive
0:39 straight in. Now, you might be
0:41 wondering, am I in the jungle because of
0:43 this t-shirt? I can't tell you that. But
0:44 what I can tell you about this agentic
0:46 operating system is why it's important,
0:48 and why you need one now, and by the end
0:49 of this video, you're going to be able
0:51 to build anything you want to into your
0:53 own agentic operating system. Now, when
0:56 it comes to the Hermes agentic OS and
0:59 the Claude agentic OS, the way that I
1:00 want you to think about it is it is
1:03 essentially one location for everything
1:05 connecting together. Let me kind of
1:07 explain why and how this works, and what
1:10 the big issue is holding you back at the
1:11 moment. And that's the fact that we have
1:13 tools everywhere. We have Claude over
1:15 here, maybe doing some coding on Claude.
1:17 We have ChatGPT, we've Grok, Gemini, we
1:19 have open Claude, anti-gravity, Cursor.
1:21 The point is that everything is all over
1:23 the map. So, the first problem we have
1:27 is what we call context isolation, data
1:28 and information isolation. And so, by
1:30 building an agentic operating system
1:32 that looks like this, where we can bring
1:33 in so many different aspects and
1:36 avenues, we can what's called unify our
1:38 AI footprint. The second thing we can
1:40 do, we can unlock new capabilities that
1:42 you cannot get just using Claude or
1:44 Hermes by itself. It's super important,
1:47 and it can drive genuinely new insights,
1:48 stuff that you wouldn't be aware of
1:50 otherwise. Now, let me double click onto
1:51 this thing and give you an example of
1:53 why this is so freaking powerful. It
1:56 can, for example, dream. Now, one of the
1:57 things that you can do with Hermes and
1:59 Claude code right here, if I just show
2:00 you this in the operating system here
2:02 for a second, is that what this can do
2:05 overnight is think and dream about
2:06 everything you've done. It has full
2:09 access to your entire conversations in
2:10 Hermes. You'll full access to all of
2:13 your conversations in Claude, in
2:15 ChatGPT, in Gemini. And based on that
2:17 information, based on the skills you're
2:19 using or the skills that you're not
2:21 using, it will generate for you on a
2:23 daily basis ways that you can improve.
2:26 This might also, for example, be based
2:28 on your goals. If you have specific
2:29 goals, maybe you're trying to launch an
2:31 email list, maybe you're trying to grow
2:32 your business, get more roofing
2:34 companies, whatever it is, it
2:35 understands that and can give you
2:37 dynamic insights by connecting these
2:39 data together. So, it can dream. This
2:41 dreaming function is insane. It is
2:44 literally an agent thinking about you in
2:45 a nice way. We like it then when it does
2:47 that 24/7, even when you're thinking
2:49 about it or not. It can reread your weak
2:51 surface patterns, conversations, skills
2:53 that we use, what changed, what stuck,
2:54 without even asking about it. It can
2:56 give you a short morning brief every day
2:58 with a couple suggestions, and it can
3:01 compound progress um without willpower.
3:02 The second thing I want to touch on here
3:04 is, you know, personas. Every persona,
3:06 every skill in one view, we can isolate
3:08 and look at all of our different skills
3:10 that we're using on a regular basis. And
3:12 then we also have cost breakdown, and
3:13 I'm going to show you a skill in this
3:15 video whilst we're building out that is
3:16 going to reduce the amount of money that
3:18 you spend using Hermes or Claude code.
3:20 It's It's blowing it right now as well.
3:22 But the idea is we can see live usage.
3:25 We can see by the hour, by the day, how
3:27 much money we're spending on what AI and
3:29 where, and whether or not we need to
3:30 downgrade our plan, whether we're
3:32 wasting money, and how much context we
3:34 have left. So, let me show you what this
3:35 looks like in practice, cuz if you don't
3:37 understand what an agentic operating
3:39 system is, you're not going to know
3:41 exactly what you're building, or if
3:42 you're using this one, how you add onto
3:44 it effectively anything that you want
3:45 to. So, if I come over to the agentic
3:47 operating system here, the idea, as I
3:49 said, is to unify everything together.
3:51 If we look at Hermes, what I can first
3:52 of all see is every connection that I
3:54 have that is available to me or if
3:56 you're building this up for clients,
3:57 that's available for them. I can check
3:59 out what's available globally. I can
4:00 check out what my Hermes agent has
4:02 access to. I can see the version that's
4:04 running. I can see the models. I can see
4:06 the memory and how many times it's been
4:08 activated or used in this week. And then
4:10 here, what we also have is the ability
4:13 to chat to Hermes and say, "Hey, what
4:15 day is it?" Whatever the hell's going
4:17 on. I no longer just have to chat to
4:18 Hermes on my phone, for example. I can
4:22 chat to it in this beautiful
4:23 all my previous conversations here on
4:25 the left-hand side if I want to go back
4:27 and reference it. Something that you
4:29 can't really do in Telegram, so it's
4:31 great that it breaks that down for us.
4:32 Then we get to the point of practically
4:34 doing things. If you think about the
4:35 fact that you're chatting with Hermes
4:36 and you're executing, you're building
4:38 code, you're building projects, that is
4:40 freaking awesome. But sometimes, you
4:42 know, things extend beyond one
4:44 conversation and we want a dashboard to
4:46 be able to show us what our goal is and
4:48 where we're at in that process. So you
4:50 can use here the essential mission
4:51 control feature, which I highly, highly
4:53 recommend you use. All you literally do
4:55 is you come down, you copy this prompt,
4:56 then shoot it to Hermes. And then when
4:58 I'm in Hermes, I literally come down, I
5:00 just drop in this prompt, for example.
5:01 It will ask me questions and it will
5:04 build out for me a mid-term goal. That
5:06 could take weeks. And it will identify,
5:07 for example, what are the things that
5:09 Hermes can do for us, what are the
5:11 things that we need to do, etc., etc.
5:12 and double click onto what you're trying
5:13 to focus on. So here, for example, I
5:15 might say, "Hey, I would like to grow my
5:18 YouTube channel by 1,000 subscribers,
5:19 for example." Now, the reason why this
5:22 is so cool and it's so helpful to have
5:25 this kind of visual intelligence lab for
5:26 your Hermes operating system is you can
5:28 visualize and see the questions that
5:30 you've got. So Hermes, for example, with
5:31 this prompt will come back and ask
5:33 questions. And as you can see, it will
5:34 ask further questions and clarifying
5:35 questions, like what's your current
5:37 subscriber count, what's the main
5:39 content type and niche, how many videos
5:41 you currently have, um have you run any
5:42 other subscriber growth campaigns? And
5:44 then based on that, we'll build a plan
5:45 for us that we can use together. And
5:46 then when you come back here over, you
5:48 can actually see this. So, this would be
5:49 a goal for Hermes. You can go through,
5:50 you can check out the brief that's given
5:52 out, which is fantastic. I can come back
5:54 to actions. I can even see what my role
5:56 is here to record and upload the videos.
5:58 It's fantastic and it works dynamically
6:00 together. Underneath that, we've got the
6:01 Pantheon. This is the bit where
6:03 basically we're going to use this to
6:05 have a visual representation of our
6:07 skills. Now, you can have skills in
6:08 Hermes. It's a fantastic The guys have
6:11 done such a great job building it. They
6:13 really have. I personally like to have a
6:14 visual overview of this stuff. So, for
6:16 example, I can down. I can add in
6:17 personas. I may go for I don't know,
6:18 let's go for Athena for instance. And
6:19 then once I've chosen Athena, I can pick
6:22 the job description, the system prompt.
6:23 It's really cool. I think it's pick the
6:25 preferred model. So, Hermes can delegate
6:27 to that specific model. Again, it does
6:29 the same thing here for Mercury, um
6:30 where I just have like autopilot and
6:32 cron jobs. I might use them over a free
6:33 model because it doesn't need
6:35 intelligence. So, having visualization
6:36 for your personas, an easy way to
6:39 connect to GitHub, also just an overview
6:40 of your user profile, your agent memory,
6:42 your sold MD, so you can review it and
6:45 amend it. The ability to connect to your
6:47 Obsidian data visually and get that
6:49 connected. As well, crucially guys, and
6:51 this is one of the core points here, is
6:53 connecting the data lake with Claude
6:56 code and Hermes agents so that both of
6:58 them can basically lean in and create
6:59 insights from one another. All you do is
7:01 literally copy this prompt here uh and
7:02 just throw it over and then you're ready
7:03 to rock and roll. And then of course,
7:05 down here we've got um some different
7:07 skills it's got. Now, outside of Hermes,
7:08 you're going to basically launch it and
7:11 set it up with other platforms. So, here
7:13 for instance, I show my AI spend across
7:14 all of my different platforms, which I
7:15 highly recommend that you do in your
7:17 operating system across Claude code,
7:19 Cortex, Antigravity. This is the
7:21 dreaming functionality I was telling you
7:23 about, where every night it dreams and
7:24 finds out stuff and it's come back with
7:27 some really freaking interesting um
7:28 basically ideas. And it's told me stuff
7:29 like, "Dude, you don't need to be on
7:31 this plan because you're using only 20%
7:33 of your usage." It'll come back and tell
7:34 you loads of interesting things. And it
7:35 can Here it'll come down and tell you
7:37 the different sources you can tell the
7:39 Hermes agent and its context is
7:40 relevant. Now, we've got mission
7:41 control, which which is reflected, so we
7:43 could also pick those up in Claude if
7:45 you want to. We've got a list of our
7:47 skills. Then crucially, our memory,
7:49 what's plugged into our stack, which is
7:50 really key. Schedule tasks that we have
7:53 going out, and then just just general
7:54 data on how much we're using it. Now,
7:56 one cool thing on memory here, this
7:57 realistically is one of the coolest
7:59 things about Hermes is the way that it
8:01 does its memory. I've covered it on
8:02 different videos on my channel, but one
8:04 of the things and the reasons I love
8:05 running Hermes on my on my MacBook
8:07 basically is because it just makes the
8:09 whole system so much simpler because one
8:10 of the really cool things I can do is
8:12 just see everything together in one
8:15 location. I might Obsidian, my local
8:16 Claude, I can connect to my Pinecone. I
8:18 can do anything, and Hermes just has all
8:21 that context. Now, the idea here is that
8:23 the better context it has, the better
8:25 decisions it can make because it has
8:27 that context. Now, for example, one of
8:28 the things I added to this one as a for
8:31 instance was a really clear onboard. I
8:32 actually spent 7 hours on the
8:33 onboarding. One of the cool things that
8:35 this does is we'll go through and
8:37 actually detect all the softwares on
8:38 your computer. Now, if you're deciding
8:40 to build this one from scratch, you
8:41 don't need to add an onboarding, but in
8:43 reality, if you've got clients you want
8:44 to bring on board, I would recommend
8:46 having some onboarding just to make that
8:48 easy so it knows the right locations to
8:49 grab everything. It knows what it's
8:51 looking at. Um they can add in your API
8:53 keys. Um you This is a really important
8:55 one because what we want to do is
8:57 understand an ROI on your time, so you
8:59 know exactly how much time and money
9:01 you're saving with skills. And again,
9:02 this is an overview of the dream
9:03 sequence when you want to far off, etc.,
9:06 etc. And so now you know exactly what an
9:08 agentic operating system is and how it
9:10 works. The next thing we're going to do
9:11 is build something together on this that
9:13 is going to solve a problem that you
9:15 probably didn't even realize that you'd
9:17 had. And that is the fact that we often
9:19 build and create beautiful things with
9:21 Hermes agent or Claude code, but
9:23 invariably, we lose it. We say, "Hey,
9:25 build me, for example, um an overview of
9:27 this or build me a plan for this." But
9:29 it's like it's it's what we call the
9:31 artifact, and they are temporary in
9:33 nature, but there's no way to actually
9:34 look and store them in one location.
9:36 Now, I'm going to talk about artifacts
9:38 in this kind of document visualization
9:39 system, which is so freaking cool, But
9:42 but this will apply to anything you want
9:43 to build. You want to build a trending
9:45 signal report. For example, I built this
9:47 in my community, which is a trending
9:48 Instagram signal report that just shows
9:50 you the trending outlier content that
9:53 exists for Instagram videos, right?
9:54 Let's say I want to build this. It's the
9:56 exact same process. It doesn't actually
9:57 make any difference in terms of the way
9:59 that we're building it out. So, here's
10:00 the thing. What we want to do is take
10:03 the artifacts in this example that live
10:05 in the chat. Um what I mean basically
10:06 want the ability to see it now in our
10:08 beautiful Claude Code operating system.
10:10 So, I would like the ability, for
10:12 example, in the Hermes section here to
10:14 see this wonderful update. So, I can
10:16 say, "Hey, I would like a document
10:17 management section." So, if I build a
10:20 beautiful overview deck, if I build a
10:22 presentation, if I build an invoice, I
10:24 want the ability to seal that sort of
10:25 stuff there, and you'll see my exact
10:27 process for how I built that out.
10:28 Because the issue is it's like we've got
10:30 a brilliant assistant, but she's got
10:31 nowhere to store it, guys. She's got no
10:33 memory on the computer, no space in the
10:35 filing cabinet, which is a big issue.
10:36 And there's a way that we can do this to
10:39 actually save you a ton of tokens with
10:40 Hermes. So, for example, the issue we
10:42 get when we give it a new repo, say for
10:44 example, I'm giving it this to work
10:46 with, right? It has to manually
10:47 understand so much about the
10:49 interdependencies, the inter inter
10:51 basically how it all connects together,
10:52 right? Now, think about it from this
10:54 point of view. What we can do is use a
10:57 new repo called Code Graph, which is a
10:59 brand new system, okay? And so, the idea
11:01 with this is that before Hermes starts
11:02 building, we want to basically create a
11:05 graph of the dashboard that we're
11:06 building together so it can easily
11:08 navigate through it. And the idea here
11:09 is that Hermes is not going to have to
11:11 reread the entire blueprint every time
11:12 we want to go ahead and use it. It is
11:15 82% cheaper, you'll use 86% fewer
11:17 tokens, which is crazy. So, let me show
11:19 you exactly what this looks like, and we
11:20 can go ahead and build it. So, this is
11:21 the repo. I'll put a link for you down
11:22 below. All I'm going to do is come down
11:24 and copy this link right here. And by
11:25 the way, if this sounds like I'm
11:26 speaking Italian, I'll put a link for
11:28 this down below, which is a full Claude
11:29 Code course that will take you from
11:31 beginners to power features, memory
11:33 systems, Hermes apps, building anything
11:35 websites. It will get you so far up to
11:37 speed. I always says I get so many
11:39 messages on this course. It is the best
11:41 piece of content I've ever created. I'll
11:42 put a link for it down below in the
11:44 community if you want to check that out.
11:46 That will get you flying. Now, once
11:47 we've copied this link, we're going to
11:48 shoot up Hermes and we're basically say
11:50 this, "Hey there, I'd like to go ahead
11:53 and install this GitHub and clone it,
11:55 please." Now, as we know, GitHub is just
11:58 fancy speak for place we store files.
11:59 Now, what we're going to do once we've
12:00 done this is we're going to point it
12:03 towards our existing dashboard. I'd
12:04 recommend that you use something like
12:06 this whenever you're dealing with a big
12:08 project. If it's kind of small, like you
12:10 know, a couple pages, not too many
12:11 complexities, not too much
12:13 inter-interconnectivity, you don't need
12:14 to use it. But when it gets bigger, it's
12:16 going to save you so many tokens. So,
12:17 the first thing I'm going to do is just
12:18 point it towards where I'm coding. So,
12:20 I'm like, "Hey there, I'd like you to
12:23 use Code Graph to understand this
12:26 project on my computer for my Hermes and
12:28 Code Code Operating System. Once you
12:29 understand that, just let me know." And
12:30 then I'm going to come down I'm just
12:32 going to paste and the actual um link of
12:33 where that is. And again, if you don't
12:34 know, and you've been building this in
12:36 Code, you can say, "Hey Code, where is
12:37 it?" Or just copy the link from your
12:39 computer. Beautiful. So, that was done.
12:40 Now, we begin of the exact process I
12:41 already use when building this out. So,
12:43 let's go ahead and go on hands-free mode
12:45 with Glider. Awesome. So, what I would
12:47 like you to build for me is a document
12:49 interface. So, the problem I'm trying to
12:51 solve in is when we build any documents,
12:54 invoices, HTML overviews, it exists on
12:56 the chat. Maybe it exists on a laptop,
12:58 but there's no visual way for me to see
12:59 this in our dashboard right now. So, in
13:02 the dashboard, specifically, um
13:04 underneath, I would like it to go
13:07 underneath the skills, but above the
13:09 commands, what I would like to see,
13:12 basically, is a interactive area where I
13:14 can see all of the documents. Let's just
13:15 have it going from left to right, maybe
13:17 four to five in a column, something
13:19 beautiful. I can click on it and I can
13:21 visually see it there. Um and this would
13:23 be anything that we build together. So,
13:24 one of the ways I think we may want to
13:26 do this is have a desktop file where
13:28 whenever you create something, you have
13:30 a Hermes folder. And then, whatever's in
13:33 that folder is reflected in the
13:34 dashboard. And from that, I should be
13:36 able to do any of the actions that I
13:38 would like. Okay, really simple. Now,
13:39 what I also like to do is this, which is
13:41 before you get started, confirm that you
13:43 understand my stated intention, and ask
13:45 me any clarification questions. And
13:46 then, what I'd like to do once we're
13:48 past that part of the process, is to
13:51 spin up a mock-up in a localhost for me
13:52 to have a look at, and then we and then
13:54 we can begin the build and finalize it.
13:55 And just like that, Hermes asking some
13:57 clarification questions, so let's come
13:59 and answer it. So, document types, yes,
14:01 it should be let's do images, let's do
14:03 any kind of document you want to. The
14:05 key thing for me is that it's easy to
14:06 navigate. I should be able to filter by
14:08 type. I want it to be gorgeous and
14:09 interactive. I might give you some
14:11 assets for that later, but I I should
14:13 encompass every document type. Folder
14:15 location, whatever you think to, as long
14:16 as you just save it and it's uniform and
14:18 it fits within Hermes, that is
14:20 fantastic. Preview versus action, yes, I
14:23 would like a visual preview if that's
14:24 possible. I just like to think just
14:25 think about how we can make it as
14:28 user-friendly as humanly possible to do
14:29 that. And then, in terms of live
14:31 updates, yes, the second you create
14:32 anything, I would like it reflected in
14:34 my dashboard. I'd also like the ability
14:37 to remove it there and delete it if
14:38 required. And in terms of existing
14:40 patterns, I would like you to follow the
14:41 existing design structure that I have
14:42 already created within my dashboard. And
14:44 once we've done that, guys, we send it
14:45 off and let Hermes work its magic in the
14:47 background. So, you see the point about
14:48 it coming back and clarifying its
14:50 understanding is super important. Now,
14:52 I'm doing this for this feature, but you
14:54 can imagine it works for essentially
14:55 anything you want to create inside your
14:57 own Agentic operating system. And then,
14:58 guys, after you've done all the back and
15:00 forth, I've built it and I've refreshed
15:01 the page, you explain what you want, and
15:03 then we can check out the visual
15:06 hierarchy, and look at this. And I had
15:08 it create the images for me as well and
15:09 do all this sort of stuff. And so, this
15:11 is what we've gotten with this beautiful
15:13 document set here. So, I've got text
15:15 contacts, I can click into it, I can see
15:16 what it's about, I can see different
15:19 ideas, I can filter by code, data, HTML,
15:21 markdown, or just text if I want to,
15:23 which is cool. I've got the HTML here,
15:24 so if I want to open that up, I can
15:26 click on this. This could be an invoice.
15:27 It's a fake invoice, I think I'm billing
15:28 people for office hours, apparently.
15:30 Pretty crazy, who knows?
15:31 Then we can come back over. You've also
15:33 got this delete function, which is cool,
15:35 and we can also, if we want to, search.
15:36 So I can come down here and go for
15:39 weekly notes, and it's there. Now,
15:40 question is, well, how do we sync this
15:42 up? So say for example, you grab this
15:43 dash one. So this one's in my community.
15:45 Let's say you wanted to sync this up.
15:46 What would we do? All we do is come down
15:48 to install prompt, like so. And you can
15:50 see you've got everything here. I just
15:52 come down, this explains everything I
15:53 wanted to do, basically. The types, the
15:55 data, it says all visible. And what
15:56 we're asking Hermes to do, and I
15:58 recommend that you do them, if you're
16:00 building this, is get it to give you a
16:02 five-word overview of what the thing is,
16:04 and then maybe like some other logic
16:07 that says add no more than 13 to 14
16:09 words underneath here. So you can at a
16:12 glance see exactly what that's about. So
16:13 for example, we're going to go ahead and
16:14 install this together. I go to install
16:16 prompt. I just copy the prompt, and then
16:17 I bring it over to Hermes. So I'm
16:19 Hermes, I'm going to come down and just
16:21 throw this by in, and let it go. And
16:23 this will explain everything for me,
16:24 which is super cool. So this is now
16:26 saying it's become the full location for
16:27 saving stuff, which is cool. So let's
16:29 give her an example. Let's say that
16:31 we're building an invoice to Who do we
16:33 want to work with? Let's say we want to
16:34 work with, I don't know, Dana White,
16:36 UFC, I'll say. Hey dude, write me an
16:38 invoice to Dana White, go find out UFC's
16:42 address, and make it for $50,000
16:44 um for being super duper cool with a
16:46 beautiful right hook, or whatever the
16:47 thing is. We're going to send that one
16:49 off. All right? And basically,
16:51 um this is just going to be a beautiful
16:53 HTML invoice. Uh I'm also going to do
16:55 very quickly is {forward slash} steer,
16:57 right? Which means it won't kick it off
16:59 into say, "For this one, just use some
17:02 fake details in terms of like my address
17:04 and all that core information. This is
17:06 just for test purposes, all right?" And
17:07 then just drop that one off. And it's a
17:09 nice little We're going to steer it in
17:11 one direction without interrupting its
17:12 flow. Of course, you can also if you
17:15 want to do background tasks tasks is for
17:17 separate tasks that aren't to do with
17:18 the initial thing that we're asking.
17:19 Now, this works properly. When we load
17:21 up a dashboard, we'll be able to see
17:23 that document. And I just think this is
17:24 so handy to have. Like, you're not going
17:26 to lose that stuff. I also, when I wrote
17:27 this, I'd encourage you to do this as
17:29 well, is what you're not including. So,
17:31 for example, if you're working on a Git
17:33 repo, right? We don't want to include
17:34 that necessarily, right? Because it's
17:36 not really a document that we're
17:38 creating. And we've also got the ability
17:39 to remove them. So, it's saying that the
17:41 document has been created. Well, only
17:42 one way to find out. Let's click on
17:44 this. Full screen this bad boy. Come
17:46 over to the Hermes Agent. And then let's
17:47 go down to the bottom and see it. This
17:49 is the very first I'm testing this live
17:50 with you right now. I'm going to come
17:52 down here. So, this is the Aurora Labs
17:53 invoice, but the one up here is Dana
17:55 White. So, let's open this one up and
17:56 have a quick look. There we go. Again, I
17:59 didn't give it any specifics, but you
18:00 get the idea now. I actually have to do
18:01 this. I actually have an invoice
18:03 template I tend to use, which is really
18:04 cool. But again, this could be done in
18:06 any way. And again, now if I come down
18:08 and I refresh this, as a for instance, I
18:10 can all the way to the bottom, I can now
18:12 view basically all of my invoices and
18:13 documents that are created in Hermes.
18:15 And this is just one example. But if you
18:17 want to grab this entire thing, it's
18:18 down below in the school community. I'll
18:20 put a link for you. The price is going
18:22 up in the group on the 13th of June,
18:23 just FYI. It's literally one-click
18:25 download, go through the onboarding, and
18:27 you will be flying. I'll also include
18:28 that if you want to build this yourself
18:30 as well, this installation prompt, so
18:31 you can just literally feed this to your
18:33 Hermes Agent. And so, we figured out how
18:35 to build anything with an Arrogant
18:36 Operating System. The next thing you
18:38 need to do is understand all of the
18:41 hidden capabilities in Hermes Agent,
18:42 which you may not know about. And we're
18:43 going to do that by watching this video
