0:02 I have spent hundreds of hours in Hermes
0:04 agent building, experimenting, and I've
0:06 seen every single masterclass. And in
0:07 this video, I'm going to take you
0:10 through the seven levels of Hermes agent
0:12 step by step. So, you know what to
0:14 expect at each level, the exact mistakes
0:16 that you need to avoid, and what you
0:18 need to do to progress to the next level
0:19 so you can save more time, make more
0:21 money, and get light years ahead of
0:23 everybody else. And if you're new, I'm
0:24 Jack. I built and sold my electric
0:26 startup with a gazillion customers. Now,
0:28 I'm building my own AI companies, and I
0:30 share here the stuff that actually
0:32 works. So, if you haven't already, grab
0:33 that beautiful coffee,
0:35 and let's dive straight in. Now, there
0:37 are seven levels to this Hermes agent,
0:38 and if you make it towards the end of
0:39 the video, they're going to be saying,
0:41 "Hey, who is this super intelligent
0:43 Hermes individual I see before me?"
0:44 That's what they're going to be saying
0:45 about you when you make it to the end of
0:46 this video. Now, we're going to go
0:48 through every level of Hermes from one
0:51 prompt to an agent that literally does
0:53 wonderful and fantastical things and a
0:55 lot of mistakes. Now, here's the thing.
0:57 What level zero is where everybody's at
1:00 cuz just for context, nobody's using the
1:02 you're seeing more of this content cuz
1:04 you watch it, but just like the context,
1:07 barely anybody is using AI, let alone
1:08 Hermes agent, so you're already ahead.
1:10 Now, let's begin with level one, which
1:12 is the beginner. This is basically the
1:14 download and go individual. Essentially,
1:16 what's happened here is you understood
1:19 that it's an agent and not a chatbot. We
1:20 know that chatbots you can have
1:22 conversations with, like Claude and
1:24 ChatGPT. Hermes is something that
1:26 actually goes ahead and does things for
1:28 us. And to reach level one, what you
1:29 needed to have done is actually gone
1:32 ahead to basically the website, grab the
1:33 information, and download Hermes onto
1:35 your computer. If this sounds like I'm
1:36 speaking German, click this link on
1:37 screen. I'll show you the full process
1:39 of getting it set up, but effectively,
1:40 you've gone and you've installed it on
1:42 your computer. Now, you've installed it
1:43 on your computer, and you're connected
1:45 to Telegram or your messaging platform
1:47 of choice. Now, at level one, you're
1:49 using it for simple tasks. So, you might
1:51 say something like, "Hey there, I am
1:52 going from Dubai to a random country in
1:55 Europe for 3 months. Do some research,
1:57 write me a list of the best countries to
1:59 go to with some insights, and just save
2:01 that HTML file on my desktop so I can
2:03 check it out for me later when I come
2:05 back. Send that off, and literally
2:07 Hermes agent is going to do this for us,
2:08 and just save things for us to check
2:10 whenever we want to. And as you can see,
2:12 Hermes agent completes things for us. It
2:14 uses tools, it's got the chat. I can
2:16 just do anything we want to in those
2:18 current environments, and now I can even
2:19 open this up and see its output. And as
2:21 you can see, I open up and it's giving
2:22 me a nice little breakdown of all the
2:24 things that I want to do. Key difference
2:26 here is that you're using Hermes agent
2:28 for anything and also tool calling to
2:29 build interesting things. Now, the
2:31 biggest drawback at level one is you're
2:34 just using it for long one-shot prompt
2:35 things. You need to make sure that
2:37 you're collaborating with Hermes agent.
2:39 Ask it to understand your stated
2:41 intention. Go back and forward and just
2:43 be mindful of your context window. You
2:45 could probably expect to save about 30
2:46 minutes a day when you're using it in
2:48 this type of fashion. And this brings us
2:50 on to the second level of Hermes agent,
2:52 and one of its strongest features is
2:55 that if you don't use it, you are not
2:57 utilizing Hermes properly. Now, with the
2:59 apprentice at this level, effectively
3:01 Hermes is getting to know you as an
3:03 individual. Now, the key thing about
3:05 Hermes agent is it the more that it
3:07 knows you, the better it actually gets.
3:08 And I can show you here, for example,
3:10 this memory system. The more informa-
3:12 the more you speak to Hermes agent, the
3:13 more information that you give it, the
3:15 higher performance it will actually get.
3:16 But what do I mean by that? Let me
3:18 explain. So, Hermes itself has a very
3:20 cool memory system. For example, you can
3:22 say to Hermes agent, "Hey, what was that
3:24 thing Could you just tell me one thing
3:26 that we spoke about 2 days ago, please?"
3:28 You can actually get Hermes agent to
3:30 recall specifically things that happened
3:32 on days. Now, the cool thing is Hermes
3:35 will commit to its own internal memory
3:36 the more that you actually talked about
3:38 it. But it isn't even just about its
3:40 memory on a service level, cuz this is
3:43 just basic at level two. It's also its
3:45 ability here to know who you are and
3:48 tailor its specific advice based on what
3:49 you're trying to actually get. And as
3:51 you can see, it's talking about a llama,
3:52 which is a video that came out a couple
3:54 of days ago on my channel. Now, you are
3:57 missing out on Hermes magic unless it
3:58 knows about you. So, what you want to do
4:00 effectively is use what we call a Hermes
4:03 soul.md. What do I mean by that? Well,
4:04 if I show you here for example on the
4:05 LS, if I come down here, you can
4:08 actually categorize your user profile
4:10 and actually physically give it a soul.
4:11 This is stuff like, how do I want you to
4:14 address me? Information about you and
4:15 basically your business or whatever it
4:16 is that you actually are physically
4:18 doing. And I'll put a link down below
4:20 for the soul.md Hermes template that
4:21 takes everything, mission goals, your
4:23 business, key metrics, all that sort of
4:25 stuff. The beauty the beautiful part of
4:27 Hermes, okay, is that if two people
4:28 speak to Hermes and ask it the same
4:30 question, they should get radically
4:32 different answers because it understands
4:35 context. It's all about context
4:36 engineering. It needs to know about you
4:38 and your information, your data so it
4:39 can best help you. And that's one of the
4:42 biggest differentiator is that we start
4:45 to see at level two on our seven-level
4:46 scale. That's why we call it the
4:47 apprentice. And the key point here is
4:50 that Hermes never starts from zero. It's
4:52 the agent that remembers your birthday.
4:53 Its memory is one of its most coolest
4:55 features. And it should And it should
4:56 know that stuff about you. So, at this
4:58 level, we should be looking at roughly 5
5:00 hours a week that we're claiming back.
5:02 Key thing is if it knows you, you can
5:04 stop carrying all the information in
5:06 your head. And remember this, when we're
5:07 thinking about how much time it saves
5:09 you, there's such a thing as context
5:11 switching, right? You It isn't just how
5:12 much time did I spend doing something?
5:14 You lose things when you switch between
5:15 tasks. So, having one of the world's
5:17 most intelligent personal AI assistants
5:19 is an unbelievable game changer and why
5:21 it's blowing up so much and it's going
5:22 to be such a huge thing in the future.
5:23 Key details, it remembers your
5:26 preferences, it can recall past chats,
5:27 and there's no need to re-explain. Now,
5:29 custom instructions are one thing, but
5:32 if you don't use skills and commands
5:34 properly in level three, you're not
5:36 utilizing its full potential. What do I
5:37 physically mean by that? Well, we're
5:40 looking at two key things here, okay?
5:41 First thing I want to touch on here is
5:43 some of the key commands and controls
5:44 that will radically change your
5:46 experiences as home with Hermes. At this
5:48 level, you're starting to basically
5:50 really utilize its functionality and how
5:52 cool it freaking is. Let me show you
5:54 exactly what I mean. First of all, I
5:55 might give it a prompt that sounds
5:56 something like this. Hey there, I would
5:58 like to go online and find for me what
6:01 are what is the best advice to scale my
6:02 LinkedIn and acquire clients in that
6:04 channel. Do some deep research for me
6:06 and find that out. Okay. So, you can
6:08 give Hermes a prompt like this. We could
6:09 be in the gym, we could be anywhere.
6:10 Now, crucially, I can do a couple of
6:13 things. I can do {forward slash}, I can
6:16 do steer. And steer basically will guide
6:19 it in one direction. I specifically want
6:21 you to look at top creators on LinkedIn
6:24 who are already doing very well and also
6:25 YouTube. Now, what this means is I'm
6:28 kind of like it's it's going to use this
6:30 and think about this prompt before it's
6:31 even completed. Now, and it's a great
6:33 little hack. Another one I can do is
6:35 background, which basically means whilst
6:37 it's doing that task, fire this one off
6:39 in the background. So, I could say, "Hey
6:40 there, I'd like to go ahead and do some
6:42 research on the best way for me to grow
6:44 out my Instagram. Specifically, go and
6:45 find me great Instagram creators and
6:47 build me a beautiful list of ways that I
6:49 can improve my Instagram." Now,
6:51 background tasks themselves happen in
6:53 the background and it doesn't even
6:54 interact with the core thing that Hermes
6:56 agent is doing. So, the idea here is
6:58 that we're using more of the features
6:59 that exist within Hermes. We've got
7:01 handoff, we've got clear to basically
7:03 the functions. And again, there's lots
7:05 of different things you can do. Now, the
7:07 other really interesting one here is the
7:10 model. Now, at this level, at level
7:12 three, we're basically using what I call
7:14 model agnostic behavior. What the hell
7:16 does that mean, Jack? What it basically
7:19 means is that we tag in the best model
7:21 for the job. That's simple. In fact, I
7:23 can even show you with the Hermes app.
7:25 If I pull this up and even if you see
7:26 here, for example, in the operating
7:28 system for Hermes, if I come over here,
7:30 you can see the model that we currently
7:32 have tagged in. In my previous video, I
7:33 also showed you how you can run this 100%
7:35 100%
7:36 privately, which is crazy. Now, the
7:38 point here is that we can see the model
7:39 we're dealing with. We're talking to
7:41 ChatGPT 5.5 and if I have a
7:42 conversation, I've got that. Freaking
7:44 awesome. So, for example, if you can
7:45 now, you can see the different models I
7:46 have within Hermes Agent right.
7:49 Anthropic, GitHub, OpenAI Codex, Open
7:51 Router, etc. So, I speak to people all
7:53 the time about Hermes Agent. Now, some
7:55 say, "Jack, my credits credits went very
7:57 quickly." Your core model, okay, think
7:59 of it like this. You want to connect
8:01 your ChatGPT subscription to Hermes
8:04 Agent. Super duper important. That lets
8:06 you use your $20 ChatGPT subscription
8:09 for anything. It is a coding powerhouse.
8:12 Now, the most intelligent model is Opus
8:14 4.8. But, the reality is you don't need
8:15 it for everything. We don't need
8:17 astrophysicists, you know, basically
8:19 building Legos for us. But, it's not
8:23 required. Sonnet 4.6 is a very good
8:24 middle ground of like really good
8:26 performance, but it's not going to burn
8:28 your cost. These are API costs. The
8:30 other model you need to have in here as
8:33 well is Grok. Grok also connects via
8:35 OAuth, which means open authentication.
8:36 You literally
8:38 connect it, you sign in, and you're
8:40 basically using Grok within the engine.
8:42 Now, the cool thing about Grok, again,
8:43 it's on your subscription and you get
8:46 access to the entirety of X. So, you can
8:48 say, "Hey, what's happening this week
8:49 with websites?" And it'll tell you all
8:51 the top fine posts. It's freaking
8:52 amazing. Then, we want to tie in the
8:54 right models. You want to make sure that
8:56 you've got Open Router as well. Open
8:58 Router is a very important piece of
8:59 software that effectively lets us
9:01 connect to anything in the world that we
9:03 want to. Why do I like it? Well, let me
9:04 show you this. If I come over to
9:06 rankings, you can see we have access to
9:08 every model. You can access free models.
9:10 You can access any model that you wish.
9:12 It's really freaking impressive. This is
9:15 how I personally access 4.8
9:17 with Claude. And essentially, we can
9:19 ratchet it down whenever we want to. We
9:21 can even use DeepSeek V4 Flash, which is
9:23 fantastic. So, we get to basically
9:26 ratchet it up based on performance and
9:28 cost and also privacy. So, we get to
9:30 move those dials up to get the right
9:32 models we want to. And at level three of
9:34 the system, guys, what we're doing is
9:36 leveraging the right model for the right
9:39 task. This is also the level where we
9:41 begin to introduce skills. Now, when it
9:43 comes to skills, what I think is really
9:45 important is the skills are only as good
9:47 as the model doing the skill and what
9:49 the skill actually is. So, for example,
9:51 in the Pantheon here, I've got in the
9:53 operating system. One of the things that
9:54 I'm very specific about when I add my
9:56 skills is specifying three things. So,
9:57 let's go with the alchemist in this
9:59 case, right? I specify what the job is,
10:01 a description of that job, and then a
10:03 system prompt, which is completely
10:05 optional. And then crucially, I come
10:07 down and I actually outline what the
10:10 model is. Why? Because I don't want 4.8,
10:12 which is really expensive, doing
10:13 something that we could delegate to a
10:15 free model or something on Deep Seek.
10:16 And of course, we can connect this
10:18 entire thing. Then if I want to change
10:19 my prompts, I can visually come and see
10:22 them and make adjustments as I want it
10:24 as I come and go. Now, why is this so
10:26 important? This really, if you're down
10:28 to it at level three, are some of the
10:29 biggest time saves that you're going to
10:31 get because it can do things like go and
10:33 do my competitive research, go and find
10:35 for me all the outliers that exist on
10:36 Instagram. And you can build software
10:38 that looks like this. And this, for
10:39 example, if you're trying to grow on
10:40 Instagram as a for instance, shows you
10:42 all of the outliers. This I scraped this
10:44 from my community. Why is this so cool?
10:45 Well, this would be something that you
10:46 could have spent hours doing. And this
10:48 is why this can save you up to a day a
10:51 week cuz you can now delegate tasks and
10:53 automations and research to Hermes and
10:55 it will execute for you perfectly. Now,
10:56 this sounds like I'm actually speaking
10:58 Portuguese. I'm going to put a link down
10:59 below for this full Claude Claude
11:00 course. I'll take you from foundation
11:03 and setups, building websites, power
11:05 features, memory systems, Hermes cycle,
11:07 stuff I have never covered on YouTube
11:08 before. I'm I'm going to put all down
11:09 below. People message me about this
11:11 course all the freaking time. You also
11:13 get complete immediate access to this
11:15 entire operating system. I'll put a link
11:16 down below so you can check that and
11:18 upgrade if that's valuable to you. And
11:20 just as a FYI, by the way, our price on
11:22 community is popping up on the 13th. So,
11:23 if you see it ahead of time, we'd love
11:25 to welcome you. So, we've covered this
11:28 in level three and it does lead us very
11:31 nicely onto the fourth level of Hermes,
11:32 which is the integrator. Now, in the
11:35 integrator section, this is where really
11:37 Hermes goes to a completely new level.
11:38 And if you look at this graphic as a for
11:40 instance, okay. What we're really doing
11:42 is connecting it and giving it the
11:45 ability to connect to your world because
11:46 the more things that you touch and
11:48 interact with that we can plug into
11:50 Hermes, the more it can really do. A lot
11:51 of the stuff that we've covered so far
11:53 is you speaking to Hermes and Hermes
11:54 doing things like, "Create me a report
11:57 that does X." or "Go and research Y." In
11:59 level four, we really unlock the new
12:01 capabilities when we give it access to
12:03 things like reading our emails, things
12:04 like creating calendar appointments with
12:05 us and our friends and our our
12:07 colleagues and potential business
12:09 partners, looking at Notion for like
12:10 things like our to-do lists, plugging
12:12 into Slack so I can get immediate
12:14 updates. For example, maybe you're on
12:15 ClickUp, right? One of the things that
12:17 we could do with ClickUp is have our
12:18 team, this is one of the things that I
12:20 implemented, basically giving updates on
12:21 that. And now basically when I want to
12:23 know what's going on in my business, I
12:25 can say, "Hey Hermes, what were the
12:26 numbers this week?" And the really cool
12:28 thing, and I really stress this upon you
12:30 again, is that you give it text, but
12:32 also numbers. So, if I'm saying, "Hey
12:34 Hermes, I'm thinking about this from a
12:36 strategic point of view, what are your
12:38 thoughts on this strategy?" And one of
12:39 the things my Hermes agent will know to
12:41 do, depending upon what I'm talking
12:42 about, is he can go to different
12:43 databases. Say you're connected on
12:45 ClickUp or whatever you are, it can go
12:47 through the latest updates that you had
12:48 that week from your team. And so, it
12:50 will actually get Did you know that
12:52 you know, Kevin or I don't know, Sarah M
12:54 or whatever actually submitted a report
12:55 that unlocked this new thing that's
12:57 really relevant for you to know. So, the
12:59 idea is you don't have to rely on your
13:00 memory You shouldn't have to rely on
13:02 your memory. Plugging Hermes into your
13:05 world will create new optionality for
13:06 you. And this is a key thing that
13:08 everybody's missing. How do I actually
13:09 plug in new things? Well, it's so
13:11 unbelievably straightforward. Let me
13:12 show you what I mean. So, say we have
13:13 our Hermes agent here, right? And again,
13:15 it's even working now in the background
13:16 answering our things before. I'm going
13:17 to come down here. I'm going to do a
13:19 forward slash and we're going to stop,
13:20 which is going to kill all the
13:22 background processes and that'll be the
13:23 end of it. But let's say I want to
13:25 connect some. So, one of the big ones I
13:26 I to connect to is Granola. They're not
13:28 sponsoring this video or anything, but
13:30 the point with Granola is they will let
13:31 you chat to any of your meeting
13:33 histories. For example, I could say,
13:35 "Hey dude, what was the title of the
13:37 last meeting I had and can you tell me
13:39 one action I took from it, please?" Now,
13:40 when you connect anything to Hermes
13:42 Agent, sometimes you'll just store the
13:44 API key on your desktop. Other times,
13:47 you can literally log in by via an MCP,
13:48 which is freaking awesome. And bear in
13:49 mind, guys, this is one of the cool
13:51 benefits of having this on your desktop.
13:53 Like, it's already got all of your
13:54 sign-ins and credentials. So, you don't
13:56 need to duplicate this. I mean, I'm not
13:57 getting paid tens of thousands of
14:00 dollars to sell you some kind of VPS. I
14:01 tell you exactly what I do in these
14:03 situations, which is run it locally on
14:05 my computer. I don't really see the
14:07 point of value in running it on a VPS,
14:09 to be honest with you. So, it's just
14:10 really cool and this makes it a lot
14:12 easier to do things like this. So, we
14:13 would just come down and we click on
14:14 allow access in this case. And as you
14:16 can see down here, one action you took
14:18 away from AI Automation Cafe was to
14:19 collaborate with someone in the workshop
14:21 regarding an iOS knowledge graph. There
14:22 you go, you can see exactly how that's
14:24 connected and we even set the one up
14:25 together. And the other thing to bear in
14:26 mind is if you're ever not sure how to
14:27 connect something, just say, "Hey
14:29 Hermes, I want to give you the ability
14:30 to do this thing. What do you need from
14:32 me to do it?" And just follow through
14:34 the instructions. One little side hack
14:36 for you, don't ever paste those API keys
14:38 directly in Hermes. You can save them on
14:40 your desktop or it'll give you a
14:42 terminal command. You can enter that in
14:44 and it will throw basically explain
14:46 everything to you step by step. Now,
14:47 where most people get stuck at this
14:50 level is they basically give it the MCP
14:51 candy store. We give you Imagine giving
14:53 a five-year-old all the candy in the
14:54 world in a candy shop. It would go
14:56 freaking and ballistic. It's the same
14:58 thing here. Now, the more MCPs that we
15:00 give to Hermes Agent, the bigger its
15:03 context window actually grows. Now,
15:04 there are cool tools we can do like lazy
15:06 loading and stuff, but fundamentally, we
15:09 don't want to give it loads of garbage
15:11 that we're not physically using, so its
15:12 context window doesn't balloon
15:14 significantly. That's just one watch
15:15 out. And the idea is the fewer number of
15:17 keys we can give it, the better. Now, a
15:18 couple of things I really want to bear
15:20 in mind with this. First of all, emails,
15:21 I personally do not give it the ability
15:24 to send emails. It can draft them. It
15:25 can add things to my calendar, but it
15:27 can't send emails just yet. So, I kind
15:30 of like to keep that one eternally
15:31 rather than sending it out. At some
15:32 point, it will be fine, but at the
15:34 moment, I just like to keep that
15:36 requirement to myself. And so, really
15:38 level four is about crushing your admin.
15:41 It should be able to do anything you
15:42 need it to do. Anything that you would
15:44 personally do on a laptop, Hermes should
15:46 be able to do that for you. And if it
15:48 can't, just ask yourself, why doesn't it
15:50 have that access? And you just connect
15:51 it to Hermes. It's a really good
15:53 exercise to do and will save you so much
15:55 time. Now, level five is when it starts
15:57 to get very interesting. We call this
15:59 level the orchestration level.
16:01 Effectively, where Hermes becomes not
16:04 only an executing agent, but a manager
16:05 that manages other agents to do
16:07 different tasks. It can, in fact, spin
16:11 up multiple agents. And every single
16:13 update that you see with Claude Opus 4.8
16:14 can now do this cool thing. It can do
16:16 ultra work. Hermes can do the exact same
16:17 thing. Remember, someone asked me
16:20 recently, Jack, what's better, Claude or
16:22 Hermes? It's like asking me, what is
16:24 better, Lewis Hamilton or Ferrari?
16:26 They're different, right? Hermes is the
16:27 If you think of Hermes is more kind of
16:29 like the the actual agent system, and
16:31 then the actual driver, the race car
16:33 drivers are things like Claude and
16:35 ChatGPT. This is more of a system that
16:36 surrounds it. Now, the key thing here is
16:39 that Hermes itself can use Opus 4.8 and
16:41 all of these individual models to do
16:43 many different things at the exact same
16:44 time. And we just mentioned, didn't we,
16:46 about the Pantheon? This idea that we
16:47 can actually go ahead, if I bring up the
16:49 right actual tab here, and show you
16:51 this. All these different systems, these
16:52 different skills that we want to build
16:54 out to do various different things. But
16:55 what is really cool about Hermes agent
16:58 is that we can actually ask it to do
16:59 multiple things at the same time. Let me
17:01 show you what I mean. And this is really
17:03 where the sub agents come to play. So, I
17:05 can say, "Hey there, I would like you to
17:09 create for me a team of agents. I would
17:11 like one of the sub agents to be, call
17:13 it ChatGPT 5.5. I want to use DeepSeek
17:16 V4. And I'd like to task DeepSeek V4
17:18 with going and doing some comprehensive
17:22 research on the best industries to sell
17:25 websites to. I want GPT-4, excuse me,
17:27 GPT-5.5 to critique him, and I want him
17:28 to go around in a loop until they've
17:30 solved the issue. And then I would like
17:32 to have Opus 4.8 as an orchestrative
17:34 agent that will oversee it like his
17:36 team, spin up as many sub agents as you
17:38 see fit, but a lot of the work I want
17:41 done by the very cheap DeepSeek V4. And
17:42 as you can see, it's actually going
17:44 ahead and doing that for us. So, we have
17:47 multiple agents in Hermes working in
17:48 parallel. Now, this is the reason why
17:50 this is even better that in some
17:52 respects than Opus 4.8 is mainly because
17:54 we're not just using Claude here. Now,
17:56 obviously, you can connect these things
17:57 to Claude, which is fine, but again, I
18:00 get to bring in any model I want to to
18:02 do anything that I want. You can do this
18:03 within Claude, but I also think it's one
18:05 of the coolest things that we can do in
18:07 Hermes, which is completely model
18:10 agnostic. We bring in the best model for
18:12 the job, and we have these fantastic
18:14 parallel agents working for us. Now, on
18:16 level five itself, this is where you can
18:19 really explode your output if you know
18:20 what you're applying for. Remember, the
18:23 biggest hallmark basically of of a
18:25 beginner automator is that you automate
18:27 stuff that simply should not exist. The
18:30 idea is this is high-leverage when you
18:32 know the high-leverage thing that you're
18:33 focusing in on, right? We don't just
18:34 talk about AI here. We talk about in the
18:37 context of winning, like in life, like
18:39 in business, in our physical health, in
18:41 our relationships. We want to be the
18:43 best version of ourselves, and that
18:44 means we know the problem that we're
18:46 solving. AI is an accelerant to that
18:48 end. So, the point is you can start to
18:50 think your output if you know what
18:51 you're focusing on and what your
18:52 bottleneck is in your particular
18:55 business or situation. So, it's a real
18:57 explosive feature. But it brings us very
19:00 nicely onto the sixth level, where it
19:01 takes one step further, and this one is
19:04 known as the builder. So, first thing to
19:05 understand about Hermes agent is that it
19:08 does ship real things, anything. I'm
19:09 getting fired at right now. Anything you
19:12 can build in Claude code, you can build
19:14 in Hermes agent. And honestly for me
19:16 guy, this was the coolest unlock. To be
19:19 at the gym, doing my bench press, I used
19:20 to decline, incline, whatever you want
19:23 to do, and being able to actually code
19:25 things and build things is a freaking
19:26 unbelievable. If you're building
19:28 software, the ability to have those
19:30 conversations, spin up the different sub
19:32 agents, and build software whilst you're
19:34 doing anything else you want to is
19:36 literally game changing. It is no longer
19:39 just a chatbot. This is software. You're
19:41 building software. And I'm telling you
19:42 right now, that is the future. Building
19:45 these micro softwares, these micro apps
19:46 is insane. If you have the idea If you
19:49 can imagine it, you can build it. That's
19:51 kind of the point here. Hermes lets you
19:53 do this remotely. And in my view, no
19:55 other app uh in terms of like, you know,
19:57 Codex and Claude really compete with
19:59 Hermes remote ability. You can hack it
20:00 and make it hacky, but then what you
20:02 start to realize is you're basically
20:04 just rebuilding Hermes anyway. So, this
20:06 is why this is so freaking sick. And at
20:08 level six as well, you start to use its
20:10 automation features. And what do I mean
20:11 by that? I mean about asking it to do
20:14 things at certain times. Taking what we
20:16 call asynchronousity, which means that
20:18 most people do this. I answer, you
20:19 respond. You You ask question, I
20:22 respond. What people in level six do, at
20:23 the builder level, is they're actually
20:25 adding scheduled tasks. They're saying,
20:27 "Hey Hermes, um I need to go and have a
20:29 conversation this weekend with, you
20:31 know, I don't know, McKinsey and Co. Um
20:33 do some deep research over the next 24
20:35 hours. And then on Friday afternoon,
20:36 make sure you give me a fresh report on
20:38 my desk. Find the email addresses of all
20:40 the board the members of the board and
20:42 actually draft me an email. Or you say,
20:43 "And also, send me a reminder on
20:45 Thursday evening that I need to go ahead
20:46 and do this thing." And you're basically
20:48 creating these scheduled tasks. And it's
20:50 unbelievably easy to do. For example,
20:52 you just say, "Hey there, I've got a big
20:54 meeting on Sunday. Could you just send
20:56 me a reminder, I don't know, midday at
20:58 12:00 on Saturday Dubai time to do some
21:02 preparation for my meeting with Nvidia.
21:03 I want you to do some deep research
21:05 between now and then. Find out who who
21:06 members of the board are. And then also
21:08 go ahead and find me five interesting
21:10 things to talk about in my meeting.
21:11 Whatever the thing is, I can send that
21:13 off. And I'm just explaining it. I'm
21:15 yapping in English. By the way, when I'm
21:17 not using Galactica, which I'm using for
21:18 my speech dictation faster in the world
21:20 right now, I typically do my voice
21:22 message when I'm talking to Hermes, and
21:25 I just yap and I talk because the speed
21:26 at which you can get information out is
21:28 four or five times faster, which is why
21:29 I speech type. It's like, honestly, the
21:31 biggest productivity hack I've had and
21:33 something I use with Hermes all the
21:34 time. Now, Hermes is setting a reminder
21:35 to remind me of something. I mean, this
21:37 sends me a daily brief every single day.
21:38 It's freaking incredible. And then it's
21:40 also going to do some research and come
21:41 back and make that reality. Look at
21:42 this. Right now, set a wonderful
21:44 reminder research for June 6, you know,
21:46 midday. It's like an executive
21:48 assistant, but it never forgets. And I
21:49 And it's like the most intelligent thing
21:51 in the world. You're only really limited
21:53 by how you're physically applying Hermes
21:56 to a situation. And what level six is
21:57 going to give you is something really
21:59 powerful. And it's this idea I mentioned
22:01 of asynchronicity, that it ships without
22:03 you, it does things without you, and
22:06 your relationship goes from being I say,
22:08 you respond to something completely
22:09 asynchronous, and it does that. So,
22:12 things get built even when you're not
22:13 thinking about them. That's like the
22:16 huge unlock for level six with Hermes.
22:19 And that then takes us to the doorstep
22:21 of the most powerful level of Hermes
22:23 agent. And that is the agentic power
22:25 user. And the idea behind the agentic
22:27 power user is we have one operating
22:30 system for every AI. What do I
22:32 physically mean when I say that? Well, I
22:34 best show you. So, this, for example, is
22:37 the agentic operating system that I use.
22:38 Now, here, obviously, we've got a full
22:40 overview of all the brands that we're
22:41 connected to or by different
22:42 integrations that we have. We can talk
22:45 to Hermes. I've been over the incredible
22:47 functionalities it has, its ability to
22:49 create mid-term goals for your business.
22:51 And it breaks down in steps and gives
22:53 you like eight different steps of what
22:55 Hermes is going to do, what you're going
22:57 to do, and vice versa. The pants and
22:58 look at our different skills. And this
22:59 is where it gets really interesting,
23:00 right? After we've got the memory, we've
23:02 got the Obsidian integration, and we've
23:04 got all our documents, which is crazy.
23:04 If I want to look at for example
23:06 something Hermes made for me, I can look
23:08 at this. There is nothing else to do
23:10 this right now. Like, you're going to
23:11 build PDFs, you're going to build
23:12 documents, you're going to build
23:14 softwares, but they just live in the
23:15 chat or they're just filed somewhere.
23:17 The ability to actually physically see
23:19 in a dashboard all the stuff that Hermes
23:21 does for you is incredible. Like, let's
23:23 say that you want to write an invoice to
23:25 the UFC for your UFC debut. You want to
23:26 do a head kick or something like that.
23:28 Well, okay, it's great that it's in the
23:29 chat and it can send things, but when I
23:31 want to come back and look at it and do
23:32 various different things, I can do that.
23:35 But, the real unlock is connecting your
23:37 world together. Remember, context
23:39 engineering is one of the biggest trends
23:40 that you're going to see. It is
23:42 exploding. And the idea here, for me to
23:44 see my spend across every model, Claude
23:46 Code, Codex, Antigravity, to have the
23:49 ability to have basically Hermes' agent
23:51 dream for me. What do I mean by that? I
23:52 mean, it's going to look at every
23:54 message that I've said to it. It's going
23:56 to look at all the stuff that I've done
23:58 not only in Hermes, but in Claude Code,
24:00 in Antigravity, in Codex, in anything
24:02 else that I'm touching. I'm going to use
24:04 all of that data, all of the skill to
24:07 proactively think about other ways that
24:10 I can improve this. Like, what does Jack
24:11 What is Jack not doing that he should be
24:13 doing? What are the skills that he's not
24:15 using that he should be using? To do
24:17 research and feed that back to you. And
24:19 crucially, the ability and the I've got
24:20 here in the memory section right now to
24:22 see everything from my Obsidian memory
24:24 that connects everything together and
24:26 literally even the Claude Code memory
24:28 and the Hermes memory, the agentic
24:31 operating system users are individuals
24:33 that don't operate in silos. They have
24:36 their entire AI connectivity in one
24:38 location. And it basically means that I
24:40 do coding in Claude Code, cool. I do
24:42 coding in Hermes, but they're no longer
24:44 separate. That gap is completely bridged
24:46 and it unlocks capabilities that people
24:49 don't even realize actually exist. And
24:50 at this level, not only is it running
24:52 without you, it reports and does work
24:53 for you whilst you are actually
24:56 physically sleeping. So, three truths
24:58 you really need to understand. It's not
25:00 a better coder than Claude code. It is
25:01 literally using Claude code. That's
25:02 really important to understand. The
25:04 number here is that you can't really
25:06 have over a thousand agents running
25:09 simultaneously. The agreed number of
25:10 agents that we think can be done does
25:12 vary based on different models, but
25:13 realistically it's not going to be a
25:15 thousand. And the magic itself isn't
25:17 something you need to wait 17 years to
25:18 do. This is something you can build and
25:21 learn in a couple of videos if you just
25:23 apply yourself. Now, understanding where
25:24 you're at and those seven levels and
25:26 where they fall apart, the next thing we
25:28 need to do is embrace the most powerful
25:30 level, and that's when exactly how to
25:32 utilize this agentic operating system,
25:33 which we're going to do together in this
