WEBVTT 00:00:00.000 --> 00:00:02.720 We have a very passive relationship with AI right 00:00:02.720 --> 00:00:05.780 now. Think about your daily routine. You open 00:00:05.780 --> 00:00:08.060 a tab, type a prompt, and wait for an answer. 00:00:08.279 --> 00:00:11.359 Then you take that answer and do the actual work 00:00:11.359 --> 00:00:14.919 yourself. But imagine an AI that doesn't just 00:00:14.919 --> 00:00:17.539 wait around for your questions. Imagine one that 00:00:17.539 --> 00:00:20.039 actually acts. It runs tasks in the background. 00:00:20.179 --> 00:00:22.039 It checks your servers. It messages your phone 00:00:22.039 --> 00:00:24.379 with updates while you sleep. It is a massive 00:00:24.379 --> 00:00:26.899 behavioral shift, honestly. We are moving from 00:00:26.899 --> 00:00:30.129 an AI that just answers to an AI that... takes 00:00:30.129 --> 00:00:32.350 independent action. It's the difference between 00:00:32.350 --> 00:00:36.289 having a smart encyclopedia and, well, having 00:00:36.289 --> 00:00:38.490 an actual employee. Okay, let's unpack this. 00:00:38.750 --> 00:00:41.170 Welcome to the Deep Dive. Today, we were looking 00:00:41.170 --> 00:00:44.609 at a huge stack of GitHub documentation, developer 00:00:44.609 --> 00:00:47.689 blogs, and Reddit threads. All about a project 00:00:47.689 --> 00:00:49.950 called Hermes Agent. Exactly. Our mission is 00:00:49.950 --> 00:00:52.270 to explore what it takes to build a self -hosted 00:00:52.270 --> 00:00:54.750 AI personal assistant. we're going to cover what 00:00:54.750 --> 00:00:56.950 hermes actually is we'll see how it compares 00:00:56.950 --> 00:00:59.369 to heavyweights in the space too right and we 00:00:59.369 --> 00:01:01.789 will break down the five core pillars of its 00:01:01.789 --> 00:01:05.790 brain and crucial for anyone listening we're 00:01:05.790 --> 00:01:07.769 going to look at how to safely set it up yeah 00:01:07.769 --> 00:01:10.390 so it works strictly for you without you know 00:01:10.390 --> 00:01:12.810 breaking your computer exactly so let's start 00:01:12.810 --> 00:01:15.829 with what hermes actually is because the repositories 00:01:15.829 --> 00:01:18.200 are going viral right now They really are. I 00:01:18.200 --> 00:01:20.599 mean, for a long time, we treated AI like a very 00:01:20.599 --> 00:01:23.000 smart search box. You asked a coding question, 00:01:23.219 --> 00:01:25.980 you copied the answer, you pasted it into your 00:01:25.980 --> 00:01:28.060 editor. You were the one moving the data. It 00:01:28.060 --> 00:01:30.540 was useful, but it was highly manual. You were 00:01:30.540 --> 00:01:33.459 still the bottleneck. Exactly. Now, looking at 00:01:33.459 --> 00:01:35.780 the developer community, user expectations are 00:01:35.780 --> 00:01:38.200 fundamentally shifting. People don't want a search 00:01:38.200 --> 00:01:40.819 box anymore. They want a personal operator. They 00:01:40.819 --> 00:01:43.140 want an AI that remembers their preferences. 00:01:43.400 --> 00:01:46.060 They want it to understand their ongoing projects. 00:01:46.599 --> 00:01:48.840 And Hermes isn't just another chat bot you log 00:01:48.840 --> 00:01:52.000 into. It is an open source framework for building 00:01:52.000 --> 00:01:55.579 your own personal assistant. And self -hosted 00:01:55.579 --> 00:01:58.040 is a key term here. Yeah. You don't just log 00:01:58.040 --> 00:02:00.299 into a website. You run it yourself. You can 00:02:00.299 --> 00:02:03.079 put it on a VPS. That is a rented remote computer 00:02:03.079 --> 00:02:05.859 running constantly in the cloud. Right. Or you 00:02:05.859 --> 00:02:07.540 can run it on a Mac Mini sitting on your desk. 00:02:07.700 --> 00:02:10.180 You could even run it on an Android phone using 00:02:10.180 --> 00:02:13.289 an app called Termux. Right. And because it lives 00:02:13.289 --> 00:02:15.990 on your infrastructure, it is always there. It 00:02:15.990 --> 00:02:17.650 doesn't go to sleep when you close a browser 00:02:17.650 --> 00:02:21.710 tab. It is a persistent entity. It is also multi 00:02:21.710 --> 00:02:23.669 -channel. This is the part that fascinates me. 00:02:23.750 --> 00:02:26.770 You don't open a special AI app to talk to it. 00:02:27.009 --> 00:02:30.330 You talk to it via Telegram, Discord, Slack, 00:02:30.590 --> 00:02:33.490 or WhatsApp. The documentation points to Telegram 00:02:33.490 --> 00:02:36.090 as the easiest starting point. And that choice 00:02:36.090 --> 00:02:38.530 changes the psychology of how you use it. Yeah, 00:02:38.550 --> 00:02:40.509 it makes Hermes feel like a real human assistant 00:02:40.509 --> 00:02:43.250 living inside your phone. You just text it, you 00:02:43.250 --> 00:02:45.389 are at the grocery store, you get an idea, and 00:02:45.389 --> 00:02:47.669 you just send a message to your bot. And it actually 00:02:47.669 --> 00:02:50.430 uses tools. Look, standard AI is like a smart 00:02:50.430 --> 00:02:52.789 dictionary. It knows a lot of things, but it 00:02:52.789 --> 00:02:54.729 just sits on the shelf. Right. Hermes is like 00:02:54.729 --> 00:02:56.930 giving an assistant their own desk. You give 00:02:56.930 --> 00:02:59.409 them a set of tools, a memory system, a calendar, 00:02:59.569 --> 00:03:01.930 and a phone number. It can run actual terminal 00:03:01.930 --> 00:03:04.849 commands. It uses browser automation to look 00:03:04.849 --> 00:03:08.030 at websites. It even has vision to analyze images 00:03:08.030 --> 00:03:10.530 you send it. That is the crucial difference here. 00:03:10.669 --> 00:03:14.669 The workspace is where it runs. The tools are 00:03:14.669 --> 00:03:17.550 what it uses to manipulate that workspace. The 00:03:17.550 --> 00:03:20.189 calendar is for its scheduled tasks. It isn't 00:03:20.189 --> 00:03:22.509 just generating text. It is executing logic. 00:03:23.090 --> 00:03:26.449 Whoa. Beat. I mean, imagine an assistant that 00:03:26.449 --> 00:03:29.430 remembers your exact workflows. And it can run 00:03:29.430 --> 00:03:31.710 them in the background without you typing a thing. 00:03:31.810 --> 00:03:34.069 It's wild. It's pulling reports and formatting 00:03:34.069 --> 00:03:36.469 data while you literally go for a walk. That 00:03:36.469 --> 00:03:38.469 is the real promise here. It isn't about writing 00:03:38.469 --> 00:03:40.810 a better poem. It creates reusable workflows 00:03:40.810 --> 00:03:43.370 for the tedious digital chores you repeat often. 00:03:43.629 --> 00:03:45.629 But this makes me wonder, does this mean we're 00:03:45.629 --> 00:03:48.389 replacing our standard chat GPT habit entirely? 00:03:48.870 --> 00:03:51.889 No, it just moves AI from conversation to... 00:03:52.240 --> 00:03:55.060 actual background task execution okay that makes 00:03:55.060 --> 00:03:56.560 sense it is a different kind of tool entirely 00:03:56.560 --> 00:03:58.960 and that brings us to the broader landscape yeah 00:03:58.960 --> 00:04:01.360 if we're going to invest time setting this up 00:04:01.360 --> 00:04:03.860 we need to know where it fits we don't want you 00:04:03.860 --> 00:04:07.500 using a hammer to drive a screw exactly let's 00:04:07.500 --> 00:04:09.560 look at how armies compares to other big names 00:04:09.560 --> 00:04:12.620 in the agent space specifically claude code and 00:04:12.620 --> 00:04:15.599 open claw right if we connect this to the bigger 00:04:15.599 --> 00:04:18.920 picture the ai agent space is fragmenting really 00:04:18.920 --> 00:04:22.240 fast Six months ago, an agent was just a buzzword. 00:04:22.399 --> 00:04:25.660 Now, different tools have very specialized jobs. 00:04:26.040 --> 00:04:28.379 Let's start with Cloud Code. Going through the 00:04:28.379 --> 00:04:31.079 developer blogs, it is clear this is built for 00:04:31.079 --> 00:04:34.509 focused, deep work. desk work. Exactly. Cloud 00:04:34.509 --> 00:04:36.449 Code is what you use when you are sitting at 00:04:36.449 --> 00:04:38.850 your computer staring at a massive code base. 00:04:39.189 --> 00:04:41.930 It excels at local environment management. It 00:04:41.930 --> 00:04:44.629 edits your files, manages your repositories, 00:04:44.689 --> 00:04:47.129 and works deep inside your terminal. It is a 00:04:47.129 --> 00:04:49.829 dedicated coding partner. But it is not a mobile 00:04:49.829 --> 00:04:52.550 first assistant. You aren't going to text Cloud 00:04:52.550 --> 00:04:54.970 Code on WhatsApp while you're driving to remind 00:04:54.970 --> 00:04:57.290 you about a meeting. No. Or to summarize a news 00:04:57.290 --> 00:04:59.810 article. Right. It is heavy machinery. Then on 00:04:59.810 --> 00:05:01.050 the other end of the spectrum, you have something 00:05:01.050 --> 00:05:03.449 like OpenClaw. The repositories show this is 00:05:03.449 --> 00:05:05.769 a much heavier system. Okay. It is basically 00:05:05.769 --> 00:05:09.000 a multi -agent gateway. Wait. A gateway system. 00:05:09.240 --> 00:05:12.000 So it connects messaging apps, multiple specialized 00:05:12.000 --> 00:05:15.420 agents, and external plugins into one massive 00:05:15.420 --> 00:05:18.600 architecture. Yeah. That sounds incredibly complicated. 00:05:18.939 --> 00:05:21.899 It is. It is powerful, but it can feel very brittle. 00:05:22.120 --> 00:05:24.879 It is really designed for power users, startup 00:05:24.879 --> 00:05:27.620 founders, or developers building complex commercial 00:05:27.620 --> 00:05:29.899 platforms. So when you have that many moving 00:05:29.899 --> 00:05:33.220 parts, if one plugin updates, the whole house 00:05:33.220 --> 00:05:36.029 of cards might fall. Exactly. It requires constant 00:05:36.029 --> 00:05:38.509 maintenance. Which leaves us with Hermes Agent. 00:05:38.709 --> 00:05:41.930 It sits in a very specific practical lane. It 00:05:41.930 --> 00:05:44.810 is lightweight. It is telegram first. And it 00:05:44.810 --> 00:05:46.790 is designed to be a self -improving personal 00:05:46.790 --> 00:05:49.470 assistant. Hermes is for when you are walking 00:05:49.470 --> 00:05:51.790 around with your phone, but you still want your 00:05:51.790 --> 00:05:53.430 digital work to move forward. You want something 00:05:53.430 --> 00:05:55.870 that grows with your specific habits. And as 00:05:55.870 --> 00:05:57.850 a bonus, the architecture is great for testing 00:05:57.850 --> 00:06:00.329 new open source language models. Yeah, definitely. 00:06:00.529 --> 00:06:03.170 But hold on. If I am using cloud code for my 00:06:03.170 --> 00:06:05.410 programming and I am setting up Hermes for my 00:06:05.410 --> 00:06:07.810 daily schedule, aren't I just giving myself a 00:06:07.810 --> 00:06:10.350 management headache? I mean, it feels that way 00:06:10.350 --> 00:06:12.790 at first, but separation of concerns is actually 00:06:12.790 --> 00:06:15.120 it. s practice here right you wouldn't ask your 00:06:15.120 --> 00:06:17.699 accountant to also design your logo exactly you 00:06:17.699 --> 00:06:20.879 want specialized tools for complex tasks hermes 00:06:20.879 --> 00:06:23.740 acts as the project manager while claude is the 00:06:23.740 --> 00:06:26.339 factory worker so if i'm just coding an app is 00:06:26.339 --> 00:06:29.699 hermes the wrong tool yes use claude code for 00:06:29.699 --> 00:06:32.519 deep coding and hermes for daily assistant tasks 00:06:32.519 --> 00:06:35.620 got it now we really need to look under the hood 00:06:35.620 --> 00:06:39.480 to understand how hermes manages to be so fiercely 00:06:39.480 --> 00:06:42.620 independent we have to examine its brain Yeah, 00:06:42.660 --> 00:06:44.759 the architectural docs break this down into five 00:06:44.759 --> 00:06:47.439 core pillars. These five pillars are the exact 00:06:47.439 --> 00:06:50.259 mechanisms that turn it from a reactive, forgetful 00:06:50.259 --> 00:06:53.939 chatbot into a proactive, reliable agent. Pillar 00:06:53.939 --> 00:06:56.420 number one is memory. And honestly, this is the 00:06:56.420 --> 00:06:59.129 biggest friction point with AI right now. Most 00:06:59.129 --> 00:07:02.250 AI models wake up with total amnesia. Oh, absolutely. 00:07:02.509 --> 00:07:04.769 Every single section starts totally fresh. That 00:07:04.769 --> 00:07:06.990 is incredibly frustrating. You don't want to 00:07:06.990 --> 00:07:09.250 explain your business model, your project context, 00:07:09.310 --> 00:07:12.189 and your personal communication style every single 00:07:12.189 --> 00:07:14.689 time you open a chat. Ermi solves this at the 00:07:14.689 --> 00:07:17.290 root level using two specific markdown files. 00:07:17.649 --> 00:07:19.910 Let's break those down. The first is the user 00:07:19.910 --> 00:07:23.269 file. This file is entirely about who you are. 00:07:23.370 --> 00:07:25.629 Right. It holds your professional role, your 00:07:25.629 --> 00:07:28.259 preferences, and your writing style. You can 00:07:28.259 --> 00:07:30.939 literally write in the file, keep my email simple, 00:07:31.240 --> 00:07:34.560 warm, and direct. And the mechanism here is crucial. 00:07:34.860 --> 00:07:38.199 Every time Hermes wakes up to do a task, it secretly 00:07:38.199 --> 00:07:40.519 reads that file first. It gives itself a crash 00:07:40.519 --> 00:07:42.800 course on who you are before it even sees your 00:07:42.800 --> 00:07:45.500 prompt. The second file is the memory file. This 00:07:45.500 --> 00:07:47.579 holds your project context. Yeah. So business 00:07:47.579 --> 00:07:49.879 information, server environment details, recurring 00:07:49.879 --> 00:07:52.279 facts, stuff that doesn't change daily. Yeah. 00:07:52.639 --> 00:07:54.600 But the docs warn you have to be careful here. 00:07:54.720 --> 00:07:57.660 You really do. You only save durable facts, not 00:07:57.660 --> 00:08:00.660 temporary task statuses. If you save every little 00:08:00.660 --> 00:08:03.000 to -do item, the file gets bloated and the AI 00:08:03.000 --> 00:08:05.439 gets confused. Exactly. Pillar two is skills. 00:08:05.699 --> 00:08:08.199 If memory dictates what the AI needs to remember, 00:08:08.399 --> 00:08:11.060 skills dictate how it repeats a task. They act 00:08:11.060 --> 00:08:13.259 as reusable recipes for your repeated digital 00:08:13.259 --> 00:08:15.540 chores. These live in individual skill files. 00:08:15.660 --> 00:08:17.779 They contain step -by -step instructions. They 00:08:17.779 --> 00:08:20.199 also use YAML front matter. Let's define that. 00:08:20.790 --> 00:08:23.589 Hidden code tags that tell the AI when to trigger 00:08:23.589 --> 00:08:26.430 skills. So instead of just guessing how you want 00:08:26.430 --> 00:08:29.170 to report formatted, it reads the hidden tags, 00:08:29.410 --> 00:08:31.730 sees the trigger word, and pulls up the exact 00:08:31.730 --> 00:08:33.950 blueprint you created last week. And I have to 00:08:33.950 --> 00:08:36.629 admit, I still wrestle with prompt drift myself. 00:08:37.110 --> 00:08:39.909 Yeah. Beat. Yeah, it's so common. You know, having 00:08:39.909 --> 00:08:42.690 to constantly rewrite instructions because the 00:08:42.690 --> 00:08:45.110 AI slowly forgets what I want it to do by the 00:08:45.110 --> 00:08:48.820 10th message in a chat. This skill system. completely 00:08:48.820 --> 00:08:51.460 fixes that. It absolutely does. You stop asking 00:08:51.460 --> 00:08:53.559 the agent to improvise every time. You give it 00:08:53.559 --> 00:08:55.960 a rigid track to run on. Which brings us to pillar 00:08:55.960 --> 00:08:58.779 three, the soul. This lives in the soul file. 00:08:59.039 --> 00:09:01.620 This one's fascinating. This file gives the agent 00:09:01.620 --> 00:09:04.419 its actual personality and tone. You can make 00:09:04.419 --> 00:09:06.620 it warm and empathetic. You can make it highly 00:09:06.620 --> 00:09:09.539 sarcastic. Or you can make it strictly technical, 00:09:09.759 --> 00:09:12.500 returning only code and no apologies. And this 00:09:12.500 --> 00:09:15.360 matters deeply if Hermes is, for example, drafting 00:09:15.360 --> 00:09:18.100 customer support replies for your business. A 00:09:18.100 --> 00:09:21.100 generic robotic assistant sounds totally forgettable. 00:09:21.120 --> 00:09:23.259 You want it to sound like your brand. How does 00:09:23.259 --> 00:09:26.379 the sole pillar actually change the output practically? 00:09:26.679 --> 00:09:28.759 It dictates the tone, preventing your assistant 00:09:28.759 --> 00:09:31.220 from sounding like a generic corporate robot. 00:09:31.440 --> 00:09:33.740 Right. It literally changes the weights of the 00:09:33.740 --> 00:09:36.860 words the model chooses. Pillar four is cron 00:09:36.860 --> 00:09:39.480 jobs. Let's define that jargon too. Scheduled 00:09:39.480 --> 00:09:42.279 tasks that run automatically at a specific time. 00:09:42.600 --> 00:09:44.879 This is the feature that turns Hermes from a 00:09:44.879 --> 00:09:48.080 reactive tool into a proactive operator. You 00:09:48.080 --> 00:09:50.480 aren't just messaging it, it is messaging you. 00:09:50.759 --> 00:09:53.899 You schedule a market research summary for 7 00:09:53.899 --> 00:09:56.659 a .m. or database backup every Friday at midnight. 00:09:56.899 --> 00:09:59.200 It just happens without a prompt. And pillar 00:09:59.200 --> 00:10:01.740 five is the self -improving loop. This is the 00:10:01.740 --> 00:10:04.700 absolute magic of the system. It is the process 00:10:04.700 --> 00:10:07.039 of turning your manual corrections into permanent 00:10:07.039 --> 00:10:09.580 skills. You ask Hermes to do a piece of work. 00:10:10.059 --> 00:10:12.919 It does it, but maybe the output is messy. You 00:10:12.919 --> 00:10:16.340 correct it. Then Hermes saves that newly improved 00:10:16.340 --> 00:10:19.080 workflow. It searches its past sessions to keep 00:10:19.080 --> 00:10:21.460 continuity. It literally rewrites its own instructions. 00:10:21.740 --> 00:10:24.299 So the next time it runs that task, the output 00:10:24.299 --> 00:10:26.480 is simply better. It learns from its mistakes. 00:10:27.059 --> 00:10:29.620 Okay, we are going to talk about how to safely 00:10:29.620 --> 00:10:32.700 set this architecture up next, right after this. 00:10:33.519 --> 00:10:36.259 Sponsor, here's where it gets really interesting. 00:10:36.869 --> 00:10:39.389 We know how its brain works with the markdown 00:10:39.389 --> 00:10:42.509 files and the cron jobs. Now, how do we safely 00:10:42.509 --> 00:10:44.970 give it a body in the real world? Because we 00:10:44.970 --> 00:10:48.110 definitely do not want an autonomous AI accidentally 00:10:48.110 --> 00:10:50.950 destroying our personal files. Setup and security 00:10:50.950 --> 00:10:53.769 are absolutely critical here. Remember, Hermes 00:10:53.769 --> 00:10:56.009 can run terminal commands. That means it can 00:10:56.009 --> 00:10:58.769 delete files. It can install software. That is 00:10:58.769 --> 00:11:01.690 incredibly powerful, but very risky. The documentation 00:11:01.690 --> 00:11:04.009 strongly recommends running it on a separate 00:11:04.009 --> 00:11:07.100 VPS or an isolated Docker container. Let's define 00:11:07.100 --> 00:11:09.379 Docker container, a secure digital box keeping 00:11:09.379 --> 00:11:11.919 software isolated from your main computer. Think 00:11:11.919 --> 00:11:14.500 of Docker like putting the AI in a soundproof 00:11:14.500 --> 00:11:17.019 room inside a shared house. If the AI makes a 00:11:17.019 --> 00:11:19.179 massive mess, the house stays perfectly clean. 00:11:19.279 --> 00:11:21.320 You just delete the room and start over. It keeps 00:11:21.320 --> 00:11:23.580 everything safely organized. You log into your 00:11:23.580 --> 00:11:25.840 remote server, install Docker, create a clean 00:11:25.840 --> 00:11:27.740 folder, and then run the official installation 00:11:27.740 --> 00:11:30.399 script. Right. You do not just download it to 00:11:30.399 --> 00:11:32.279 your desktop. During that setup, you connect 00:11:32.279 --> 00:11:34.799 it to Telegram using a tool called Botfather. 00:11:35.370 --> 00:11:38.529 This gives you a unique bot token. But the community 00:11:38.529 --> 00:11:41.350 forums highlight a massive security warning here. 00:11:41.470 --> 00:11:45.230 They do. You absolutely must lock the bot strictly 00:11:45.230 --> 00:11:49.110 to your specific Telegram user ID. Yes. If you 00:11:49.110 --> 00:11:51.570 skip that step, anyone who finds your bot on 00:11:51.570 --> 00:11:54.149 Telegram can start texting it. And since it has 00:11:54.149 --> 00:11:56.649 terminal access, random strangers could be running 00:11:56.649 --> 00:11:59.169 malicious commands on your private server. It 00:11:59.169 --> 00:12:01.129 is a huge vulnerability. Another strict rule. 00:12:01.370 --> 00:12:05.049 Never, ever paste your API keys, your passwords, 00:12:05.230 --> 00:12:07.470 or your tokens directly into the chat window. 00:12:07.669 --> 00:12:09.929 You store those secrets safely in .env files. 00:12:10.409 --> 00:12:13.149 That's a hidden digital safe used specifically 00:12:13.149 --> 00:12:15.529 for storing your private passwords. And you use 00:12:15.529 --> 00:12:17.889 a .gitignore file to protect them from being 00:12:17.889 --> 00:12:20.730 uploaded to public code repositories. Clean security 00:12:20.730 --> 00:12:23.500 habits from day one are essential. The developers 00:12:23.500 --> 00:12:25.720 also recommend disabling any tools you don't 00:12:25.720 --> 00:12:27.980 actually need yet. If you aren't using browser 00:12:27.980 --> 00:12:31.000 automation, turn it off. Limit the attack service. 00:12:31.139 --> 00:12:32.919 So why shouldn't I just run this locally on my 00:12:32.919 --> 00:12:35.519 MacBook right now? An autonomous agent with terminal 00:12:35.519 --> 00:12:38.000 access could accidentally break your personal 00:12:38.000 --> 00:12:40.379 computer files. That is a terrifying thought. 00:12:40.600 --> 00:12:43.279 So we put it in a safe space, we lock it down, 00:12:43.480 --> 00:12:46.480 then we build our first workflows. Yeah, and 00:12:46.480 --> 00:12:48.659 the community guides say you should not build 00:12:48.659 --> 00:12:52.220 complex multi -step automations on day one. Start 00:12:52.220 --> 00:12:55.500 very small. The very first workflow you build 00:12:55.500 --> 00:12:58.100 should be connecting a private GitHub repository 00:12:58.100 --> 00:13:01.340 for daily cron job backups. This makes total 00:13:01.340 --> 00:13:03.899 sense. You are backing up the memory and skill 00:13:03.899 --> 00:13:06.659 files. You are literally saving the brain and 00:13:06.659 --> 00:13:09.080 habits of your assistant. If the server crashes, 00:13:09.360 --> 00:13:11.139 you don't lose the weeks of training you put 00:13:11.139 --> 00:13:13.940 in. The second recommended workflow is a 7 a 00:13:13.940 --> 00:13:16.279 .m. morning briefing. You set a cron job for 00:13:16.279 --> 00:13:18.960 Hermes to pull your top priorities, scan relevant 00:13:18.960 --> 00:13:21.779 industry news, and flag urgent calendar reminders. 00:13:22.100 --> 00:13:23.820 It sends it to your phone right when you wake 00:13:23.820 --> 00:13:26.399 up. It makes Hermes feel genuinely helpful immediately. 00:13:26.659 --> 00:13:28.879 You get a clean, customized start to your day. 00:13:29.000 --> 00:13:31.419 The third recommended workflow is a weekly server 00:13:31.419 --> 00:13:34.120 health check. Reviewing disk usage, checking 00:13:34.120 --> 00:13:36.980 running services, and scanning error logs. Hermes 00:13:36.980 --> 00:13:39.120 lives on a server, so ask it to manage that server. 00:13:39.320 --> 00:13:42.159 Exactly. Checking disk usage isn't just asking, 00:13:42.259 --> 00:13:46.379 is the disk full? It's having the AI run the 00:13:46.379 --> 00:13:49.240 command, read the output, notice that a log file 00:13:49.240 --> 00:13:52.480 is eating up 80 % of the drive, and texting you 00:13:52.480 --> 00:13:54.700 to ask if it should clear it. It built a great 00:13:54.700 --> 00:13:56.799 maintenance habit early on. We have the setup, 00:13:56.899 --> 00:13:59.080 we have the safety checks in place. What does 00:13:59.080 --> 00:14:01.019 this actually look like in the wild when it starts 00:14:01.019 --> 00:14:05.009 working along? This is where we look at the real 00:14:05.009 --> 00:14:07.649 -life applications. Sure, you can use it as a 00:14:07.649 --> 00:14:10.490 simple personal assistant for reminders, but 00:14:10.490 --> 00:14:13.210 the architecture scales up beautifully. You can 00:14:13.210 --> 00:14:15.659 use it as a full business operator. monitoring 00:14:15.659 --> 00:14:18.159 a Discord community, summarizing hundreds of 00:14:18.159 --> 00:14:20.759 YouTube comments to find common questions, or 00:14:20.759 --> 00:14:23.159 drafting technical support replies based on your 00:14:23.159 --> 00:14:25.480 actual documentation. What's fascinating here 00:14:25.480 --> 00:14:28.000 is that the value isn't day one perfection. You 00:14:28.000 --> 00:14:30.360 have to adjust your expectations. The real value 00:14:30.360 --> 00:14:33.580 is day 30 automation. Exactly. Think about a 00:14:33.580 --> 00:14:35.720 content creator workflow. Hermes could draft 00:14:35.720 --> 00:14:38.340 scripts, pull research, and create data diagrams. 00:14:38.889 --> 00:14:41.370 You can give it a cron job and say, every Friday 00:14:41.370 --> 00:14:44.370 at 4 p .m., review my audience comments from 00:14:44.370 --> 00:14:47.129 the week and create 10 new content ideas. It 00:14:47.129 --> 00:14:49.409 just happens. And you can interact using voice, 00:14:49.509 --> 00:14:51.850 which is huge for accessibility and speed. You 00:14:51.850 --> 00:14:53.610 can be out on a walk, not looking at a screen, 00:14:53.730 --> 00:14:55.889 and just brain dump a chaotic voice message into 00:14:55.889 --> 00:14:58.450 Telegram. Hermes will transcribe it, parse out 00:14:58.450 --> 00:15:00.409 the actionable items, and create a structured 00:15:00.409 --> 00:15:02.929 task list in your system. It also has vision 00:15:02.929 --> 00:15:05.429 capabilities. You can feed it UI screenshots 00:15:05.429 --> 00:15:08.370 from an app you were building or complex data 00:15:08.370 --> 00:15:11.250 diagrams and ask it to check for visual errors. 00:15:11.490 --> 00:15:14.409 Or you can use it as a dedicated, relentless 00:15:14.409 --> 00:15:16.990 research assistant. You ask it to monitor specific 00:15:16.990 --> 00:15:20.370 AI news feeds during the week. Compare three 00:15:20.370 --> 00:15:23.429 different software tools. Pull the pricing. format 00:15:23.429 --> 00:15:25.669 it into a table, and save the findings directly 00:15:25.669 --> 00:15:27.710 to its memory. And this all feeds right back 00:15:27.710 --> 00:15:29.730 into that magic self -improving loop we talked 00:15:29.730 --> 00:15:32.669 about earlier. You start with one real messy 00:15:32.669 --> 00:15:35.509 task. You let Hermes attempt to help. You review 00:15:35.509 --> 00:15:38.330 the output and you correct it. You say, no, don't 00:15:38.330 --> 00:15:40.769 use bullet points. Use numbered lists and keep 00:15:40.769 --> 00:15:43.289 the summaries under two sentences. And then you 00:15:43.289 --> 00:15:45.509 save that exact pattern as a permanent skill. 00:15:45.950 --> 00:15:47.870 What happens if the first research reported generates 00:15:47.870 --> 00:15:50.990 is completely useless? You correct it, then tell 00:15:50.990 --> 00:15:53.850 it to save that improved format as a skill. It 00:15:53.850 --> 00:15:56.330 is so incredibly practical. You are not just 00:15:56.330 --> 00:15:58.210 sitting around waiting for the big tech companies 00:15:58.210 --> 00:16:01.370 to make the base AI model smarter. You are actively 00:16:01.370 --> 00:16:06.149 teaching your specific instance how to work perfectly 00:16:06.149 --> 00:16:09.190 for you. Beat. you build a fundamentally better 00:16:09.190 --> 00:16:11.450 way of working together. By offloading these 00:16:11.450 --> 00:16:14.210 routines, it drastically reduces the number of 00:16:14.210 --> 00:16:16.350 small, exhausting decisions you have to make 00:16:16.350 --> 00:16:18.610 every single day. So let's bring this all together. 00:16:19.450 --> 00:16:22.309 Hermes Agent and tools like it represent a fundamental 00:16:22.309 --> 00:16:24.710 shift in computing. We are moving from renting 00:16:24.710 --> 00:16:27.350 intelligence one prompt at a time to owning an 00:16:27.350 --> 00:16:31.029 evolving operator to sex silence. It builds its 00:16:31.029 --> 00:16:33.730 own memory. It builds daily habits. It builds 00:16:33.730 --> 00:16:35.990 complex workflows tailored exclusively to the 00:16:35.990 --> 00:16:38.490 messy reality of your life. It is definitely 00:16:38.490 --> 00:16:40.690 not magic out of the box. I think that's important 00:16:40.690 --> 00:16:43.370 to stress. It requires genuine guidance and discipline. 00:16:43.649 --> 00:16:46.269 You have to actively prune its memory and take 00:16:46.269 --> 00:16:48.490 the time to create those foundational skills. 00:16:48.809 --> 00:16:51.690 But the compounding return on your time is immense. 00:16:52.009 --> 00:16:54.950 Once you get those skills dialed in, it operates 00:16:54.950 --> 00:16:57.370 silently in the background, constantly pushing 00:16:57.370 --> 00:16:59.710 your work forward while you focus on the big 00:16:59.710 --> 00:17:03.009 picture. It essentially becomes a custom, lightweight 00:17:03.009 --> 00:17:05.490 operating system for your professional life. 00:17:06.250 --> 00:17:08.130 Before we wrap up today, I want you to think 00:17:08.130 --> 00:17:10.470 about this. Look closely at your daily routine 00:17:10.470 --> 00:17:14.150 right now. What is the one tedious digital chore 00:17:14.150 --> 00:17:17.250 you do every single morning? Pulling data, checking 00:17:17.250 --> 00:17:20.069 a dashboard, formatting an email that you can 00:17:20.069 --> 00:17:21.990 hand off to an autonomous assistant right now. 00:17:22.230 --> 00:17:24.309 Thank you for joining us on this deep dive. We 00:17:24.309 --> 00:17:26.069 covered a lot of technical ground today, but 00:17:26.069 --> 00:17:27.970 hopefully it gave you a blueprint for the future. 00:17:28.210 --> 00:17:31.049 So what does this all mean? It means your new 00:17:31.049 --> 00:17:33.829 assistant is waiting for instructions. Out hero 00:17:33.829 --> 00:17:34.109 music.