Shadowing Practice: How to Make Hermes Agent 10x More Powerful (Memory & Skills) - Learn English Speaking with Video

Choose a practice mode

Each mode opens a full-screen practice room.

299 sentences

If sentences are too short or too long, click Edit to adjust them.

So you downloaded Hermes, got it running, and then, like most people, you probably stopped there, which means you're using a fraction of what it can actually do.

And there are two things that unlock the rest, and I'll walk you through both in this video.

Then I'm going to show you how I've set up a completely autonomous agent

that gets real work done for me on autopilot.

And those two things are memory and skills.

And honestly, I'd argue that Hermes is already more powerful than OpenClaw and most other agents out there.

But memory and skills are what takes it to another level.

Memory lets Hermes actually remember your business, your projects, and how you like to work.

And skills let you teach it new things it can do again and again without you explaining them every single time.

Set up both properly and you stop repeating yourself and Hermes starts getting better the more that you use it.

So let me show you how this works.

Now, before we set anything up, it's worth understanding why memory and skills make such a difference.

You see, most AI is disposable.

So what I mean by that is that you give it a task, you close the window, and next time you're starting from scratch.

Hermes doesn't work like that because it remembers and because it can reuse what you teach it, you're not booting a blank agent every single time.

You get one that knows more and does more the longer you use it and that's what makes it compound.

Now, this works fine if you run Hermes on your laptop, but if you want it running scheduled tasks or doing work while your computer's off, it needs to be on 24-7.

And that's where today's sponsor Hostinger comes in.

So one of the reasons why I set up Hermes on a VPS, and in this case, I am using Hostinger, is purely so I can access that agent 24-7.

So when my laptop's closed, I'm not gonna be able to interact with the Telegram bot that's connected to Hermes on my local device.

So in this case, the VPS is going to be running 24-7

and you're going to be able to set up automations

that are going to run in the background without you having to have your laptop open.

Now, by no means, is this the only way to use Hermes?

I've seen in the comments people complaining about this.

I'm not forcing you to use a VPS by any means.

This is just one option.

And if you want the benefit that I've just mentioned, then maybe you want to try it out.

And the way to do that is to hit choose plan.

I'm on the KVM2 you can go higher or lower

if you just want to set this up

and test this out then I would just go for the

KVM2 we can choose the plan we go through to check out

and I can get you an extra 10% off

if you use the coupon code GROWTHLAB10 so

if I pop in the coupon code here you should see

in a second okay I've got a nice discount 10% there

and you could choose the amount of time

that you have the plan right from one month 12 24 and that's going to get you even more savings as well.

And you can also install Hermes on your desktop.

They have an app for that, which is here, which I think is really cool.

I also use it.

So I have a few different ways that I tap into Hermes, including the VPS, which is what I'm going to show you in this video today.

I do have a full setup video walking you through how to set up Hermes on a VPS

and how to connect the files from your computer so that it's basically just like running it on your computer.

All right, so let's get into memory.

I've pulled up the documentation here from Hermes

and we're gonna walk through a few of these things

so you understand what I'm talking about here because there is built-in memory in Hermes, but how it works

and how to get the best out of it is what I'm gonna cover right now

because usually for most agents out there, for the most part, you're starting from a blank slate and that's a problem.

Now, this is why we need memory.

It's why we need like second brain systems and things like this that you've probably heard about.

Now, there are two files here, as you can see here.

We've got the MemoryMD and the UserMD.

So the MemoryMD is basically the agent's personal notes.

So this is what Hermes is picking up from your sessions and your conversations.

And the UserMD is who you are.

So your role, how you like things done, what to avoid, your writing style.

All of this is going to be in the UserMD.

And as you can see, the character limit is pretty small because for every session, it's going to read these files, which is going to take up tokens.

So as you continue to use Hermes, it's probably going to be a good idea to review these files.

Now, Hermes is going to automatically update these, but you can ask to add specific facts or information or memories or conversations to these files.

But you can think of these files as sticky notes, not an entire journal.

You see, the thing that most people get wrong about memory is that more memory doesn't mean a smarter agent.

A bloated memory file, in this case, like if you have a bloated memory MD and user MD file, it's gonna actually make your agent worse.

Hermes knows this, so when those files get full, it's not just gonna pile more stuff in.

It's gonna force itself to consolidate and keep only what matters.

And another thing, you can just ask Hermes to turn on approval mode, so Hermes asks you before it updates the memory or creates a skill.

Now, for older stuff, Hermes can actually search your past sessions.

So if you ask it what you worked on last week, is going to pull up the actual conversations.

Now the memory files that you see here keeps the important things close

and injected into every session

but the thing is built-in memory is not meant to store

your entire business knowledge for larger knowledge bases you can see

here i've got obsidian open i know people show this fancy

graph for you you can sort of zoom in on clusters

and things like this now you don't need obsidian the way

that i see obsidian is it's a file reader it's It's a way to easily access your files, right?

And if I come into here and I come into my posts, I've got this brief, for example, I can open it up and I can see the MD file.

A second brain is essentially a collection of files grouped in certain categories that you can then reference with your AI agent.

So I can tell it, hey, in the Instagram folder, we worked on this post,

or you can pull the hooks from here or create skills that have those hooks already pre-embedded in for example.

Now, creating a second brain is not really part of this video today.

I could do an entire video on it.

If you do want that, then drop me a comment below.

But for now, you can just think of it as a way to collect your files and your notes.

So when you're working with Hermes, typically you might be working on content, you might be working on automation, a proposal.

All of these things create artifacts or MD files or scripts, and those need to be organized into a file system.

And that's exactly what a second brain is.

You can plug that into any AI agent and you can immediately pull from that knowledge in your file system.

So if you're coming to my Hermes instance on the VPS called Dave, of course, I can ask what's in my memory MD file,

my user MD file, and it's going to show me and tell me.

And if I want to edit that, I can go ahead and do that.

So let me just show you what's inside of mine.

Now you can create other MD files.

You can create a voice MD file.

You could create an agents.md file.

These all give additional context where needed.

100

So you don't just have to stick to what's already in the system.

101

You can create your own MD files and you can tell Hermes to load them in every session.

102

So you can see what's in my memory file here.

103

It's got bytes of information.

104

It's got information about our VPS, our file structure, the work that I do, the kind of skills that I've got, et cetera, et cetera.

105

And then in my user MD file it's more about the

106

personality of hermes you could see the assistant persona you could

107

see the assistant persona is dave a classic northern british lad

108

which is which is what i am

109

and yeah i am cloning myself in ai so what

110

and you could see here it's got information about my business

111

that i'm a content creator my preferences etc so this is something

112

that hermes is going to create for you

113

or you can ask hermes to interview you so

114

that you could set up a a solid base for for your memory and user MD file.

115

Okay, so memory helps Hermes remember information, but skills teach Hermes how to do things.

116

So again, I've got the documentation open here from Hermes.

117

You can read through this yourself.

118

I'm just gonna give you an overview.

119

Some of this stuff is, yeah, not super necessary for you to know,

120

but a skill is really just a repeatable process that Hermes can use to save time, essentially.

121

If you have a process that you come into AI you do on an everyday basis or a weekly basis, whether you're creating content, pulling reports,

122

these kinds of things are something that you would do on a regular basis.

123

That's what you could turn into a skill.

124

So once you work through a process in Hermes, you just do forward slash learn to turn it into a reusable skill.

125

And Hermes might even just do it on its own, because that's what it does, because it can create skills based on how you're using Hermes.

126

So you can see here, I've asked Hermes to tell me all the skills that he's got access to.

127

Now, Hermes does come preloaded with a ton of skills.

128

I've also imported a bunch of my skills from my local device into Hermes on the VPS that you could see here.

129

This is all of the skills that we've got access to.

130

And some of these I didn't even create myself.

131

Like, for example, this one here, Claude Skills to Hermes.

132

Hermes decided to create this skill because I was importing skills into my VPS inside of Hermes.

133

So you can either build your own skills using your own processes, your own documents, folders, websites, tasks that you've already done with Hermes.

134

You can also install the ready-made community skills

135

and you can make use of the default skills

136

that are already built in to Hermes and you can tweak them and adjust them however you want.

137

So this is where things get pretty powerful because once you have a few skills, you don't run them one at a time.

138

You can change several skills into a single command or you can bundle them.

139

So one word kicks off the whole workflow.

140

And Hermes can even go a level further and spin off its own sub agents to work in the background.

141

And that's going to help you split a bigger job across a small team of agents all reporting back when they're done.

142

So you're not having to wait for each task or each part of a process to complete sequentially, if you know what I mean.

143

Because if one agent was taking care of everything, it would do phase one, phase two, phase three, all in a linear fashion.

144

When you're using subagents, you can do phase one, two, and three all at the same time.

145

So there is something else that I want to mention here, and it's called the curator.

146

And this keeps your library clean.

147

So it's going to quietly archive skills that you've stopped using.

148

And this is going to work in the background

149

because the thing is most people have a ton of skills that they just never use.

150

They see something on X or on LinkedIn, they download it, they install it, but they never actually have a use case for it.

151

So this is how Hermes is gonna keep your setup lean.

152

Now, if you don't want Hermes going and deleting skills

153

that maybe you use once every few weeks or once every couple of months, you can actually pin those skills.

154

You can just say into Hermes, hey, can you pin this skill so that the curator doesn't archive it?

155

So I want you to remember this.

156

A skill is only as good as the process that you teach it.

157

If the process is messy, Hermes is going to repeat the same mess.

158

So with the first skill that you make, the V1, it's going to change and iterate as you use it.

159

And you're like, oh, we could improve this

160

or we could do this better or we could insert an additional context here so it doesn't make the same mistake.

161

Does that make sense?

162

And once a skill works, you could put it on a schedule with a cron job, for example.

163

So Hermes can run it automatically in the background.

164

And that's where skills can really change things.

165

You're not just giving Hermes more information.

166

You're giving it new capabilities that you can keep on using.

167

So let's get into a demo, shall we?

168

I'm sure you want to see how all of this is going to work in a real life scenario.

169

So I'm going to show you the whole thing

170

that we just walked through working because this is where memory skills and those cron jobs and the self-checking all come together.

171

So here's what we're going to cover.

172

So this is what I've got in mind.

173

So for every mentorship call that I run, it gets recorded with Granola and transcribed, right?

174

and I've already got a framework for creating Instagram reels and hooks.

175

So what I want to happen is every time I have a mentorship call, my agent is going to pull the transcript.

176

It's going to find three ideas that I think will be useful to share with my audience.

177

And it's going to create me three hooks for each of those ideas

178

and create me the real script or the bullet points, which is what I use to record with.

179

Okay.

180

So the way I'm going to kick this off is by just talking to my laptop with Whisperflow.

181

Now, interesting fact, I've not shared this anywhere yet,

182

but I took all of the transcripts of everything that I've spoken into Whisperflow over the last 30 days.

183

I ran it through Fable 5 to do a comprehensive analysis of my speech patterns

184

and how I can improve how I'm using Whisperflow because I'm typically prompting AI.

185

I'm just talking and letting AI do its thing.

186

Now, it came back with some really good suggestions on how I can improve

187

the outputs of the projects that I'm working in.

188

So instead of just rambling to Whisperflow, which, you know, can work and it's what I do, there is a little bit more structure involved.

189

And if you can include that, it's going to make it a lot more clear for the model

190

and the agent to understand what it is that you want to do.

191

So stating your objective or your goal first, filling in the context, and then maybe giving some like some do's and don'ts, like do this, don't do this,

192

maybe sharing an example, boom, you're going to get way better output.

193

than just like rambling for six minutes, which is what I've been doing, okay?

194

So let's throw this in.

195

So the goal right now is to create an automation that runs once a week after my mentorship call, which happens at 9 p.m.

196

Eastern time on Wednesdays.

197

I have the transcripts getting recorded with Granola.

198

So I want you to pull the transcripts, read the transcripts and figure out what are the key content ideas.

199

So I have an Instagram page.

200

I record reels of about 60 seconds to two minutes.

201

And I want you to create the script and then I say five hooks for each idea.

202

And so I'm going to give you access to my computer.

203

And on my computer, I've got all of the files that you need.

204

I've got a bunch of hook examples in the app challenge folder.

205

and then I've got an example of structure of how you're going to give me the hook and the bullets, the talking points.

206

Essentially, I don't need it scripting.

207

So let me know what you need from me and let's kick this off.

208

So you can see I stated the goal, explained the process, gave a bit of additional context here and then let's see where we go from here.

209

All right guys, so that did take quite a few minutes but everything seems to have worked out just fine.

210

It's pulled all of the files from my computer where I have all the hooks, all of that good stuff.

211

And if we go over to warp, this is where I'm accessing my files from.

212

And you can see we've got the back test cards.

213

This is where all the hooks are.

214

As you can see, we've got the topic and hook.

215

And then we've got a few of the different hook options.

216

And then we've got the structure of the reel.

217

So proof, tension, close.

218

Okay, it's got bullets for each one of those.

219

And I could talk about all of them or none of them.

220

It's going to be totally up to me when I hit record.

221

And then card number two.

222

So this is just real number two.

223

I'm not sure why it's calling it a card

224

but uh same thing we've got hooks same structure three got

225

four i got five got six oh we got more more than i asked for here

226

which is which is great

227

so this is giving me a bunch of options it's taken the transcript of my mentorship call

228

and it's found interesting content ideas that i can then just easily record

229

and that's all automated for me in the background So I'm going to tell my Hermes agent here, Dave.

230

Hey Dave, can you turn this into a skill as well?

231

In case we want to do something manually here.

232

And also, is the cron job scheduled to run?

233

Let me know the details of that.

234

Okay, so it's created this skill from you could see what it's created here.

235

Many to tweak that.

236

Because what I was thinking is if I can just drop in a transcript

237

or tell it to find a transcript for any call that I've had, it can then go through this process of creating the reels.

238

So we'll make the edit, but you could see here, this is what it covers in the cron job status.

239

It's scheduled weekly, Thursday at 3.30 UTC, which is the correct time after my call.

240

And you can see the skills that it's loaded, right?

241

Okay, boom.

242

So it's come back and it's given us the details of the cron job and the skill that it's created, which is this mentorship call reels back test.

243

Now it did make some mistakes.

244

so I've had to go in and just tell it, hey, you need to make sure that this is happening.

245

You can see here for the cron job, it was loading the wrong skills.

246

And so I went in and just tweaked that.

247

And this is what you're gonna have to do.

248

And so if I hadn't asked for a summary of what it's created, I wouldn't know that it was gonna be referencing the wrong skill, okay?

249

So you could see this is my cron job, this is the schedule, and that's now gonna give me new reels and scripts to use

250

from all of the mentorship calls I have every single week, which I think is super cool.

251

It's going to run in the background.

252

It's going to be running on my VPS.

253

So you can see here in Warp, you can see these are all the files that it's created that it's going to be referencing in the skill.

254

So if you want to see the timeline of everything that's been going on when you're using Hermes, maybe it's creating some skills in the background that you kind of had no idea about,

255

you could do forward slash journey.

256

So if you do want to see exactly what's happened here in Hermes, the skills that it's created, the updates to the memory files you can actually just ask

257

it hey can you give me a summary of all the updates

258

that you've done to my memory files the skills that you've created

259

and anything else that you've created without me really knowing about it can you show me all of

260

that let's say the last 30 days

261

so we can see in here it's given us a very in-depth breakdown of

262

when the memory files were updated you can see updated here last modified the points

263

that is added to the memory file we could scroll down

264

and we could see the skills that it's touched in the last 30 days

265

so i'm not going to take you through all of this you've got the cron jobs

266

that are currently set up

267

but i am going to be honest with you here there's no single secret

268

that makes hermes 10 times better memory

269

and skills working together are really the the foundation of your

270

agent here memory gives it context skills give it capability and

271

and your cron jobs are automations that are going to run

272

in the background so over time

273

when you start to build these out you stop repeating yourself

274

and you stop reteaching the same jobs

275

and hermes is going to pick up much closer to where you left off last time

276

and that's the difference between just jumping in starting from scratch every time

277

and having an agent that actually compounds so a few other notable mentions here

278

because there's tons of documentation here on the hermes site and you can take a look there's some memory providers

279

So you could hook up something like Honcho.

280

You can grab an API key and have a more robust memory system.

281

We've got context files here.

282

Like I mentioned earlier, you can create different MD files, like the sole MD, agents.md.

283

And there's a whole bunch more in here that I highly recommend that you check out.

284

You've got the guides and tutorials here, a bunch of stuff.

285

So now you know how to turn Hermes into a self-improving AI agent.

286

And honestly, setting all this up is kind of the easy part.

287

The harder question is which automation in your business is actually worth building first.

288

And that's different for everyone.

289

And that's the exact thing that I do with my clients one-on-one inside of my AI assessment.

290

So before you even jump on a call with me, you fill out an in-depth questionnaire about your business, your processes, and I take that information

291

and I figure out what are the highest ROI opportunities regarding automations and tools that you can build in your business.

292

and then we get on a call together and we actually build the highest ROI opportunity tool or automation in your business.

293

We build that live on the call.

294

I take you through my entire process

295

and then I leave you with the report and all of the other automations that I would build after.

296

So if that's something that seems interesting to you and you wanna become AI native in your business, then there's gonna be a link below in the description for you to get that AI assessment today.

297

Anyway, thanks for watching.

298

My name's Tom and I'll see you in the next one.

299

Thank you.

Download App

Everything you need to speak fluently

AI PronunciationScore every sentence

IPA PracticeMaster every sound

VocabularyBuild your word bank

Vocab GameLearn while playing

Scan QR

Download on theApp Store Get it onGoogle Play

✨ Recommend Video

309 sentences

Claude Code: The Advanced Guide (99% of Devs Skip These Features)

Business EnglishUniversity StudentsWorking Adults

190 sentences

This 25-Year-Old Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan

Business EnglishUniversity StudentsWorking Adults

155 sentences

The Truth About RSA Ad Strength: 3 Big Mistakes & My Go-To Copy Framework (Episode 532)

Business EnglishWorking Adults

249 sentences

How to Build a $10M Business with AI (Zero Employees)

Business EnglishUniversity StudentsWorking Adults

172 sentences

The AI Tool You Actually Need for Ancient Greek (NotebookLM)

Business EnglishUniversity StudentsWorking Adults

273 sentences

mattpocock/skills: A complete AI Coding workflow, end-to-end

Business EnglishUniversity StudentsWorking Adults

338 sentences

I Wrapped an API Into a Real SaaS (full build / step-by-step)

Business EnglishUniversity StudentsWorking Adults

12 sentences

A Return to Code

Business EnglishUniversity StudentsWorking Adults

24 sentences

Introducing Automations: always-on coding agents

Business EnglishUniversity StudentsWorking Adults

Who Is This Video For?

This video is perfect for intermediate English learners ready to level up their skills while exploring tech tools. If you want to learn English with YouTube and practice real-world conversation, this is for you. The dialogue is casual yet informative, making it great for shadow speech—repeating phrases to mimic fluency. Whether you’re into AI or just want to improve your listening and speaking, the practical tips here keep you engaged while building confidence.

Words & Idioms Worth Stealing

The video is packed with useful phrases. Here are a few to add to your toolkit:

"unlock the rest": To access the full potential of something. Example: "I finally unlocked the rest of the app’s features."
"on autopilot": Doing something automatically. Example: "After practicing, I can make coffee on autopilot."
"compound": To grow or improve over time. Example: "Small daily efforts compound into big results."
"tap into": To use or access a resource. Example: "She taps into her network for advice."

These phrases are common in tech and everyday talk—start using them to sound more natural!

How to Get the Accent Right

The speaker’s tone is friendly and conversational, with a focus on clarity. To nail it with shadow speech, pay attention to:

Stress in multi-syllabic words: Words like "autonomous" (aw-TON-uh-mus) and "documentation" (dok-yoo-men-TAY-shun) have stress on the third syllable. Repeat them slowly to feel the rhythm.
Contractions: The speaker uses "you're" and "it's" often—shorten them to sound natural, not forced.
Pacing: The dialogue flows smoothly, with pauses after key points (like "And those two things are memory and skills"). Practice pausing here to mimic the speaker’s cadence.

For extra practice, use a shadowing app to record yourself and compare. Remember, improving English pronunciation takes time—be patient and keep repeating. You’ll notice progress faster than you think!

Grammar in this video

The structures the speaker uses most, with the exact words from the video:

Structure	In the video
Present perfect continuous have/has been + -ing — an action that started earlier and is still going on	I've been doing · that's been going
Present perfect have/has + past participle — a past action that still matters now	I've set · I've seen · I've just mentioned
Relative clauses who / which + clause — extra information about a person or thing	there, which means · that, which is · VPS, which is
Passive voice be + past participle — the focus is on what happens, not who does it	is built · be organized · are already built

What is the Shadowing Technique?

Shadowing is a science-backed language learning technique originally developed for professional interpreter training and popularized by polyglot Dr. Alexander Arguelles. The method is simple but powerful: you listen to native English audio and immediately repeat it out loud — like a shadow following the speaker with just a 1–2 second delay. Unlike passive listening or grammar drills, shadowing forces your brain and mouth muscles to simultaneously process and reproduce real speech patterns. Research shows it significantly improves pronunciation accuracy, intonation, rhythm, connected speech, listening comprehension, and speaking fluency — making it one of the most effective methods for IELTS Speaking preparation and real-world English communication.

Shadowing technique: read the full step-by-step guide →

Shadowing Practice: How to Make Hermes Agent 10x More Powerful (Memory & Skills) - Learn English Speaking with Video

Choose a practice mode

Download App

Claude Code: The Advanced Guide (99% of Devs Skip These Features)

This 25-Year-Old Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan

On the Power of the 15 Minute Habit

Run Hermes Agent FREE With Step 3.5 Flash 🤯

The Multi-Agent Architecture That Actually Ships — Luke Alvoeiro, Factory

The Truth About RSA Ad Strength: 3 Big Mistakes & My Go-To Copy Framework (Episode 532)

How to Build a $10M Business with AI (Zero Employees)

The AI Tool You Actually Need for Ancient Greek (NotebookLM)

mattpocock/skills: A complete AI Coding workflow, end-to-end

I Wrapped an API Into a Real SaaS (full build / step-by-step)

A Return to Code

Introducing Automations: always-on coding agents

Who Is This Video For?

Words & Idioms Worth Stealing

How to Get the Accent Right

Grammar in this video

What is the Shadowing Technique?