Hermes Voice Agent is INSANE! artwork

Hermes Voice Agent is INSANE!

AI News Today | Julian Goldie Podcast

July 5, 2026

Meet Hermes Apollo: A Local, Voice-Activated Jarvis Inside Agent OS (Build Apps, Briefings & Browser Control)The script demonstrates Hermes Apollo, a voice-activated assistant inside Agent OS that responds quickly via ChatGPT real-time and can run agentic tasks through a backend agent using GLM 5.2.
Speakers: Julian Goldie
**Julian Goldie** (0:00)
Apollo, how are you today, sir?

**SPEAKER_2** (0:02)
Impeccably well, sir, and ever ready to assist you. And how are you feeling today?

**Julian Goldie** (0:07)
Yeah, I'm doing great, I'm doing great. Can you just tell me a quick joke, mate?

**SPEAKER_2** (0:12)
Certainly, sir. Why did the scarecrow get a promotion?
Because he was outstanding in his field.

**Julian Goldie** (0:18)
What's a good word to learn in Chinese?

**SPEAKER_2** (0:22)
A splendid word to start with might be xie xie, which means thank you, always useful and ever so polite.

**Julian Goldie** (0:31)
Could you build out a snake game?

**SPEAKER_2** (0:35)
Building that now, sir.
The snake game is ready for you, sir. Enjoy.

**Julian Goldie** (0:39)
Could you build down a countdown timer for 24 hours?

**SPEAKER_2** (0:46)
Building that now, sir.

**Julian Goldie** (0:48)
Then once that's completed, we'll actually get the preview as well down here. So you can see a bunch of stuff that's built out recently for me, like for example, 30 minutes ago, 26 minutes ago, 21 minutes ago, seven minutes ago, etc. This is a pretty powerful system that we're using here with Hermes Apollo as a voice activated agent. It's basically like a modern day Jarvis plugged into Hermes agent. Now, if you're wondering, okay, how does it work step by step? Let me guide you through it.
So basically, you say the word and then you can run this on Windows or Mac. You don't have to just use Mac. And it's basically like a voice copilot inside agent OS. So I can say the wake word and have the wake word switch on and it can answer, it can run agentic tasks. It remembers what you tell it, reads the morning brief as well. So we can click on the briefing section here and it can actually run briefs for us. And we can see our history as well. So we can see previous histories from the past as well in terms of daily briefings. And then it will just pull that up in the background whilst I'm talking to you, which is pretty amazing stuff. So you can see here that it's got the suggested focus, how many action items are opening, the open action items.

**SPEAKER_2** (1:58)
A measured Saturday, sir. Your notes suggest a diligent tidying pass.

**Julian Goldie** (2:04)
I'm just going to mute that, but you get the point.
Pretty powerful stuff. And then we can see our previous briefings. We can see what we've worked on, what's on our mind, what are the main headlines this week as well, which is pretty cool. And the way this runs is basically, we have Chat Cheapity Real-time. That's why it's so fast to reply.
And then we have Jarvis connected on the back end as an agent. And I'm using Glm 5.2, but you could use whatever you want. I think the Glm 5.2 is a good balance between speed and actually getting really nice outputs from it as well. And you can see when it's speaking, it glows up a different color. And also if it's reading out a briefing like it just did a second ago, then that'll be quite different to the color that would use when it's responding to you. So you know what it's doing, etc. So it's pretty powerful stuff what it can do. And you might also wonder, okay, how can you use this? Well, I can say Apollo from across the room and it's listening. It can be in wall mode so we can have it full screen as well, running whenever we need it. We can see our previous history over here, which is pretty cool. We can switch between weekly and monthly briefings as well.
And it could just reread the briefs to me.
And it's running locally as well.
So it's got a wake word. It's hands-free. It hears you, it does speech to text. It has a brain so it runs with Glm 5.2 or you can switch to agent mode as well. So you can actually switch over here to agent mode. And this can also control your browser and that sort of thing as well for you. So for example, if we go back to normal mode here and we switch on real time again, can you just open up juliangoldie.com and then it just does it like so. It's pretty insane. So it's really, really fast to respond. Very agentic and you can see the previous history as well of your chats as well, which is great. Plus you've got everything you've built over here too, which is really, really useful. I haven't seen anyone else create something like this that's just available. We give it away inside of Agent OS. You could build it from scratch, but if you want, I'll set up inside the AI Profit Boardroom. And here's some example prompts that you could run. So you could ask it like, okay, you know, for maths, you could ask it for marketing ideas. You could ask it to open like your Kanban board up or that sort of thing. You could even ask it like, how much disk space you have left or what are the biggest files in your downloads folder, all sorts of different things. You could use it for remembering things as well. So this is actually plugged into our memory galaxy inside Obsidian. The great thing about that is it can recall memories. And also when we use it, like you see here, it was just updated two minutes ago, it will actually recall the memories that I give it as well. And then we can use it for content creation, for building things as well. You might notice before that actually gives a preview of what it's built. And this is all working alongside everything else that we have inside the system. So we have Hermes Oracle that can put in the latest trending news from Twitter. And then we can create SEO content and social media content based on that, plus say update to the latest headlines. And we also have Hermes Astros. And Hermes Astros basically what it does is it looks at your competitors and you can change your watch list as well. So we can have competitors that we're monitoring and then we can come up with new content ideas that are inspired by the original competitor. And you can change the watch list. It's super easy to use. It can generate ideas on tap.

3 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000775483668