Video of the Day20:35Alex Finn · Demo
ChatGPT Voice Mode Turns Your Voice Into a Chief of Staff for AI Agents
A creator's real-time demo of ChatGPT's new voice mode as an always-on dispatcher for AI agents across every device he owns.
July 27thThe argument in one line.
Grok Bot ships as a ready-made team of named AI agents with their own cloud computers, trading the customization of open-source tools like Hermes and OpenClaw for a zero-setup experience built for non-technical users.
Read if. Skip if.
READ IF YOU ARE…
- Someone evaluating AI agent tools who wants to run tasks — browsing, logging into services, editing files — without setting up an open-source harness first.
- A non-technical creator or solopreneur who wants a team of agents doing background work without model selection or environment tinkering.
- Anyone curious how agent-to-agent messaging or a 'teach a task' screen-recording feature could replace repetitive manual workflows.
SKIP IF…
- You already run Hermes or OpenClaw and specifically need open-source customization — model swaps, UI changes, self-hosting.
- You want a deep technical comparison of the underlying models rather than a first-impressions product walkthrough.
The full version, fast.
Grok Bot is a newly launched multi-agent AI product built around named personas — each with its own cloud computer — that can browse, log into accounts, and complete tasks locally or in the cloud while the user watches or takes over. Its standout features are a 'teach a task' screen-recording mode that turns a demonstrated workflow into a reusable skill, agent-to-agent messaging so a new agent can pull context from an existing one instead of being re-briefed, and scheduled 'routines' the agent can recommend on its own. Compared with open-source tools like Hermes and OpenClaw, it trades deep customization for a setup-free experience, making it the stronger default for non-technical users and a weaker fit for anyone who needs to control the model or the interface.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:40
01 · Cold open
Grok Bot is introduced as a direct competitor to Hermes agent and OpenClaw, powered by a new model, with the video framed as a verdict on whether to switch.
00:40 – 01:31
02 · What Grok Bot is
First-run experience: the user is greeted with a named, avatar-having AI agent. The product's whole pitch is feeling like a team of coworkers rather than one chief-of-staff assistant.
01:31 – 02:17
03 · The built-in cloud computer
Each agent can run on the user's local machine or its own cloud computer, watchable in real time, with the option for the user to take over control directly.
02:17 – 03:36
04 · Teach a task
Clicking 'teach a task' starts a screen recording; the user demonstrates a workflow (reading newsletters, buying on Amazon) and the agent converts the demonstration into a repeatable skill.
03:36 – 04:23
05 · Logging in and account access
The built-in computer lets the user log the agent into real accounts — demoed by logging Grok Bot into a video-editing account so it can edit the presenter's videos.
04:23 – 04:59
06 · Clarifying questions
Grok Bot is shown proactively asking follow-up, multiple-choice-style questions before acting on a taught skill, which the presenter compares favorably to Claude's plan mode.
04:59 – 06:16
07 · Agent-to-agent communication
A demo shows one agent messaging another named agent ('Barry') to retrieve what content it has already repurposed, so a new agent doesn't need the full context re-explained.
06:16 – 06:39
08 · Plugins marketplace
A day-one plugins marketplace lets agents pull in context from other apps.
06:39 – 07:40
09 · Routines (scheduled automations)
Scheduled tasks, called routines, can be created manually or recommended by an agent based on the work it's already doing — shown with a weekly video-repurposing routine.
07:40 – 08:13
10 · Mobile app
The mobile app offers full desktop parity, including watching and taking over a cloud computer session from a phone.
08:13 – 09:08
11 · Brain dump to reverse prompt
The presenter's preferred onboarding method: tell the agent everything about yourself, then ask 'what can you do for me?' so the agent generates the task list instead of the user.
09:08 – 09:58
12 · Recap of differentiators
Three strengths are recapped: the built-in multi-agent approach, the persistent cloud computer per agent, and out-of-the-box agent-to-agent context sharing.
09:58 – 11:22
13 · Does it replace Hermes and OpenClaw?
Verdict: Grok Bot wins on zero-setup, crash-free productivity for non-technical users; Hermes and OpenClaw win on full open-source customization, which most users don't actually need.
11:22 – 12:10
14 · Close
Closing pitch to try the tool during its beta rollout and a standard subscribe ask.
Lines worth screenshotting.
- Grok Bot gives every AI agent its own cloud computer, so tasks can keep running even after the user closes their laptop.
- A 'teach a task' mode records the user's screen while they demonstrate a workflow, then turns that recording into a reusable skill the agent can repeat.
- Agents inside Grok Bot can message each other directly, letting a newly created agent pull context from an existing one instead of the user re-explaining their business from scratch.
- Scheduled automations in Grok Bot are called 'routines,' and an agent can recommend which routines to build based on the work it's already doing for you.
- The fastest way to onboard a new AI agent is a 'brain dump to reverse prompt' method: tell it everything about yourself first, then ask what it can do for you.
- Grok Bot's mobile app has full parity with desktop, including the ability to take over a running cloud computer session from a phone.
- Open-source tools like Hermes and OpenClaw offer full customization — model swaps, UI changes — that a closed product like Grok Bot cannot match.
- For roughly 90% of users who don't need deep customization, a zero-setup multi-agent product beats an open-source harness that requires tinkering and can crash.
- The clearest differentiator of Grok Bot versus single-agent tools is that it's built around a team of separately named personas rather than one assistant.
- New AI agent tools are being released so fast that testing each one directly, rather than reading about it, is the only reliable way to know if it fits your workflow.
Three signals for choosing an AI agent tool
AGENT TOOL TRADEOFFS
Whether an AI agent tool should be judged on how fast it gets you productive out of the box or on how much you can customize under the hood depends entirely on your own technical comfort.
02What Grok Bot is
- A tool built around multiple named agent personas signals it's optimized for delegation across several ongoing workstreams rather than one general-purpose assistant.
- Naming and personifying an agent is a UX choice meant to make delegation feel like managing a team rather than issuing commands to software.
03The built-in cloud computer
- An agent that runs on its own cloud computer, rather than your local machine, can keep working after you close your laptop — a real capability gap versus browser-extension-based agents.
- Being able to watch an agent's cloud screen in real time, and take manual control mid-task, turns a black-box automation into something you can audit and correct on the fly.
04Teach a task
- Recording yourself performing a task once and having the agent convert that demonstration into a repeatable skill removes the need to describe a workflow in words.
- Learn-by-demonstration onboarding is most useful for idiosyncratic, person-specific processes that would be tedious to explain from scratch every time.
05Logging in and account access
- Giving an agent its own login to your accounts, not just API access, is what lets it operate inside tools you don't have a formal integration for.
- Keeping the controlled computer visible at all times, rather than hiding the agent's actions, is what makes handing over account access tolerable from a trust standpoint.
06Clarifying questions
- An agent that proactively asks clarifying, multiple-choice questions before acting produces more accurate results than one that guesses and executes immediately.
- The quality of an agent's follow-up questions is a better predictor of whether it will do what you meant than the sophistication of its first response.
07Agent-to-agent communication
- Letting agents message each other to share context eliminates the need to re-explain your business or preferences every time you spin up a new agent.
- Agent-to-agent context sharing becomes valuable the moment you add a second or third agent to handle different parts of your work.
08Plugins marketplace
- A day-one plugins marketplace for pulling in context from other apps signals a product being positioned as a platform, not a single closed tool.
09Routines (scheduled automations)
- Scheduled automations only save time if the agent itself can recommend which ones are worth building, rather than requiring you to invent every automation from scratch.
- Being able to review past runs of a scheduled automation is what makes it trustworthy enough to leave running unattended.
10Mobile app
- Full feature parity between mobile and desktop, including taking over a running cloud session, matters for any tool meant to run unattended tasks you may need to interrupt on the go.
11Brain dump to reverse prompt
- Telling an agent everything about yourself up front and then asking 'what can you do for me?' shifts task discovery from the user onto the agent.
- This reverse-prompt method solves the common 'I don't know what to even ask the agent to do' problem by having the tool generate the task list instead.
12Recap of differentiators
- The three differentiators that matter most when comparing agent tools are built-in multi-agent support, persistent cloud execution, and inter-agent context sharing — not raw model quality alone.
13Does it replace Hermes and OpenClaw?
- A closed, zero-setup product will usually beat an open-source agent harness for non-technical users, because setup friction and crashes are a bigger barrier than any feature gap.
- Open-source tools remain the better choice specifically when you need to swap the underlying model, change the UI, or modify how the agent works under the hood.
- For most users, the ability to customize an agent tool matters far less than the ability to get it running without tinkering.
Terms worth knowing.
- Grok Bot
- The multi-agent AI product reviewed in the video, built around named agent personas that each get access to their own cloud computer.
- Hermes agent / OpenClaw
- Open-source AI agent tools positioned as Grok Bot's direct competitors, valued for being fully customizable but requiring more setup and technical comfort.
- Teach a task
- A Grok Bot feature that records the user's screen while they demonstrate a workflow, then converts the recording into a reusable agent skill.
- Routine
- Grok Bot's term for a scheduled, cron-job-style automation that an agent can run on a set interval, which the agent itself can recommend building.
- Brain dump to reverse prompt
- An onboarding method where the user tells an agent everything about themselves up front, then asks the agent what it can do for them.
Things they pointed at.
Lines you could clip.
00:00
“SpaceX AI just released GrokBot, a direct competitor to Hermes agent and OpenClaw.”
09:58
“So the question becomes this, does it replace Hermes and OpenClaw?”
10:01
“If you want an out of the box experience where you just open it up and you immediately start getting productivity done... then I would say Grokbot is the best.”
10:44
“The downside to not being open source or tinkering is that this isn't going to be as customizable.”
11:24
“The space moves so fast that you just need to try every tool that comes out.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
SpaceX AI just released GrokBot, a direct competitor to Hermes agent and OpenClaw. It's powered by their brand new AI model, Grok 4.6, and it can do anything a human being can do on a computer while also having its own computer in the cloud.
I've been testing it the last week, and it is a really, really good tool. But the question becomes, should you rip out Hermes and Open Claw to use it? Well, in this video, you're about to find out.
Let's lock in and get into it. So this is Grokbot. This is what you're gonna be greeted with when you come into it for the first time.
You'll get your first AI agent, which for me was Slate. You get these agents with these cool logos and profile pictures, which I really enjoy.
They also get their own names as well, so they feel kinda human like, which I think is a really cool way to approach this. The whole point behind Rockbot is they want you to feel like you have a team of coworkers working for you.
I think while Hermes agent and Open Claw go for this feel of having, like, a , Grok is really going for here's a full team of different personas doing work for you on their own computers at all times. What we'll do first here is I'll I'll go through what's special about GrokBot, what the differentiators are, give you a demo, then I'll talk about should you use this over Hermes or not.
But starting off what makes it different, again, is that emphasis on multiple named agents doing work for you at all times. The other big differentiator here is each one has their own computer.
So each one can either use your local computer or have their own computer that lives in the cloud.
And in fact, you can watch everything it's doing on its cloud computer in real time. Not only that, you can click open here, and you can actually go in and control their computer for them.
So if you wanna come in here and you wanna control their browser and do anything inside here, you can do that as well. It's a fully operational cloud computer that your agent and you have control over.
But why would they want you to have control over this computer? Why would they wanna give you such visibility? What's the point?
Well, two different reasons. Number one, you can teach your Grock bot how to do things right in this computer.
So if you click teach a task, it starts recording your screen. And so if I wanted to go now and show it how to read my newsletters, if I wanted to go and show it how to buy things off Amazon, I can do that.
Or if I wanted to show it how to find other AI YouTube videos, I can teach it any workflow I want. I'm right here through the browser. For you, maybe you have some sort of process at work where you have to create a special type of presentation or a document that comes from other types of information.
You can walk through that process here where you click around, find the information, build the document for it, hit the stop button, and then now that agent, as you can see, learns that process and learns from that demonstration.
So it can repeat it over and over and over again and turns it into a reusable skill. As you could see here, it's now watching that demonstration I did, and it will be able to turn that into a skill. That's really cool.
Claude and Chad GBT added similar features over the last few weeks, but neither is as intuitive as Grok is right here. It's as easy as just opening up and clicking teach a task.
Another reason why having a built in computer is really, really cool is it allows you to actually log in to accounts you want your Grock bot to have access to. So I want Grock bot to edit my videos for me.
Right? I don't have my own team like a lot of other YouTubers out there. I don't have editors or thumbnail bill.
I do everything myself. But now what I wanna do is actually have Grockbot try to edit this for me. So I'm gonna go in.
I'm actually gonna log in for Grock so that it can access account and edit my videos for me. And it's all done in this really intuitive user interface.
Everything is really easy to understand. When you need to log in to things or you need action, it's really easy to understand, and you can see the computer at all times that it's controlling. So you can constantly monitor what your agent is doing.
Here's another really amazing part about Grokbot that I like a lot. Uh, it is really good at prompting you for questions. If you teach it new skills, if you ask it to do things, it does a really good job of asking you further clarifying questions so we can understand what you're trying to do more.
So I showed it that screen recording. It's like, okay. What should we turn this skill into?
And then it even asked me if I wanna do a dry run. It is really, really good at asking questions, which has always been kind of my favorite part of, like, using Claude. The plan mode is excellent.
The multiple choice questions it gives you is excellent. I always like that about it. They bring that into here, they do it in a really good level.
Another really awesome part about Grokbod is the agents can communicate with each other. So if you have a working for you, you can be like, hey.
Can you work with Barry on this? Hey. Can you work with the other agent on this?
And they'll share their knowledge and context together to get really good tasks done. Let me show you. So Barry's been doing a bunch of content repurposing for me.
I'm gonna say, can you work with Barry to see which content he has repurposed for me already just so they can be up to date on the work each other has done? This is really important to do if you decide to create new agents so that they can share context instantly rather than having to copy and paste all the context over.
So as you can see here, it's, uh, saying it's gonna ping Barry, message Barry, ask Barry, I'll bring back what he's already repurposed as soon as he replies.
Got the message from Barry. As you can see, you can look at their messages between each other, which is really, really cool. See how they're talking to each other, and you can say, here's all my repurposed inventory, and it sends it over.
That's amazing. That saves you a ton of time. Right?
You start a new agent. Maybe you have one agent that responds to your business emails that you just created, and you have another agent that's been doing a lot of other business work for you.
And you wanna be, hey. Get all the context from this other agent, all the work you've done for me on my business. Instead of having to reexplain your entire business over and over and over again to your new agents, you can say, hey.
Talk to my other agent, see what we've done already, and they go and they text message each other and get the context. That is really cool and saves a lot of time. That is a truly differentiated feature than any other AI agent I've seen out there.
Speaking of context sharing, there is an entire plugins marketplace. So you can see all these different plugins that will allow you to take in context from other apps as well.
That is really, really nice to to see on a day one app. That is really amazing. Cronjobs, they have them.
They are called routines. As you can see, I already have a routine here for repurposing, uh, my weekly videos into newsletters.
By the way, if you wanna keep up to the end of these videos, make sure to like and subscribe down below. Turn on notifications. All I do is make amazing videos about the latest AI tools.
So I'm doing live boot camp about GrockBot this week in the Vibe Coding Academy. Make sure to check that out down below. Linked in the description.
Best decision you ever make. Join that community. Best AI community on planet Earth.
But the routines are really amazing as well. You can decide when to run them. You can see past runs.
Give it specific instructions. Test it out right now. Standard cron job stuff.
But what I really like about these routines is you can go in to any of your agents and say, hey. What routine should we make out of this? You can get, like, recommendations on which routines to make, and it'll automatically spin them up and create them for you.
You So don't have to sit there and, like, manually build out each and every routine. They have a mobile app. The mobile app is excellent.
It is very, very simple and clean. I think they hired a new team of designers over there at SpaceX AI. They are excellent.
You go in. You can see the computers. You can take over the computers from mobile, which is really, really nice.
Even in mobile, come in here, control the computer, log in to accounts, whatever you need to do. All the same functionality as desktop. There really is no difference whatsoever, but in a really clean, nice package makes getting work done on the go really nice.
That is a must download as well. As for how to use it and what you should do when you first get into it, you should use my favorite method for introducing yourself to new agents, which is the brain dump to reverse prompt methodology.
That is you go in, you tell it everything about yourself. As you can see, that's what I did here. And then you reverse prompt it.
So you tell it all about yourself, then you go, hey. What can you do for me? And as you can see here, it gives you a bunch of really nice recommendations, recommends doing repurposing for me, ask me to connect my Gmail account, or recommend tools to connect, or recommend tasks it can do, or recommend routines it can install.
And then you don't really have to do as much thinking. You lean on the intelligence of the agent to do these things. You don't have to come up with every single task for yourself.
It's like the number one question I get all the time is like, oh, boy. What do I even do with these agents? What should I even do?
You don't have to think about it. You lean on the agent. You tell it about yourself.
You say, hey. What can you do for me? You'll get way more done that way.
So to recap the differentiators in the strengths of Grok Bot, one, the kind of multi agent approach is built in. Neither Open Claw nor Hermes is built for this type of approach.
It's as easy as clicking the plus button. You have a brand new agent with a brand new name, persona, all that really, really easily. Two is the built in cloud computer.
So not only can it control your computer like a standard agent, but it is built in from the ground up to have its own computer so you can watch it do its work on your screen. And you can also teach it through that computer how to do your daily tasks that you normally do on your computer so it can just offload that for you.
And then another strength that a multi agent approach is the fact that they can communicate with each other out of the box standard. They're able to coordinate and share context and information, which will save you tons of time. So the question becomes this, does it replace Hermes and OpenClaw?
I will say this. If you want an out of the box experience where you just open it up and you immediately start getting productivity done, no kind of tinkering, model selection, harness selection, UI setup necessary, then I would say Grokbot is the best.
It just works out of the box. There's no crazy setup needed. There's no crashing.
Right? That is really the challenge with Hermes and Open Claw is there's tons of setup and tinkering need to be done. And if you're not super technical, it becomes challenging, and it will crash a lot.
Grok bot just works out of the box and helps you out immediately. Now the downside to that, the downside to not being open source or tinkering is that this isn't going to be as customizable.
I can customize Hermes and OpenClaw literally any way I want. It's completely open source. If I wanted to make it powered by an open source model, if I wanna change how the UI works, if I wanna change something under the hood, I can do that.
You can't really do that with GrockBot. But for, like, 90% of people who don't need to do those things, who just want, like, a a fleet of agents that do work for them, Grockbot's gonna be perfect, actually. Grockbot is going to be really, really good.
But either way, no matter which persona you are, I think you need to download it. I think you need to use it. I think you need to try it out to see if it fits into your workflow.
Because when it comes to AI, the space moves so fast that you just need to try every tool that comes out. You need to test them out to see what works for you.
Some of them will work for you. Some of them won't, but you gotta at least give it a try. Because if you find the ones that fit into your workflow really, really well, it will up your productivity a ton.
And for me, GrockBot has been that. It's excellent. Try it out.
Just released today. I think it's a slower beta rollout. Think by Friday, everyone will be able to use it.
It's Tuesday right now. So if you can't use it now, keep trying over the next few days, but they have started rolling out this beta today. Let me know what you think.
Is this differentiated enough from Hermes and Open Claw for you? Let me know down below. I hope this was helpful.
Subscribe. Turn notifications if it was. I'll see you in the next video.
The Hook
The bait, then the rug-pull.
A brand-new multi-agent AI product just launched with named personas, their own cloud computers, and agent-to-agent messaging — and the obvious question is whether it's good enough to replace the open-source agent tools already doing that work.
Named ideas worth stealing.
09:08list
Grok Bot's Three Core Differentiators
- Built-in multi-agent approach with named personas
- Each agent gets its own persistent cloud computer
- Agents communicate and share context with each other
The presenter's recap of what sets Grok Bot apart from single-agent, chief-of-staff-style tools like Hermes and OpenClaw.
Steal forframing any product comparison between an all-in-one platform and a single-purpose assistant
08:13concept
Brain Dump to Reverse Prompt
Tell a new AI agent everything about yourself and your work up front, then ask it 'what can you do for me?' so it generates the task list rather than you inventing every task yourself.
Steal foronboarding any new AI agent or assistant, not just Grok Bot
How they asked for the click.
VERBAL ASK
11:40subscribe
“Let me know what you think. Is this differentiated enough from Hermes and Open Claw for you? Let me know down below. I hope this was helpful. Subscribe. Turn notifications if it was.”
Soft ask placed after the full verdict and tied to whether the viewer found the breakdown useful, rather than a hard sell up front.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
