HomeVideos

Elon's New #1 AI Agent Just Did What Claude And ChatGPT Couldn't (SpaceXAI Grok Bot)

Now Playing

Elon's New #1 AI Agent Just Did What Claude And ChatGPT Couldn't (SpaceXAI Grok Bot)

Transcript

416 segments

0:00

Elon Musk just built an AI that could

0:02

replace parts of your job, and he may

0:04

have just started a completely new trend

0:06

in AI. Over the last few years, AI has

0:08

mostly helped people do their jobs

0:10

faster, but now it can observe what you

0:12

do and do it by itself. And it's not

0:15

just one AI, this is an army of AI

0:16

workers. They can talk to each other,

0:18

divide tasks, browse the internet, and

0:20

[music] actually get work done. And the

0:22

more you use them, the better they can

0:24

get at working like you. So, as usual, I

0:27

tested it out for you guys. I built 10

0:29

different bots with this. One helps me

0:31

save money every month. [music] Another

0:32

tracks my work every morning. A third

0:34

learned how to do my work just by

0:36

watching me do it once. And at the end,

0:39

I'll show you one free alternative

0:41

that's definitely worth knowing about.

0:43

Now, if you want to build your own bots,

0:45

I've put [music] 100 plus Grok Bot use

0:48

cases, prompts, and the free alternative

0:50

inside my free WhatsApp community. Grab

0:53

it from the link in the description

0:54

before starting. Now, let's dive in.

0:58

Before we open the app, here are three

0:59

cool things you need to know about Grok

1:01

Bot. First, you can teach it how to do a

1:03

job once, and it learns instantly. You

1:06

give it a role, access to the tools it

1:08

needs, and show it how you want things

1:10

done. After that, you can message it

1:12

like a teammate and assign work. Second,

1:14

it can actually go and do the work on

1:16

its own. It gets its own cloud computer,

1:18

so it can open websites, browse the

1:20

internet, use connected tools, and

1:22

complete tasks even when you are not

1:24

sitting in front of your laptop. Also,

1:26

it can be accessed using a phone. And

1:28

third, you can create multiple bots and

1:31

make them work together. One could

1:32

handle sales, another research, another

1:35

content, or whatever fits your workflow.

1:37

You can put them in the same group, so

1:38

one bot can finish its part and pass the

1:41

work to another. That is more useful

1:43

than having 10 separate AI chats open.

1:45

Now that you've understood this, let's

1:47

get into the setup. When you open Grok

1:48

Bot, your bots are on the left and the

1:50

chat workspace is on the right. You'll

1:52

also see plugins where you can connect

1:54

tools like Gmail, Google Calendar,

1:56

Slack, and other apps. So, your bot can

1:58

work directly with those tools. You

2:00

don't have to keep copy-pasting things

2:02

from your AI chat. I've already

2:03

connected a few of mine, so I'm going to

2:05

show you 10 crazy things you can build

2:07

with it, then how to build them

2:08

yourself. We'll create product image and

2:10

video ads, create a YouTube strategy,

2:12

track information on X, manage your

2:14

subscriptions and work emails, get Grok

2:16

bot to learn tasks by watching us once,

2:18

and even coordinate all the bots like a

2:20

chief of staff. So, if there's one of

2:22

these use cases that you care about the

2:24

most, you can skip straight to that

2:25

section. Let's start with the first one.

2:27

The first bot we'll build is Leonardo, a

2:29

generative AI specialist for product

2:31

images and video ads. We'll connect it

2:34

to Higgs Field, so the bot can use top

2:36

AI models for generation. For the first

2:38

prompt, you can give it a simple product

2:40

image and say, "This is the product.

2:43

Create a vintage-style product photo

2:45

shoot around it with an old-school Apple

2:47

campaign feel." For this example, I used

2:49

an old iPod mini, and these were some of

2:51

the results. Once you have the product

2:53

shots, you can continue in the same chat

2:55

and ask it to turn the product into a

2:57

short video ad using the same visual

2:59

style. This is what it created.

3:11

>> [music]

3:13

>> So, with one product image, you can make

3:15

photo shoots and video ads. The second

3:17

bot I made is a personal YouTube

3:19

advisor. You can use it to think through

3:22

your channel, test ideas, or figure out

3:24

what to do next. For example, you can

3:26

ask, "How do I get to 1 million

3:28

subscribers by next month?" Instead of

3:30

giving you generic advice about posting

3:32

more or making better thumbnails, it can

3:34

look at your channel and give you a

3:36

proper breakdown. For example, here it

3:38

shows us where the channel is right now,

3:40

what we need over the next 30 days, and

3:42

different scenarios for how realistic

3:44

that 1 million goal actually is. The

3:46

more context you give it about your

3:48

channel, the better these conversations

3:50

get. Grok can essentially become

3:52

specialist you keep coming back to for

3:55

ideas, strategy, and decisions. And this

3:57

idea of giving AI access to your real

4:00

data isn't just useful for work. It can

4:03

be pretty powerful when it comes to your

4:04

money, too. Before we move ahead, I have

4:07

a quick update for you. One of India's

4:09

most recognizable names in personal

4:11

finance and the founder of the 1% Club,

4:14

Sharan Hegde, just built something

4:16

pretty interesting. It's an AICFO and

4:19

it's completely free. Unlike a regular

4:21

finance app, you don't have to manually

4:23

track everything. It connects directly

4:26

to your banking, stocks, credit cards,

4:28

and insurance and reads all of it. So,

4:30

instead of giving you generic financial

4:32

advice, it can answer questions based on

4:34

your actual money. You can ask something

4:36

like, "When can I afford a car?" and it

4:39

can give you a timeline based on your

4:41

actual income, expenses, and savings.

4:43

You can also ask, "What are my biggest

4:45

spends this month?" or "Which of my

4:47

mutual funds are underperforming?" And

4:49

because it has access to your actual

4:51

portfolio, it can answer using your real

4:53

numbers. It can even calculate your

4:55

financial independence date

4:56

automatically. And the best part is that

4:59

it's completely free to use. [music]

5:00

I've linked it below, so you can ask it

5:02

one question about your own money and

5:04

see what comes back.

5:08

Now, let's get back to Grok bot because

5:10

the next bot can actually start doing

5:12

work for you in the background. Now, let

5:14

me show you the third bot. This can save

5:17

you a lot of time during the day. It's

5:19

called Trending Radar and its job is to

5:21

keep an eye on your X feed for you. You

5:24

can set it to run every 15 minutes. So,

5:26

instead of you opening X regularly, it

5:28

checks what is trending and gives you

5:30

the top five updates. You can see it run

5:33

here. It gives you one set of updates,

5:35

then runs again 15 minutes later, and

5:37

gives you the next set. The reason this

5:40

works in the background is because of a

5:41

feature called routines. A routine is

5:44

just a task you want the bot to repeat

5:46

automatically. In this case, the task is

5:49

to check eggs every 15 minutes for

5:51

updates. If you want to build your own,

5:53

you can create a new routine, tell it

5:55

what to do, and choose when it should

5:57

run. Let AI handle repetitive tasks for

5:59

you. Tell me in the comments what you

6:01

would use this feature for. Now, let's

6:03

move on to another feature that allows

6:05

Trending Radar to keep working without

6:08

taking over your computer. Recently, AI

6:10

has started using your computer,

6:12

searching the internet, and doing agent

6:15

tasks, right? But while that is

6:16

happening, your computer is usually tied

6:19

up. Grok solves that by giving each bot

6:22

its own computer environment. For

6:24

example, Trending Radar needs access to

6:26

X, so you can log into X inside the

6:29

computer that the bot uses. Now, every

6:32

time its routine runs, the bot can open

6:34

X, check the feed, collect the updates,

6:37

and bring the result back to you. So,

6:39

you can keep using your computer or even

6:41

close it while the bot continues

6:43

working. That is what makes these agents

6:46

so useful. The fourth bot you can build

6:48

helps you manage your messages. If you

6:50

regularly share updates with a community

6:52

or a team, you need to turn those

6:54

updates into clean, ready-to-share

6:56

messages. So, you can build a bot

6:58

specifically for this. I share AI

7:00

updates regularly with my WhatsApp

7:02

community, so let's connect it to my

7:04

workflow and give it one simple job.

7:06

Take AI updates and turn them into

7:08

properly formatted WhatsApp messages.

7:11

Then, you can ask, "What's the news for

7:13

today?" The bot collects updates,

7:15

understands the format you need, and

7:16

turns them into a message ready to

7:18

share. You can see it here. It creates a

7:20

clean WhatsApp post in the style you

7:22

want. You can also use the same workflow

7:25

to share updates with your team through

7:26

Slack. And if you want more videos like

7:28

this, where we break down the latest AI

7:30

tools and workflows, make sure you

7:32

subscribe and stay ahead. Now, let's go

7:35

to the fifth setup, building a bot for X

7:37

and another for LinkedIn. The X bot can

7:39

understand how you write, so when you

7:41

give it a topic, it can turn that into a

7:43

post around your writing style. Then,

7:45

you can build a separate bot for

7:47

LinkedIn because the way you write there

7:49

can be different from how you write on

7:50

X. So, now you can have one bot finding

7:53

the trends, one writing for X, one

7:55

writing for LinkedIn, and even one

7:57

preparing the WhatsApp version. At this

7:59

point, the workflow is already useful,

8:01

but managing four separate chats would

8:03

still be annoying. That is where the

8:05

multi-agent setup starts to make a lot

8:07

more sense. Instead of managing four

8:10

separate chats, you can put all of these

8:12

bots into one group and make them work

8:14

together. Just click on new chat and add

8:17

the agents you want. So, you can add

8:19

trending radar, X draft, LinkedIn draft,

8:21

and your WhatsApp bot into one group and

8:24

call it content OS. This becomes your

8:26

sixth setup. You can also customize the

8:28

group as per your choices. Now, you can

8:31

give the whole group one instruction.

8:33

For example, you can ask it to find the

8:35

top five trending posts for the day and

8:37

then turn them into content for X,

8:39

LinkedIn, and WhatsApp. Trending radar

8:41

finds the updates first. Then, a bot

8:44

called X draft writes the X version,

8:46

LinkedIn draft creates the LinkedIn

8:48

post, and the WhatsApp bot prepares the

8:50

community message. So, the same

8:52

information moves through your entire

8:54

content workflow without you

8:55

copy-pasting it from one chat to

8:57

another. Once everything is ready, you

8:59

can review it, make changes, schedule

9:01

it, or send it through the connected

9:03

tools. So far, we've dealt with bots

9:05

that were already set up. Now, let's

9:07

build one from scratch. For the seventh

9:09

bot, let's create something practical.

9:11

If you pay for a lot of software

9:13

subscriptions, after a point, it becomes

9:15

hard to remember what is active, what is

9:18

renewing soon, and what you may not even

9:20

need anymore. So, you can create a bot

9:22

for exactly that. Click create new bot

9:24

and it asks for your requirements. Let's

9:26

give it one simple job. Track all my

9:28

subscriptions from my email. Tell me

9:30

what is active, what is renewing soon,

9:33

and what needs attention. Also, create a

9:35

clean dashboard so I can see everything

9:37

in one place. If Gmail is already

9:39

connected, the bot will go through the

9:41

relevant emails and organize the

9:43

information for you. It groups

9:44

everything into clear sections like

9:47

active subscriptions, upcoming renewals,

9:49

and things you should review. Then it

9:51

goes and builds the HTML dashboard for

9:53

the whole thing. And because the bot can

9:55

keep checking your inbox over time, you

9:57

do not have to rebuild this list

9:59

manually every month. The eighth bot can

10:01

solve another problem that comes up all

10:03

the time. If you get a lot of work and

10:05

collaboration emails, some of them are

10:07

easy to miss when your inbox gets busy.

10:10

For example, I get many partnership

10:12

emails, so I built a bot called

10:13

partnership inbox. Its job is to scan

10:16

emails, find a partnership or

10:18

collaboration, and check whether it's

10:20

already been replied to. If you have not

10:22

replied, it prepares a draft for you. It

10:25

will not send anything without your

10:26

approval if you tell it not to. Then you

10:29

can set it to run every day at 9:00 a.m.

10:31

So, every morning you can open it and

10:34

immediately see which partnership email

10:36

still need attention. Here, it found

10:38

three unanswered emails and prepared

10:40

replies for all three. You only need to

10:42

review the reply and decide whether it

10:44

should be sent. This ninth one is a

10:46

little different because this time you

10:48

do not tell the bot what to do. You show

10:51

it. It is called teacher task. Until

10:53

now, you have been telling the bots what

10:55

to do through prompts, but some tasks

10:57

are much easier to show than explain,

10:59

especially when they involve a lot of

11:01

small clicks. That is exactly what this

11:03

feature is for. You perform the task

11:06

once, the bot watches what you do, and

11:08

then it turns those steps into a

11:10

reusable skill. For example, if you book

11:12

a cab to travel to the office every

11:14

weekday around 11:00 a.m., you can teach

11:17

the bot how you normally book a cab.

11:19

Instead of writing a long prompt

11:21

explaining every step, you simply show

11:23

the bot the process. For example, we

11:25

open the cab website, enter the pickup

11:27

location, select the office as the

11:29

destination, choose the ride, and

11:31

confirm the booking. The bot watches the

11:33

entire workflow, and it turns the steps

11:35

into a skill. Then, you can add one

11:38

instruction, "Run this every weekday at

11:40

11:00 a.m., compare the available ride

11:42

options, and choose the cheapest one."

11:44

It will now permanently run that for

11:46

you. Teacher task is super useful. What

11:48

would you use this for? Tell me in the

11:50

comments. Now, once we have this many

11:52

bots, we have a new problem. We need to

11:54

remember which bot does what. And this

11:57

is why, finally, we'll build a 10th bot

11:59

whose job is to manage all the other

12:01

bots, and we'll call it Chief of Staff.

12:04

So, if you ask, "What's the update on my

12:06

YouTube today?" your Chief of Staff can

12:08

check which agent has the right context,

12:11

message that bot, and bring the answer

12:13

back to you. And if you already know

12:15

which bot you need, you can simply tag

12:17

it inside the conversation. For example,

12:19

you can tag your YouTube expert, ask for

12:21

the latest update, and your Chief of

12:23

Staff passes the request to that agent.

12:26

Once the response comes back, you get

12:28

everything inside the same chat. This

12:30

becomes really useful once you have a

12:32

larger setup. Instead of jumping between

12:34

10 different bots and trying to remember

12:36

where everything lives, you have one

12:38

coordinator that can speak to the rest

12:39

of the team for you. Now, the last thing

12:42

I want to tell you as a feature of Grok

12:43

Bot is that it's not limited to the

12:45

desktop app. You can use it on the

12:48

phone. You can open Grok Bot on iOS with

12:51

the same account and see the same bots

12:53

you already created. It's called Open

12:55

Manus Bot. You can download it on Mac OS

12:58

or Windows, create multiple bots, put

13:01

them into rooms, and even export or

13:03

import complete teams. The biggest

13:05

difference is the model selector. You

13:06

can choose a different AI model for each

13:09

bot. So, one agent can use Claude,

13:11

another can use GPT,

13:13

>> [music]

13:13

>> and another can use Grok, depending on

13:15

the job. You also get separate tasks

13:17

inside each bot, a computer section

13:19

where you can connect a cloud computer,

13:21

and an automation section where you can

13:23

schedule tasks or use webhooks. For the

13:26

cloud computer, you will need a Box API

13:28

key. So, the setup is a little more

13:30

technical than Grok Bot. For app

13:32

connections, Open Man as Bot uses

13:34

Composio, which lets your bots work with

13:36

tools like Slack, Gmail, Google

13:38

Calendar, GitHub, Notion, and more. You

13:41

can also connect to ElevenLabs if you

13:43

want to give your bots a voice and

13:45

actually call them. And just like Grok

13:47

Bot, you can customize each bot with its

13:49

own name, title, description, and even

13:52

make one of them your chief of staff.

13:54

So, the trade-off is pretty simple. Grok

13:56

Bot is easier to set up, but Open Man as

13:59

Bot gives you much more freedom to

14:01

choose your models, tools, and how each

14:04

agent works. A few things before you go.

14:06

If this video showed you the power of AI

14:08

agents, the next question is, how can

14:10

you actually use AI to make money? We

14:13

recently made a video showing 10 ways

14:15

people are using AI to create income

14:18

streams and automate their work. I'll

14:20

put that video right here on the screen.

14:21

And if you want to build your own bots,

14:23

I've also shared 100+ Grok Bot use

14:27

cases, prompts, and setup ideas inside

14:29

my free WhatsApp community. I've also

14:31

added the setup links, including the

14:33

resources you need to try for the free

14:35

alternative. You'll find the link to

14:37

join the community in the description.

14:39

And one more thing, around 70% of the

14:41

people watching this video are not

14:43

subscribed yet. So, if you want to keep

14:45

discovering new AI tools, workflows, and

14:47

ways to stay ahead instead of catching

14:49

up, make sure you hit that subscribe

14:50

button. See you in the next one.

Interactive Summary

The video introduces the concept of AI agents that can perform tasks independently by observing workflows, using digital tools, and collaborating with each other. It demonstrates how to build specialized bots for tasks like marketing, content creation, email management, and productivity. Additionally, it highlights advanced features like 'routines' for background execution and a 'Chief of Staff' bot to coordinate multiple agents, concluding with a comparison between Grok Bot and the more flexible, technical Open Manus Bot.

Suggested questions

3 ready-made prompts