I spent months telling you that the AI

bubble is going to pop. I still believe

that. But I was wrong about something

and it's worse than I thought. [music]

I'm not walking back my original AI

bubble pop thesis. I do believe that AI

is a bubble and I do believe that it has

already popped. You can look at any

number of my videos describing the circ

circular financing scheme where

companies with infrastructure compute

GPUs pay money to AI companies but with

the contingency that those AI companies

then have to spend the money on compute

on their compute. Nvidia is doing this

famously with a number of different AI

providers. And where do the AI startups

get the money? They get it from venture

capital people who are also leveraged

ostensibly in data center construction.

So it is a circular financing scheme

with the major hyperscalers now in their

guidance for 2026 all announcing huge AI

capex for the year totaling about 680

billion across the five largest ones.

This thesis is further reinforced and I

still buy it. I still maintain that the

AI bubble has popped and the finances

are just waiting to catch up with it.

You will note in my videos and a few

very astute viewers have mentioned this

in the comments. I don't say exactly

what's going to happen when the bubble

pops and that's because I didn't really

know until now. You'll notice again if

you listen closely to all of my videos,

I' I've been talking about this. I never

say that AI is totally useless. It

doesn't do anything. I get the

frustration with it. I'm frustrated with

it, too. It's changing things very

quickly. it seems like to suit the whims

of a few technology oligarchs that are

in control of all of our decisions

around work, how much we're paid, the

amount of work we do, those sorts of

things. I share your frustration there.

But you'll note in every video I say

that AI is actually useful. It's not the

end- all beall, but it is fairly useful

for a number of things. My background is

a software engineer. I don't do much

coding on the co on the job these days

anymore. I've been in leadership for a

long time, but it has reopened up my

world for working on side projects and

sprinting through some code to just get

some stuff to happen. Just cool projects

I've been working on. It's a blast. One

of my friends put it to me this morning.

He said, "It's like being 11 again and

discovering the internet for the first

time." So, I'm going to walk you through

my current position on these tools and

my current assessment. I'll explain what

has totally changed my mind and that I

totally missed that I discovered about a

week ago that changes a lot of

implications of what's going to happen

when the AI bubble pop fallout reaches

the mainstream. I'll talk about the

historical evidence drawing a parallel

between the AI bubble currently popping

and the dot bubble. Then we'll talk

about what that pop is actually going to

look like once it proliferates outwards,

who's at risk, and what to do to protect

yourself when it all comes crashing

down. You tried chat GPT, you tried

C-Pilot. What was your honest reaction?

>> Honest reaction when trying ChatGpt,

Claude, all of these LLMs at first was

from the get-go, hey, this is fairly

useful. It's not going to massively

change how I work, but it's fairly

useful. I remember about I remember

right around the time chat GBT 3 or 3.5

came out. It was in the airport. I was

at John Wayne Airport. I was waiting to

fly back to Phoenix and I was sitting in

the lounge and I'm there early for my

flight so I got like an hour to kill. So

I pop the laptop open and I was about to

get going on a side project and I see a

hacker news post announcing that

OpenAI's got this new chat model out.

Here's the link. So I go sign up and I

start trying it right away just

expecting kind of a you know typical if

this then that chatbot and I'm quite

pleasantly surprised to find that I can

carry on at least a basic conversation

with it. It has a personality. It can

answer questions about itself and it can

even drive the conversation if I ask it

to. But at this stage, it's, you know,

no more than a party trick, a curiosity

to be sure, but not immensely

immediately helpful. As the months go

on, we get access to this at my current

place of employment where I was working.

There's some new models that come out.

We get access to some of anthropics

latest stuff, and I start using it to do

just simple automation things. I have it

write some very basic bash scripts for

me, some Python scripts. Every once in a

while, I might ask it a coding question.

And for the most part, it does pretty

well until you throw an immense amount

of complexity at it. It gives a

straightforward answer or it generates a

shell script that deletes files in this

folder every day at a certain time.

Stuff that's relatively easy to look up,

but it's genuinely saving some time for

me around this point. Once these models

reached a point where they were able to

have complex and in-depth conversation,

I'd estimate probably around the 120 to

130 IQ level. They became immensely

helpful for researching, but I still

thought it was hype because essentially

this is just a chatbot. It does not have

true autonomy. It does not have a way of

working and functioning in the world.

It's just stuck in this chat interface

back and forth, back and forth. It's

starting to reach out into the world

through some coding tools like cloud

code or codeex cursor etc. But largely

this is just a better version of the

same. It's saving some time with those

coding questions. It's able to do more

complexity work a little bit more on its

own now but it ultimately just feels

like a better mousetrap.

>> What made you write it all off? What

specifically felt like hype? Like I said

a lot has happened over the past couple

of weeks. The biggest thing that has

happened, if you have been living under

a rock and you haven't heard about it, I

even have friends that are outside of

tech texting me about it is the release

of Clawbot, now known as Open Claw. It

uses existing models behind the scenes.

So, it uses models by Anthropic,

ChatgPT, whatever model you want to plug

into it. But it's an interface layer

with this model that totally changes the

paradigm of how it works. You could

choose whichever chat interface you want

to use to interact with it. I chose

Discord and I genuinely find now I'm

able to delegate very large autonomous

and recurring tasks to the machine and

it's done them to a very high level of

satisfaction for me. Another friend who

really urged me to try this tool said

something precient about it. He said,

"Josh, this is different. There is

something special here." And I think

that's very well put. That's how I feel

when I'm using Open Claw.

>> What can you do now in 5 minutes that

used to take you hours?

>> Well, as you know, the YouTube

subscribership has been growing

explosively for my channel over the past

couple of weeks. I'm happy to make

videos, but really at heart, I am a

writer. I am not a talker in front of a

camera. I'm doing this as a necessary

evil. I prefer to write. I've kept a

blog for a very long time now in

multiple different iterations. And so,

ultimately, what I want you to do is

read the blog. And the best way you find

out about the blog and when I have new

posts is through the newsletter. You can

sign up in the link down in the

description. How's that for a nice

subtle plug? But I realize my website is

not really set up in a way where you can

find out about the newsletter, where you

can easily access my blog. It's kind of

all over the place. So I open up

Discord. I shoot openclaw a message and

I say, "Here's my goal. I want you to

modify the website. I want you to make

it easy for people to sign up for the

newsletter. I want you to make it easy

for people to find the blog, find the

YouTube. This should be central command

for all things Dr. Josh C. Simmons so

people can find my stuff easily. No more

than 2 minutes later, it had made the

code changes to the website. It had

committed and pushed the code to my

GitHub repo and it had deployed it on my

hosting service. So by the time it

reported back a couple of minutes later,

the changes that it made that I didn't

really ask it to make, I just said to

improve it to meet this requirement were

done and in production. Now, you could

argue that's not rocket science. It's

not totally reinventing the wheel,

right? It's some front-end coding, a

couple of server side methods. It's

fairly straightforward for a competent

coder. That being said, I'm a busy guy.

I work a full-time job. I'm getting the

YouTube channel off the ground. I'm a

parent. I'm active in my church. Like,

there's a lot going on. And so,

something like that, a reconfiguration

of my website, that's going to be a four

or five hour task on a Saturday that I

don't really have these days. So, I

could just send OpenClaw a message to do

that for me. It's still what I want.

It's still driven by my direction, but I

don't have to get caught up in the

details anymore. And I think it was at

that moment I realized that my friend

was right. There really is something

special here. Things have changed in a

big way. And the very strange thing

about that is I I am a believer that I

think a lot of clerical work jobs in

offices, they're just not going to be

necessary 6 months from now. Once

somebody figures out how to set

permissioning on this, there's a couple

rough edges, but it largely does the

task today. Like if someone's job is to

compile reports out of other data and

write up a report every week, like

that's not a job that a human is going

to be doing in 6 months. And the reason

that I say that it is actually going to

wipe out a lot of jobs, which is

something that until last week I was

very bearish about. I did not think this

was going to meaningfully replace office

jobs, but it's going to replace most

office jobs because essentially what

they've done, remember they have not

made a new model. These are running on

the existing models that you can access

in Cloud, OpenAI, Google Gemini. It's

using those existing models. All they've

done is they put a framework on top of

it. And the way that I like to

conceptualize it is essentially they've

given these models a body, not a

physical body, but a virtual body that

can take virtual actions. Now, sure you

had connectors in Claude and Chat GBPT

and Google with their Google Workswuite,

but they suck, right? If you ask Gemini

to modify a Google doc, half the time

it's like, I can't do that. I can't find

it. And that's all in Google's

ecosystem. If you ask OpenClaw to buy

you stuff on Amazon with a prepaid gift

card, it's like, sure, champ, hold my

beer. I got you. And that's the same

model underneath, too. So, I truly

believe this is going to replace most

office jobs. And that's already

concerning enough because you you have

to start to think what do we what do we

do about people that are unemployed?

We need some kind of a universal basic

income probably. But that would assume

that our corporate and government

overlords have some empathy for the

situation. I don't think they do. So I

don't know exactly what that looks like.

But I do know that this is going to put

most people out of work in office jobs

inside of six months. I also know that

you may think, "Hey, but Dr. J, my job

is exempt, right? It's exempt. I work

with people. I'm a salesperson and I

work directly with people and that can't

be replaced with a robot. And I think

you're wrong long term, but I think

shortterm you may be right. But there's

a flag on the play. And hear me out on

this. If we got salesman Jim, your

name's Jim, you're salesman Jim, and we

got salesman Bob. And Jim, you're

staunch about not using AI. You're just,

I don't want to use it. I don't care

what it does. I don't care how great it

is. I'm in sales. I'm not going to use

it. And then you got salesman Bob. And

Bob is like all in on OpenClaw. They got

the stuff set up. They got it plugged

into their systems. Bob has done his

research. He's been watching this

channel. He knows what's going on with

these tools and where they're headed.

And they need to do research on leads.

Guess who's going to have the more

comprehensive lead research? It's going

to be Bob. Jim, your your time is

limited. You have to sleep. You have to

eat. And honestly, you can't work that

fast compared to a computer. If Bob

wants to spin up an analysis on Acme

Corp, who you guys are trying to sell

software to, he can say, "Open claw, you

know, go off, research Acme, get all of

their stuff online, pull their quarterly

earnings, tell me what their weak spots

are, what their needs are, where my

service could slot in. Furthermore, I

want you to go to all of the key

personnel's LinkedIn pages. I want you

to research all of the information you

can find on the web about them and build

a package on what they might be

vulnerable to in a sales pitch, what

they might go for, decisions they've

made in the past, and give me a

comprehensive uh briefing document

before I walk into that meeting. Bob is

going to have better intel than you.

That's not a value judgment. That's

using a tractor instead of a horse. So,

I'm not saying that this is the answer,

and we'll we'll get into who's who's at

risk, who's safe later, but this stuff

is going to change the workforce in a

big way.

>> In 1999, everyone said the internet

would change everything. They were

right. And investors still lost 80% of

their money. How? We're always relating

this back to the dot bust. And the

parallels are immense. I reference it in

most of my recent videos where I get

under the hood with this stuff. The

thing that you have to remember with the

dot bubble is that the tech built by the

companies that went bankrupt and

irrelevant out of that bubble built the

internet that you are using to watch

this video. We would not have YouTube.

We would not have tech companies. We

would not have AI if it was not for the

dot boom preceding the dot bubble and

pup. What we have now is a phenomenon

that's less like standing on the

shoulders of giants and more akin to

standing on the shoulders of ghosts

because remember out of that tumult you

had pets.com defunct out of business did

not capitalize on the dot bubble in the

correct way and then you have a survivor

like Amazon who embraced a new type of

funding model they were willing to

innovate they were willing to work

differently they survived and now they

dominate for better or worse probably

worse the e-commerce market. So you

probably be asking the question, you

know, Dr. J, how do you reconcile this

with your bubble popup hypothesis? Then

how are you now saying that AI is

actually going to provide tangible value

in the sense that it will operate for

the cost that AI operates and it will

replace human workers? How do you

reconcile that with a bubble? And it's

simple. Historically, the dot bubble

illustrates this, right? I just said it

earlier. The tech we're using right now

was made by a lot of companies that are

now defunct that went bust because of

the bubble. The same thing is going to

happen to AI and for the first time I

really believe in both things. I believe

that one the bubble has already popped

and the financing the funding scheme is

unsustainable. The figure we always like

to quote on this channel 4 cents on the

dollar. 4 cents on the dollar. That's

what AI companies are making for every

dollar they spend only 4 cents. That's

an unsustainable business model.

Anthropic's $380 billion valuation

proves this completely underscores that

point that the funding hype is

unsustainable. It is un un there's

nothing like it even compared to the dot

bubble. The funds are are bigger. The

bets are bigger. The risk is bigger. I

also believe that this technology is

valuable and it will replace a lot of

human workers like a significant amount

like way more than I thought 2 weeks

ago. And so the.com bubble proves that

both of those things can be true. That a

very [snorts] useful and worldchanging

technology is being developed and the

funding situation is completely un

unsustainable that is surrounding it.

Both of those things can be true and

they're true here for this AI bubble.

Everyone is arguing about whether AI is

overhyped or underhyped and they are

both wrong because of this third

position I've just outlined. Or maybe

they're both right. It's overhyped.

Financially, it is underhyped for the

technology.

>> Why is it crashes and it changes

everything a darker outcome than either

side is selling?

>> Because we don't know who's going to

survive or what the landscape is going

to look like after it pops. Will it be

open AI, Anthropic, Google? Will it take

your job? Is it financially feasible

enough to make 4 cents on the dollar and

also take someone's job? What are you

going to do if it does replace your job?

and the tech monopolies are still making

this kind of money. Who should pay for

universal basic income? It's darker

because we have now less clarity about

what is going to happen when this all

comes crashing down. The darkest part of

all of this is the question, who is at

risk? Because we are all at risk. Nobody

is safe from this and we can't do

anything to stop it. I tell you, I'm not

in my own echo chamber on this. I have a

friend that has a pool cleaning

business. I have a friend that's a

plumber. I of course have a bunch of

friends working in tech in a bunch of

different types of industries in tech.

And I talked to this about all of them.

And I'm always curious to know how much

they know about AI. My parents too,

which are now almost senior citizens.

And one of the things that people always

argue, and I see this argument on Reddit

and Blind and other places, is that if

you are working in the trades or working

in some kind of physically involved

industry, you're exempt from this.

You're safe. Your job isn't going to be

replaced. And that's not true. Nobody is

The only people exempt from this are

people that have enough money to not

work anymore and live off of

investments. Those are the only people

exempt from it. Otherwise, it's us

versus them. We are all at risk for

this. I don't care if you have a

landscaping business. I don't care if

you are writing reports that are never

read that kick up to management layer. I

I don't care. We are all in this boat

together. Because remember, either this

is going to fully replace an employee

and that's the the sort of

straightforward plug-and-play scenario

or what's going to happen is if you

don't use the tool to gain a competitive

advantage in your field, somebody else

will and they will outpace you. I saw

it's it's big out here in Arizona. I saw

when I was driving to church this

morning a billboard for trash can

cleaning. It gets to be about 115°

Fahrenheit in the summer. If you have a

trash bag that's a little leaky, it gets

some blood or fat from some meat you put

in there uh on the bottom of it and it

cooks at that 115° temperature. The

smell is incomprehensible. It is lethal.

It disrupts the brain waves. It's so

severe. And so a big industry here is

trash can cleaning. They come around

with a little mobile unit. They got a

power washer, some industrial soap. They

hose it out and then they pour it right

down the sewer, which is probably not

great. I've never had my trash cans

clean. That seems like a little bit rich

beyond my pay grade. Anyways, it's a big

business out here, right? And I don't

know anyone that has a trash can

cleaning business, but like I said, I

know people in plumbing. I know people

in pool cleaning. I know that most of

them aren't using AI at all. They some

of them abhore it. Other others just

don't know how to set it up. Don't know

how to use it past using it as a basic

chatbot. But if you are using something

that takes action on its own today, it's

open claw, which is open source. That's

not like a product you buy if you think

I'm hyping it. It's open source. So,

it's only a matter of time before

somebody takes this, commercializes it,

and then puts it in a rapper that

normies outside of software development

can understand and can use that. That

will happen within weeks or months. And

by the way, whoever does that is going

to make a bajillion dollars off of it. A

bajillion dollars easy. That's why

Anthropic is probably going to buy these

guys out before they can get some ideas

of their own. I digress. Back to our

trash can cleaning business. So, you got

these trash can cleaners, right? And if

I start a trash can cleaning business up

tomorrow, no offense to those guys, but

you know, you get the machine for it,

you get the power washer, and there's

I'm sure some level of craft to it, but

at the end of the day, it's cleaning a

trash can, right? Like, you can you can

learn and be get probably pretty good at

it within a couple of weeks. There's

going to be some gotchas, but it's

powerwashing a trash can. It's

low-skilled labor. I don't even think

trash can cleaners would get mad if I

said that. But if I get, you know, maybe

a small business loan or some friends to

bankroll, hey, I'm going to do a trash

can cleaning business. All of a sudden,

I'm I'm competing with those guys. And

then I start to do analyses of

neighborhoods that are more likely to

get their trash cans cleaned or pay a

higher dollar value for it. Maybe I do

it by income. And I pull MLS records.

Maybe I have a friend who's a realer

that can get me access to the MLS

system. I pull like home sale values to

figure out like where are the highincome

houses and neighborhoods and I start to

pump this data into the AI and I have it

do automated research, find out who

lives there, uh send mailing to these

houses, optimize the route so I spend

less on gas. There are many things I can

do to increase my level of competition

against these other guys and eventually

I will put them out of business or I

will significantly negatively impact

their business. And that's a field where

it's like, hey, I, you know, it's a

trade, it's kind of a trade. I use my

physical body. I'm I'm out there. A

robot can't do that. And it's like,

yeah, probably not today. Probably not

in the next 5 to 10 years, honestly,

with the sad state of uh humanoid

robotics right now. But the other things

that you have to do as part of running a

business, if somebody comes along and

automates those with AI, you are now

competing with one person who is working

with the capacity of 10 to 100 employees

that you don't have. or if you do have

them, you need to pay all of them a

salary. And we can sit here and hem and

haw and argue about the morals of that

and whether we should or shouldn't. And

that's certainly a debate we should be

having. But the thing I'm trying to

drive home, the thing I'm trying to

explain to you in these videos is that

this is happening. This is going to

happen. It is a fantasy to think that

this AI bubble pop will happen and then

AI will go away. There is no eventuality

in which that happens. So I don't think

white collar jobs, blue collar jobs,

whatever color jobs, none none of these

jobs are safe whether directly through

replacement by AI or indirectly through

the empowering your use case, drumming

up competition, doing better analytics

on where you should focus your efforts.

Those are just those are just the facts.

I really maybe there's one or two, but I

can't off the top of my head think of

anything that is totally totally safe.

Totally safe. Hit it in the comments if

you do think of something that's totally

safe. I would I would love to hear about

it, but I've been racking my brain and I

cannot think of something safe,

something exempt from this. Even if you

do use it, even if you do use it, it's

going to force multiply your output. So,

there's going to be less people in your

position. So, it's going to drive up the

competition. I think we're already

seeing this happening in software

engineering. All of a sudden, one

engineer can accomplish way more in a

day augmented by AI. They don't need to

write boilerplate code anymore. They

don't need to spend 5 hours tracing a

function call to find the origin of some

data. They can just ask the computer and

it can relay that information with a

high level of certainty. So here's what

I want you to do this week. Here's what

I want you to do this week. If you do

not have a subscription to one of the

major AI providers right now, I would

suggest getting one. $20 a month. You

can choose from Grock, Gemini,

Chat GBT, Claude through Anthropic. I'd

say go Claude, but these things change

pretty much day by day on which one is

the best forefront provider. So, just

pick one. Ultimately, they're all

improving at a massive rate. Pick one

and try to replace Google search for a

week and just see how that goes. See

what it feels like. See if it makes

things faster for you. You probably

Google a bunch of things every week. I I

had one earlier. I had the windshield

replaced on my truck. They got a rock

chip in it. They put a new windshield in

and it's one of the fancy new ones where

it does the uh it's got a camera in it

and it does the lane assist and the

adaptive cruise control where it keeps a

consistent distance uh based on speed to

the car in front of you. Uh it's not

it's not a Tesla, it's not a Cybert

truck, don't worry. But it's um it's got

all those kind of adaptive features. So,

they have to recalibrate the camera when

they put the new windshield in and the

active lane assist was off. And I kept

thinking about it, kept forgetting about

it. And so, before I pop in my truck to

head home from church today, I just I

search it real quick on Perplexity. I'm

like, I got a Ram 1500. Active lane

assist is off. Did they up the

calibration when they were doing the

windshield? And it's like, no, actually,

they they didn't. There's a big button

that you missed on the dashboard. It's

like right there. And I look and sure

enough, I'd been looking at the

dashboard the past like three times I

drove it and there's a huge button that

has a car going out of a lane and it's

got an orange light next to it. So I

toggle it and it's back on. It's good to

go. But my point is is like I hate

YouTube how-to videos. Like would I have

had to watch like a YouTube how-to video

of some guy saying, "Hey, I'm Joe Bob

and welcome to my garage. And back when

I was five, I got my first car." And you

probably are familiar with this if you

look up recipes, right? Even for the

time saving on recipes right now there.

That's a huge It's always I grew up in a

village in Greece when I was three and

my mom used to cook this recipe and it's

like blah blah blah blah blah. How many

cups of flour do I need? Just tell me

that in the chat GPT or whatever LLM

you're using. It'll tell it to you right

away. So just try to replace Google

search with it for a week and just see

what kind of benefits you get out of

that. If that all goes well, you'll

start to develop a facility with how you

speak to the AI. This is some degree of

skill. It's not extremely highskilled,

but you do talk to it a little bit

differently and you learn as time goes

on. But you need at least a base level

facility with those skills because when

these things like OpenClaw that can

actually go out into the internet and

take actions for you autonomously start

to become common, which like I said that

that is weeks away that you will have

access to an iOS app, a web app that

lets you speak to an LLM that will take

actions on your behalf. uh and and I

mean this is a couple of weeks away. If

you are like a non-s software person,

non-coder, you will have that very very

very soon. And when it comes, you want

to be in a place where you know what

these models are capable of, how to tell

them what to do and how to process and

second guess some of that information to

make sure you're getting the most

accurate takes. Do start that this week.

Start that this week. This is one of the

most important skills you can develop

for your future career. Remember, I've

deployed AI systems at scale. I've

overseen 100 plus person departments.

When I tell you that this is different,

it's not hype. It's not like some PR nut

job or finance bro, forgive the vest, is

telling you this and they don't know

anything about the internals, how it

works. They've never really use these

tools in a meaningful way. I am telling

you that as of last week with the

release and the development of OpenClaw,

it it is different now. It is different.

This is a real emergency. Before we

wrap, if you remember a better time on

the internet, my friend in Michigan is

making these hats through abandonware.

The link is down there in the

description. They don't pay us or

anything like that. I just I like the

hats. It's a a dude I know in Michigan.

Go buy the hats. It reminds me of when

the computer was a happier, more sane,

more human time when a Carda 95 was

around. Make sure you subscribe to the

channel. Hit the bell to be notified

when new videos drop. As I mentioned,

newsletter sign up down there below. Get

all the facts and figures updated

quarterly behind this. And I want to

hear about it. What is your job? And are

you worried about being replaced? Tell

us in the comments. Let's get the

discussion going. Thank you for

watching.