You see this face?

That's the face of a man who just made a

prediction on the internet for tech and

got it right. Let's just pretend we're a

big tech Google company. It costs

$50,000 a month and you're spending $1.3

million a month on just AI agents. To

replace those with just engineers would

Well, that kind of math you I mean it's

a number that kind of math you can't

just do off the top of your head. So

let's just say 30 engineers. That's like

30 engineers worth of people working on

something. You can't just do this for

every single project. Your company at

some point going to go, "Okay, time out.

We've made a mistake. We have decided

that we let you use all the tokens you

want. That's bad. We're going to go back

to the old days. Who's the most token

efficient?" And that my friends, that

prediction came right just a short 8

days later. Oh, what do you say? What is

this right here? Oh, is that Is that Sam

Altman saying something? He also said

the cost question came up quite

suddenly. At the beginning of 2026, the

issue never came up, Altman said. People

were totally happy with the amount they

were spending, he said. Now AI costs are

a huge issue.

>> [laughter]

>> I have correctly predicted the future in

only 8 days. And so obviously off this

high, off the high of the Uber COO

saying, "Yeah, token max token maxing. I

can't justify it. I don't even know I

can't even tell what features are being

made with this token maxing." But

nonetheless, after this amazing tech

prediction, I have decided that I'm

going to make five more predictions. Now

one of them's going to be heartwarming.

One of them is going to be what I think

is a good outcome. And then the other

three, nightmare scenarios. I think all

of these are going to come true within

the next 6 to 12 months, okay? And and

here's the other thing. None of these

predictions involve AGI, all right? They

all involve what seems to be the most

obvious answer of all time. Because when

you think about it for a second, like,

how did Sam not see this coming? Did he

never work at a company? Has he never

worked at a company where you had to

argue to spend $50, but then all of a

sudden every employee can just spend 10

grand on tokens? Brother, do you not

know how companies operate? How did you

not know this? Like, did you not have to

go to a VP and argue why you actually

needed an extra 8 GB of RAM? Like, did

you not live a normal ass life where

every last little thing was scrutinized

in a $1,000 meeting to make sure you're

spending $50 correct? Because that, my

friends, that's the corporate lifestyle.

Prediction number one, some of these AI

labs, they're actually going to attempt

to trade tokens for equity.

Oh.

Oh, Yes, this actually is

happening. This is not a prediction. I

just thought I'd throw this in there.

Just

This is insane, right? Okay. So,

actually token prediction number one is

going to have to be open source

donations. You, yes, you are going to be

able to donate a million tokens from

your budget into open source so they can

run CI or these other various tools that

may have to be run in some sort of

automated fashion, which costs open

source projects too much money. So, you

can be like, "Hey, have 10 million

tokens on your friend, okay? 10 quick 10

mil for you. By the way, those 10 mil,

those are Kimme Kimme 26s, okay? I ain't

paying for I ain't paying for no Chad

Jibity 55s, all right, buddy? And I

believe we're going to actually see this

within just the next few months. Open

source token donation. Yes, this one's

kind of nice, you know? Like, oh, that's

this is kind of sweet, you know? People

out there trying to make the world a

better place via tokens. Okay, that's

that's kind of nice. It's something kind

of like protein folding, you know how

you could donate your computer? Well,

now you can donate your tokens, maybe to

cancer research.

I know the frontier modelers keep

telling us how they're going to do this,

but maybe you'll have to be the one

donating your tokens for that to happen.

But either way, token donations, highly

likely we're going to see this in the

next few months. I had to get that one

out of the way because that's the nicest

of all the predictions. The next

predictions are going to progressively

get more and more deranged. But first,

big thank you to the sponsor.

If you have not tried Composer 255 you

have to try it. Let me give you a couple

reasons. Well, first off, it's like my

favorite model, it's the one I use the

most, it's the one that I can do the

easiest back and forth with because it's

actually fast. You can actually have a

back and forth with. Second, the output,

it's really, really good. But third,

something that I think is super

important, especially today with all the

news and token cost and everything, is

it's actually an affordable model. You

can do research. You can do debugging.

You can get a whole a lot done without

breaking the bank. So, check out

Composer 255 from Cursor.

Thanks, Cursor.

Now, back to the video.

Okay, hey. So, number two is kind of a

bit of an interesting one. I think that

when companies start hiring people,

they're going to of course give them the

dollar bills, right? And then they're

going to also give them the stonks,

which of course actually just simply is

dollar bills that grow. Then, you know,

they also get like 401k matching and all

these other things. I know that's very

American of me, but you get you get some

sort of benefits that go along with it.

But I think there's going to become a

fourth category that's going to happen,

which is token budget. You're going to

be given a yearly token stipend. And

however you choose to spend that at your

company to get your work done is on you.

And if at the end of the year, your

token spend is less than your token

budget, hey, you're going to have a

bigger, fatter, nicer bonus at the end

of the year because you're getting your

work done and you're not costing the

employer too much. This is going to

obviously incentivize people to be more

efficient with their tokens. Oh, you're

like a super-duper ultra engineer and we

whole super believe in you. Here's a

100k budget. Blow it all you want or

be super efficient with what you do, you

get to keep some of that money.

Incentivizing good behavior. Of course,

this is a once again one of my nicer

predictions, but I can actually

definitely see this some sort of

incentivization

for your behavior not just blowing

tokens like wild. Okay, we do not need

100 more dashboards. Like maybe just

maybe planning things out and thinking

about things ahead of time might be the

best way. Like dog, I know you got 10k

more in you, but maybe you don't need to

get those lines out right now, okay?

Hey, take a step back. It's okay, buddy.

All right, we're starting to enter into

the unhinged territory, all right? This

is where things are going to get a

little um

strange. Prediction numero three. You

remember planning poker? You remember

Agile? Remember, you know, back in the

day long before vibe coding? You have to

You used to have to like sit down and be

like, "Mhm, this task is a is a medium."

And then everybody would play this game

of poker, which by the way, set it

aside. That's not poker. I hated the

fact that they called it planning poker

because what are people bluffing? Are

people actively lying to me? I'm pretty

sure this is called "Hey, get guess a

number as to what you think this is

going to be." And if everybody guesses

the same number, that might be just what

it is. Like it's not That's not poker,

okay? That's not poker at all. Hated

that, okay? Sorry. Either way,

we're going to see something a little

bit different, okay? My little vibe

coders, my little sweet summer children,

I know you're not familiar with that

process, but

new process is going to land and it's

going to look something like this. All

right, hey, we got this feature coming

up. We're going to play a little bit of

token poker.

Uh everybody get your cards out. How

many millions of tokens do you think

this this feature's going to take? Ooh,

I think this is I think this could be

done for 10 million Kimmy 26 tokens.

Throw that throw Throw in the hat. Some

people are going to be like, "Dude, this

is 50 million gippity extra high fast

tokens, okay? Cuz it's urgent. It's

going to cost us money." You're going to

have this entire session where people

are going to be playing poker. Which is

still not poker. About how much a task

is going to cost in tokens. You know I'm

right. You know I'm actually right. In

fact, I would not be shocked if someone

said, "We already do this. We already

token poker. We already But we already

put them token poker chips down, boy."

Like I already I believe it. I would be

I'm completely not surprised. But we're

going to replace Agile with token Agile.

Oh my gosh, the consulting class is

going to be so fun. Think about all the

consulting people that are going to come

into your company, reintroduce a new

version of Agile, and one of them is

definitely going to be token poker. All

right, so the final two predictions I

think are going to be the most likely

ones you're going to see pretty soon.

And they're going to probably re- like

wrought the most destruction on any

company. I believe these next two are

effectively what George Hotz was talking

about when he said, "I'm calling it now.

The adoption of AI agents into software

development will be one of the most

costly mistakes in the field's history."

Now, I don't think he's saying agents

are up here bad here. I think he's

saying there are some companies in which

they operate that are going to be

absolutely insane to budgets and just

just actually delete companies

completely out of existence. Prediction

number four I think can be best

represented by Microsoft's org chart.

Which is going to be instead of having a

company-wide budget for AI, which I

think will still exist, you're going to

have org-wide and team-wise budgets. And

this is going to cause people to lose

their damn mind. Because you know what's

going to happen. There's going to be

individuals out there just spending an

enormous amount of tokens, and their

whole team is is to just hate that

person. Or they're going to use their

entire team's budget and have to go back

to like hand coding and just absolutely

being so angry at what's going on. And

you know also what's going to happen

with this idea of kind of

this this these organizational wide

token spreading is there's going to be

middle managers that their entire job is

to like petition {slash} negotiate for

your team's allotted token limit. An

entire management class is going to be

created out of this. It is going to

it's going to be absolutely beautiful.

And then of course the natural

outcropping of all of this is that there

will be team members in which to save

tokens are going to start pair

prompting. Yes, you heard it here first

folks. I predict by the end of the year

you will definitely hear that phrase

paired prompting. They're going to get

together and really kind of hammer out

the perfect context {slash} prompting so

that they don't use too many tokens to

go and get the task done. And even

worse, there's going to be like these

team long activities, right? Cuz you

know how it's you you can technically

set up these agents to run for like a

day straight or 6 hours straight or

multiple days straight and they can just

spend so much tokens. So what you're

going to end up happening is these long

running tasks, the prompts to kick them

off will be like team reviewed. You're

going to have people actually on GitHub

knitting how the prompt was set up being

like, "Actually, you don't need to

threaten the grandma anymore. And

actually it's best to give them a

certain personality and then somebody

else is going to be like, "Actually, on

archive paper number 3,972,

it actually says personalities are

destructive." And it's just going to be

this non-stop shaman-like exercise where

they tell you the secrets of the prompt

and your whole team is going to be just

arguing over how to kick off these long

running tasks. It is going to be it's

just a thing of beauty. All right, the

fifth and and final prediction, which I

think is definitely the most unfortunate

because this one, if it's not already

true, will be true here very, very soon.

Which of course is that companies are

going to try to reward as hard as they

can the people who are using the most AI

by budgeting the AI specifically for the

people. So, what's going to end up

happening is they're going to review all

the get logs and be like, "Okay, well,

Brian over there, he only produced

10,000 lines of code last month, but

Timmy

Timmy produced 100,000. You know what?

Tokens for Timmy, but none for Brian.

Okay, nobody loves Brian, all right?

Everybody loves Timmy. You know this is

going to happen. They're going to find

the people just simply creating the most

change gets the most budget. And the

thing is is that this isn't even that

far from happening because if you

remember ClickUp from doing some of the

layoffs, one of their big things is the

100X organization. They're like they're

like setting forth kind of this idea

that there will exist organizations that

have 100X output. Well, how are you

going to get 100X output? You're going

to have to have an enormous budget, and

once obviously the budget hits the

fan, people are going to go, "Okay, we

need to reduce who gets to be a part of

this organization, who produces the most

code?" And it's just naturally going to

happen. They're going to find the

biggest slop cannons, and they're going

to put them into positions where they

get effectively unlimited budget. I

genuinely believe there's going to be a

company that is going to so thoroughly

code themselves into hell that they're

actually going to fail as a company due

to how much crap and burden they have

put on themselves with the promises of

the magical

sand rock computer talky thingies. I

don't even know what happened there. I

just got I got all my words mixed up.

All right, okay. I was coming in with,

you know, like the the magical sand that

talks to you. I don't even know what to

call it. What do you call it? They're

getting They're going to get owned

because somebody told them that the

magic rocks actually can predict the

future and will in fact lead them to

salvation and instead it just led them

to hundreds of thousands if not millions

of lines of burden and not technical

advantage. The name

is the primogen.