hey welcome to cloud unplugged i am here

with salman a new co-host uh do you

want to quickly introduce yourself salman

I feel like I need to have the

mic that was in the intro,

but I don't have the mic in my

hand.

So yeah, hi.

My name is Salman Iqbal and I'm co-hosting

Cloud Unplugged with the one and only John

Shanks.

It's the first time for me on a

podcast as a host.

So let's see how it goes.

I'm excited for this season.

This is good.

It's going to be good.

We've got four news topics.

We've got the Anthropic Mythos and the

Glasswing Project.

I think that's something you're going to

talk about, Salman.

We've got the Vassell Breach.

and what that means for AI,

but really was it AI or was it

more supply chain?

We'll come on to that.

Tim Cook leaving and what it actually

means to the strategy of Apple overall in

the AI

arms race that's going on.

More AI again, Andon Labs,

a researching company who have basically

created a shop all based on AI driving

the business on its own.

And then we've got your sneaky story,

which we'll kind of come on to as

a surprise outside of mine as a last

day,

which is something you're going to talk

about.

We'll kind of crack on into that a

bit later.

Do you want to kind of kick off

on the Anthropic stuff and Mythos

Glasswing project and what's been going on

there?

Absolutely.

We all care about security in software

supply chain.

Security is very important to all of us,

especially us working in the cloud as

well.

So Anthropic a few weeks ago announced

this project Glasswing,

which is a closed invite only project for

a bunch of companies, including Computing,

Cloud Native Computing Foundation,

and some other companies as well to invite

them to work on and improve the software

security supply and looking at things like

if you can figure out if there's issues

in your software supply chain.

and it uses a special model that they

specific model that they announced called

mythos or mythos i think mythos that's

what they call it mythos and the position

it got announced as a a massive fanfare

and that has a major big breakthrough in

ai cyber security and a lot of security

companies as as is always the case with

the announcements are anthropic recently

every time there's an announcement in a

specific area the stock plummets for other

companies

So they released it.

They found a lot of vulnerabilities in

FreeBSD, in Mozilla,

in a bunch of other frameworks that have

been out there.

Some vulnerabilities that didn't exist a

long time ago as well,

like they've been in the ecosystem for

twenty or thirty years.

So it managed to find those

vulnerabilities and report on it.

And there was a FreeBSD as well, OpenBSD,

all of that stuff was there,

which is good.

There's great news for the ecosystem.

We have a new system that's come out.

But the reason why I bring this to

the table is when it got announced,

the closed sourceness of this thing,

it creates a bit of an imbalance between

what the big companies are doing or maybe

perhaps what the little open source,

smaller projects will do.

That's the high level.

But then also some companies manage it,

some security companies manage to.

use different models to try and recreate

the same issues so the mythos is closed

source is a bit more expensive so i

thought it'll be interesting to discuss

what our take is on this and i'll

be interested to hear from you as a

veteran of the cloud and security space

just to hear your inputs as well

So this then, like the whole mythos thing,

are you sceptical?

Do I detect a bit of a hint

of scepticism?

Scepticism, is that right?

On whether or not this is more of

a promotional PR stunt?

Is that what you're thinking?

Or do you actually think it is?

I'm not saying it's not.

Tell us the truth.

I'm not claiming that it's a promotional

PR.

I mean,

it's a great piece of work that they're

doing in the ecosystem.

It's an amazing thing that they're trying

to do.

But sometimes there's a bit more hype than

there is.

For example,

anything that surrounds Anthropic

recently,

they announced

cloud design where you can just design

stuff automatically.

Forget about Figma.

That's what the chat was on the social

media.

Forget about Figma.

And now you can just do everything with

cloud design.

And of course, when that got announced,

Adobe stock went down,

Figma stock went down and everybody jumped

on it.

I wanted to give it a go.

because i make animations for my videos

but i'll be pleased to say that uh

i've burned all my credits within thirty

minutes because it's a has a different

credit limit but i didn't get what i

wanted and it made me a llama which

didn't even look like a llama uh as

a design

It's some like cartoonish thing that you

draw in MS Paint.

So I'm saying it is an excellent project

where you can find,

going back to Glassman,

it is an amazing project where there's a

lot more work to be done.

And that's why, hence,

they asked so many companies to

collaborate with them.

But sometimes these things are a bit

overhyped.

Aisle is a security company.

They used some of the open source tools

and also some of the closed source tools

that you have,

but not as expensive models

that they managed to find the same

vulnerabilities in the system.

And they've been doing that since last

year, since twenty twenty five.

They've been looking at these

vulnerabilities in open BSD, free BSD.

And they and they did they did announce

them.

Of course,

No such fanfare was attached to their

announcement than it was to Anthropix.

I'm guessing because a lot of companies

were involved with them.

Plus it's Anthropix.

So there's a bit of hype there.

Yeah.

So I kind of was reading about this.

There was a bit of a funny thing

because we've done a few different stories

as in there were

one again this is a bit of my

this is probably my cynicism but on the

pr side they were like kind of talking

about what it might mean to security

companies um you know like oh it costs

like twenty thousand dollars to like test

your software we did this for like fifty

dollars on a token so i was like

okay that's like one spin um is like

how much cheaper and effective it can kind

of be but

what i read though i think it was

to do with out of how many attempts

would the model actually take the its own

um free will i guess to a point

on the exploitation like if it found

something would it actually write the code

to exploit it or would it just be

like highlight the fact that it's found

something that has like maybe a

vulnerability and under the opus one like

four point six

you could run it hundreds and hundreds of

times and only twice did it ever produce

um actual exploit whereas with mythos it

actually like a hundred times more likely

to have generated an exploit for it not

just highlighted what it's found but

actually then chain a set of events i

think it was actually it looked at

multiple pathways

yeah and then found multiple ways to kind

of like change things together to get an

exploit so some some things were quite

complicated um and then it would actually

take the initiative to then obviously

write the exploit um as well so that

was kind of the concern i think which

is like well it might mean that you

know you kind of raise an exploit i

i guess the thing i have a bit

of

beef with the most i think with this

is um it being trying to be like

a positive and a negative all at the

same time so they're like gating it

through a pr event of like let's make

let's create a closed group let's call it

glass wing the glass wing project we need

like people to support us on like

validating the model on that side but on

the second side

their stance is, well,

this is great because you can now find

vulnerabilities really easily in your

software and actually it's really powerful

and we should be security testing anyway

and blah, blah, blah.

But

Actually,

the barrier for entry has just dropped.

So it's like, well,

there's a difference between a security

specialist that's literally quite niche

and somebody,

I don't know comparatively to the model,

but probably a unicorn of an individual to

be able to do what Mythos could do

is rare.

You'd have to probably spend loads of

money.

You'd have to try and find them.

There's probably not many of them around

versus now you could just be anyone that

goes on and just like get me an

export.

So the barrier for entry is reduced.

There's no culpability.

There's no legalities around it because

the laws haven't changed.

It's on the person who does the exploit,

not on their company,

even though they've reduced the barrier

for entry.

So all of that side of things,

more from a regulatory and legal stance,

Feels like, I don't know,

feels like government or body should be

responding to it saying, okay, hey,

just a minute,

you're about to drop the barrel for entry.

Almost anybody, you know,

an eight-year-old could probably come in

and be like,

get me an exploit to my,

I don't know, this website,

or go and hack my school's

website and school systems or whatever

else and it's like yeah well you know

it's not our fault it's your kids uh

maybe you should uh you know what i

mean it's like it's that place for your

newly children you're saying that that

mythos and project last week can basically

become like the red team and just go

instead of you hiring because hiring a red

team and doing this is quite expensive

that's exactly what you're saying right so

that's what,

that's kind of what they're saying.

Even they're like, cause it's like,

Oh my God, look, you know,

this is like,

it would normally cost you like,

it would normally cost you like,

to get this, like pen testers in,

or do something.

It's like, just for fifty quid,

fifty dollars guys,

you can do the same thing.

So that, they're kind of spinning it,

but through an angle of like,

business lens, but actually,

their behaviours on it, is very different,

to the promotional aspect of it.

Do you see what I mean?

Like the way they've, the,

And so it's like on one side,

you're behaving one way,

on the other side,

you're saying something else from a PR

perspective.

And on the other side,

from a government lens, you're like, well,

what's everyone else doing?

Do you know what I mean?

Like as in who else is like what

from a regulatory perspective,

which I'm not saying like we need to

regulate everything,

but what does it mean when you make

everything so accessible that means that

you could just compromise something quite

easily?

But is that any different to any of

the other tooling that already exists out

there?

Yes, the barrier to entry is high,

but over time,

a bunch of tools have come out that

software engineers can also use.

A simple example like Wireshark.

When Wireshark didn't used to exist,

you can't even see the packets or the

network.

so does that mean that even then the

person who had a bit of knowledge about

software they could start using these

tools and start hacking or doing the red

team efforts would they be culpable is

that are they culpable well that's the

that's the stance isn't it i think because

it's like it's that's that that will be

their argument you know really that's that

would be the exactly what they'll say is

that well hey just admit this has existed

already we're not doing anything different

Because we kind of are,

because literally I don't need to know

anything.

The same way as I could go to

it today and be like,

write me an email in Spanish.

I can't speak any Spanish.

Or literally,

I might just know to say hello.

So there's a difference in trying to

understand the data of packet inspection

on a network and actually understanding

what it really means.

versus just never really even having to

understand anything at all.

And then even anything you don't

understand, you just pass it back.

And you're just like,

can you deal with this?

Because I don't understand it.

And on it goes in a chain.

What do you think about it anyway?

Do you think I'm being a paranoid loon

or something?

Well, the thing about security,

how I look at it,

is it's like layers of Swiss cheese

because it has lots of holes in it.

This is the analogy people use all the

time.

I think this is just yet another tool

in that, oh,

I need to have code isolation.

I need to have automated scanning.

I need to have validation.

Yeah, great project when it exists.

It's an excellent tool.

It's an excellent model,

as you talk about.

And when they did their test, yes,

they found some of the models weren't as,

as you were saying,

weren't as good at finding these

vulnerabilities.

But some like, I can't remember, GPT OSS,

I think version four, version five,

whichever one it was,

it was good at finding it.

So I think all I'm saying is,

These tools already exist.

Claude already had a security scan too.

They already exist.

Sometimes we get caught up in the hype

when Anthropic released something and

everybody else is like, this is it.

We don't need any more security.

Everything's automated.

You don't need to do anything else.

But all that stuff still applies.

All the OWASP practices still apply.

Yes,

this tool will give you an earlier

visibility of something that's

perhaps gone wrong or has been going wrong

for a while.

You didn't catch it.

But all the good practices,

they don't get thrown out of the window.

They still have to apply.

You still have to have all those

practices.

So I think all I'm trying to say

is that we shouldn't fall for the hype.

I fell for the hype.

I just thought,

I don't need to do anything for security

anymore.

I started looking into it a little bit

more.

I was like, okay, yeah, fine.

I understand.

Yeah.

I guess it's not just falling for the

hype,

but it's the hype plus

you know, like the rate of change,

plus it being a bit like a,

it's gangster, right?

It's like protection money.

You know now that they're probably going

to bring something out now that's like,

you know,

the clawed security tool that like helps

you against other models trying to hack

you.

So in one side of like got a

model that could hack you,

on the other side they've probably got an

agent or some other tool that you also

can buy that kind of detects the hacking

or something.

So it's like basically,

It's like pay us the money or we'll

rob your house.

But if you pay us the money...

yeah and you're like that's basically

what's going to become it's a really bad

thing there's nothing like that at all

because i do really like it so i'm

like obviously slacking it off but really

use it every day but anyway that's just

how it uh i should i should put

a disclaimer i love this stuff anthropic

too right cloud code i use it all

the time but like for example this mythos

model uh we don't there's some information

out there opus costs five dollars

per million tokens for an input and the

output is like twenty five dollars per

million token but the mythos cost is

twenty five dollars per million tokens for

input and hundred and twenty five dollars

per million tokens for output going to the

point that you're making that yeah it's

like oh we can we can give you

this but it's gonna cost you a little

bit more but then you also have to

balance with

What do you have?

Do you have an alternative for it?

Can you do this yourself?

Maybe that cost is reasonable.

Exactly, it's fair.

Yeah, I mean, it is amazing.

That's the thing.

It's amazing, but it's also scary.

And it's also like, you know,

it's got the touch of everything all kind

of going on at the same time,

isn't it?

But it's kind of funny.

But I guess leading on to security,

Vassell,

Vercel got hacked kinder,

but through really context AI.

And I actually, when we were speaking,

I didn't realize until you were kind of

mentioning it to me recently that actually

it was context AI that got hacked and

then Vercel run the tail end of it

because of an employee.

So actually,

I don't know if you want to talk

about it because you know,

you obviously go to smoking is what we

call it usually.

So Vercel is a cloud platform and you

can host your applications,

your websites and whatever you want.

It is basically a cloud platform that I

have used in the past.

and a lot of people that i know

are vibe coding stuff nowadays are using

that platform to host their websites and

host everything else which is it's an

excellent product and i i've used it and

they have very reasonable pricing for the

hosting so whatever happened it's not it's

it's it's it's unfortunate that

couple of weeks ago and attackers accessed

the internal system and customer data but

not directly but via the context.ai

third-party tool that an employee was

using so an employee was using context.ai

tool using an OAuth token so they

authenticated

using the Google Workspace to use that

tool, whatever they were doing.

And contacts there got hacked,

and then whoever the attacker was managed

to access the environment variables,

the API keys,

anything that was in the Google Workspace

for OAuth, for Vercel.

So they managed to get some information.

customers were affected some some weren't

uh they did come across you know straight

away they were very good and they were

very open i was following that on twitter

um and their founder as well they told

people what needed to be done that you

can you can rotate your tokens and things

like that

But again, there were a lot of people.

They were like,

what does the rotation of token mean?

Because the barrier to entry is now low,

which is great.

Anybody can use this stuff.

So the language sometimes also needs to be

on par with what people can understand.

And people are spending the weekend.

It happened last weekend.

People are spending the weekend rotating

rotating uh uh secrets i just deleted my

account even though i didn't have any

secrets on there for now remember we

talked about security you need to have a

kill switch i hosted it somewhere else um

but that's that's what happened basically

what this shows is you

Vercel will have amazing security

protection against security threats and

attacks that might be happening.

We just talked about this security in

layers.

But a single integration can cause a

massive blast radius.

So this is something you can get

compromised like that,

which I thought is interesting to discuss.

Maybe what's your thoughts on this?

What do you classify this as?

Because obviously you get the supply

chain,

which is like normally from pulling down

other software dependencies and from one

down into the other.

And obviously it's like a Russian doll

effect of like lots and lots of things

containing other things and it all bundles

down.

And somewhere within all of that chain

could be like a vulnerability.

Whereas this is not a traditional supply

chain.

It's like...

But it is a chain in itself, obviously,

of suppliers, I suppose.

So I just wonder,

is that what it is still deemed in

the industry?

Is it still supply chain?

It still is.

I read it and they're calling it the

modern supply chain attack.

Okay.

More than infrastructure.

That's how it's been framed.

That's that's how it came across because

context is another tool.

And, you know, that's how it's been.

It's still a supply chain attack because

it is in your eyes.

Yeah, it is in your chain.

But it's

there's a lot of people,

even I use OAuth to authenticate against

various systems in various places,

but you kind of have to keep reviewing

these tools over time because of this pace

of change, John, you know,

like you touched on that before.

I think it's a little bit

hard to keep track on what's going on

what's going what's wrong it's just you

just have to be a little bit careful

all the security principles that you have

before they still apply uh you know this

is uh is it yeah so this is

i don't know what your thoughts are on

this uh i find it a little bit

like i guess there's like two bits on

it for me is one is like slightly

self-reflective of

how lazy i'm becoming as in like i

really don't like much effort anymore for

anything you know and it's it's

incremental you know it's like an

incremental thing where um you know

installing software on your machine to do

like things and then you're like oh now

it's in the cloud i don't even need

to install things it's just like there i

go to a website and then

know food delivered to your door and you

know all these like things and then you're

like oh great just anything single sign i

don't have to have loads of passwords you

know and then now you're into ai and

you're like oh great i don't have to

really keep uploading documents and my

images just give you access to my google

drive i'll give you access to one drive

like i mean you you find it it's

there rather me move it paste it so

then each time that window slides of

convenience

then the more risk kind of you are

exposed to really, because it, it does,

you know, there's like a, like a,

I guess a multiplier effect.

The more times you're doing that,

the more likely you have more things that

have more access.

And then I couldn't even, you know,

I reckon if you ask most people like,

okay,

How much single sign on, you know,

where's your identity like signed into in

this moment in time?

Like how many systems do you reckon you've

got a single sign on token access?

That's probably got some privileges maybe

on the claim.

And people probably wouldn't even know.

They wouldn't be able to tell you.

They might just think three or four and

then you'd have a look and it'd probably

be like, like,

twenty-five or fifty or something.

Yeah.

Yeah.

You know, so that's,

I think that's the scary thing.

And then auditing, that's really hard.

I think from a, so yeah,

I think it's great on one side because

it's so convenient and makes everything

easier on the other side.

guess this is the the risk isn't it

it's you just one exploit on that token

that someone's then got access to um

before you can start scooting about um

their estate and finding things out yeah

as them yeah i know it's interesting you

say you're becoming lazy but some people

argue you're becoming more efficient

because you you don't have to just single

sign on everywhere right

If we were to play the Anthropic Playbook,

we would be saying it's an efficiency

game.

That's exactly what it is.

Efficiency game.

That's exactly it.

It's an efficiency game.

Messing.

And then, so another story then.

Tim Cook,

probably a little bit of a curve.

We're not so, like, technical, you know,

technology-based, AI-based.

But I found this interesting only for the

fact that...

Tim said he wasn't retiring or anything

four weeks prior.

And then John Ternus has obviously been

announced that he's now going to be the

new CEO.

And then Tim Cook's not really stepping

down fully.

He's still there and transitioning over.

But it obviously feels more pointed

because of John Ternus's background being

in hardware and

who was responsible for the M-Series

silicon program and rolling that out,

which is obviously a massive success.

So it just kind of gave a hint

into their strategy for them as a company

that they're aiming more for the devices

and let's get a CEO that's hardware-driven

in charge who can kind of drive that

strategy and think about how can we run

models on phones and how can we compute

things locally on devices.

So that felt like quite an interesting

opportunity

thing um and on the other side you

kind of got their competitor which is

google as well and they're kind of owning

everything you know the cloud phones the

models you know and they're like so it's

like just a different play but anyway i

don't that was that was just more more

amusing on an interest amusing on an

interest of like oh kind of interesting

that they've made that decision now

well yeah i don't want to read too

much into it but tim cook was from

supply chain right so i think he was

a supply chain person before and turnus as

you say is uh he's a hardware person

and you know recently gemini four uh

google announced gemini four a bunch of

models within gemini four some of them can

run on a phone right yeah oh it's

the gemma one

uh gemini uh i i one of the

german ones yes yes but gemini you know

the google gemini model is one of one

of them right on the phone perhaps what

you're saying is right so i think they're

focusing on because now you can you can

either you have there's so many models out

there right

but hardware is only a handful of

companies that are making hardware that

can run this stuff so perhaps that's a

bit of their strategy as well to maybe

improve because the m chips are amazing

for a bunch of stuff people couldn't even

buy the uh the the mac minis because

everybody's running open claw on it i know

you have a bunch of automation running the

house doing your laundry and stuff like

that and oh yeah yeah it's all robotic

I know everything is robotic for you.

I know you got those running in your

house.

So I think Apple saw that with what

happened with Open Claw.

It got sold out.

The minis got sold out because of the

capabilities.

Perhaps they're like,

we need to double down and maybe work

more into the hardware.

Not reading too much into it,

but Nvidia is also doing like, okay,

Nvidia is just...

They're the biggest player right now in

GPU land.

So I think they might be onto something

here with even bigger market share

capturing.

Because phones, they've got it, right?

Fine.

Big market share.

And also devices too.

They're getting it.

Perhaps they are maybe more into maybe

some stuff for the data center.

Maybe they'll release their own M series

GPUs.

Yeah, exactly.

I think it is.

I think it's more if they can...

if they can get the hardware exceptionally

powerful to run models.

Because what they've been phenomenal at is

obviously the size of the phone,

the design of the phone,

and then the capability on the chip.

So that's always been their strength.

It's like you're saying, the size of it,

from a consumer perspective,

the accessibility of it,

the convenience of it.

and so it does feel like as they've

not got into the model building side that

their players let's not waste time

building models right let's people are

going to do that and let's yeah let's

focus on the hardware and let's see

probably i mean i'm just speaking as if

i know i'm just making all this up

about all this is like a hypothetical

strategy but you'd think that they'd be

like well

We can then at some point,

probably there will be models that you'll

be able to just run locally.

Security will be a big risk.

What happens if you're offline or you've

got very costly exercise that you kind of

need to do for certain things?

You can't be expected if you're on the

tube to have an experience.

So I imagine that they're really going in

on the experience side of, like,

if we can invest in that and then

run a model and we can make that

work and it's really small, then we can,

like,

go head and shoulders above everybody

else.

Yeah.

I'm ready for it.

I'm ready for it.

Are you going to be like me and

have robots and escalators and slides?

I don't even... My aim is to be...

roughly maybe fifty tons in size

eventually where ai is delivering my food

i've got an escalator i slide back down

my house i don't leave let's see i

just never just john you're trying to be

efficient with your life that's it it's

all about you you're not moving but you're

being efficient you're getting everything

exactly i mean i love to have

run some of these models on my phone

when I don't have internet,

which is anytime you're getting on a train

within UK, you have no internet.

So I'll have to run these models and

help myself with some stuff.

So yeah, for sure.

Good luck to the new CEO.

Yeah, exactly.

Good luck to the new CEO.

I want to know more about, I mean,

I'll talk about my story,

but it's a little bit,

but I want to kind of know more

about your random... Random story.

Story, yeah.

I mean,

you're going to have to break it before

I get into mine.

Yeah.

You know,

we all established that you love

efficiency.

And we know that you love a sharp

fade on your hairstyle.

You love that.

Before that,

you have to go to a barber.

And sometimes it's just walk in.

You can't even book an appointment to some

of these barber shops.

You're sitting there and you're waiting

for your turn.

And there's like five people in front of

your queue.

And it's kind of hard.

And then one of the guys,

one of the barbers is empty.

But you know, he doesn't cut that well.

So you don't really want to go to

him.

So that's a bit awkward.

And you're just waiting for the best

barber in the shop and nothing's going on.

And you're just wasting your time.

Internet is terrible.

So you can't even use any of the

models that you want to use to brainstorm.

so glide the company called glide

g-l-y-d-e they said you know what we'll

get you some clippers which are ai based

and they will do a better job at

cutting your hair than a normal person

would right so these clippers these are

actual clippers which have an attachment

so there's like a bigger clipper

attachment and the the clippers go on

And they're AI powered.

And how do they work is you put

a band on and the idea is you

give this clipper to somebody who's never

cut hair.

So you don't have to go to the

barber, right?

And then you put this clipper on and

it's got this band that you put on.

The band just tells you select what

hairstyle you want.

Yeah,

I want to fade because that's what the

clippers do, right?

They're going to fade the hair in.

And then there's a screen on the we'll

put the link in the show notes.

There's a screen on the Clipper that tells

you, okay,

now you need to move it here,

move it there.

Just rough movements like up, down,

left and right.

What you have to move or that you

Clippers on remotely.

A person still has to cut your hair,

but they don't have to be a barber.

I can cut your hair, John.

We can try this.

Oh, I see.

I see.

So basically,

it almost guides them and explains to them

how to use the clippers as AI.

And the blades, the way the blades work,

they move themselves.

So right now,

if you have to do a fade,

you got to change the setting, go in,

move it like that.

The blades are a little bit bigger.

And then the way they move,

they just move themselves.

You just put it in the vicinity.

I want to put it here.

they will do the cut say okay now

move it here we'll do the fade up

down whatever it would change the settings

would you let me do that then if

it would if i get these clippers would

you trust me to use gliding and give

you a good fade how about how about

this i'll get the clippers i'll test it

on you and then you can do this

kind of wasn't the question just that you

just that was kind of the question i

just was just saying to you is like

Would you trust these Clippers?

Are you going to trust me to do

your... Would you trust me?

I would do it.

Would you?

We'd do it live with Cloud Unplugged.

How about that?

We'd do it live with Cloud Unplugged.

This is where we both come back with

the most horrific hairstyles on the next

episode that anyone's ever seen.

If the hairstyle is terrible,

we can go bald for a couple of

weeks.

That's all right.

So basically then it...

Just so I understand then,

there's a band around, a band that's,

what, a physical band?

Yeah,

it's like a thin band that goes around.

That's just your training line.

Oh,

don't go above this line or that line.

Yeah, okay.

So basically, it's just a guideline.

A bit like when you're bowling,

it's like the kid, like you put... Oh,

yeah, that's what it is.

That is exactly what it is.

And then some people did a haircut.

Of course, you know,

it was in CES and...

vegas they got they did some haircut of

course it's not as good as a professional

doing your your your fade but you can

get the short straw if you get go

to the worst barber in the barber shop

and get a really terrible fade the

reasonable job of cutting hair the fade

wasn't amazing but i did a reasonable job

of cutting hair technology is coming there

that means john you can be way more

efficient

You don't have to go.

Everyone that was like, oh, you know,

AI is just coming for like the tech

jobs and like all this.

And then they're like, no,

it's coming for the barbers.

Nobody knew that.

The whole time it was being sidetracked

with like developers and all this time.

It's like this, you know, this.

movement going on around the side that's

about to take out hairdressers and

barbers.

Yeah, scissors will be next.

You know, they'll be like...

Hair extensions.

Where does it go from here?

We've all got amazing hair extensions.

We've all got good perms,

tight perms coming into work.

Yeah.

What's your new story?

We can end on that.

So, yeah, so there was Andon Labs,

who is like an AI...

kind of research company,

I think just mostly around agentics.

And they have opened up a store called

Andon Labs.

Well,

it's opened by Andon Labs called Andon

Market, and it's in San Francisco.

And basically the whole store has been run

and managed and operated by AI called

Luna.

Luna's done all the hiring.

um it decides what to sell done the

logo of the shop design the the shop

logo done all of these things and so

it's an experiment um that is real and

there is a store i think you can

visit it it's on a specific street in

san francisco you can kind of go there

don't know fully what's been sold i think

it's novels and books sci-fi things and

But anyway,

they're trusting it to basically as a test

to see how it operates a business and

whether it can kind of like build a

company on its own without any human

intervention.

Very unlikely, I'd say.

I think the answer is, I mean,

just based on the Clippers story that we

just heard, I think we're all safe.

Until Mythos is out and then, yeah,

everyone's going to have the best haircuts

we've ever seen.

Then it's game-changing.

yeah yeah but yeah that was that story

but anyway so that's that's uh it for

us i guess we'll be doing these um

every week where we can um and kind

of stick into the format

But yeah, it's been good fun.

Good having you on now for this season.

So it's going to be good.

First one down.

Thank you so much for having me and

asking me to do this, John.

I really appreciate that.

And thanks for all your insights.

And we know you love for efficiency.

So next time I'll bring something more

onto the table that will make you even

more efficient, John.

Oh, wow.

That's it.

Yeah.

Okay, cool.

I'm just not on the podcast anymore.

Is that efficient?

It's just someone else.

I'm going to get off this call now

and just going to tell Andon Labs.

Find a new host.

Yeah.

I want them to be.

Yeah, exactly.

Just look at John.

Efficiency.

Efficiency.

Efficiency.

Yeah.

All right.

Great speaking to you.

Bye-bye.

Bye.