Meta’s new AI picks outfits from clothes you own

Welcome to The Signal from Superpower Dailing with Maya and Theo We

are looking at a quiet launch from Meta today They

are testing an AI outfit planner that actually turns your existing closet into a fully automated travel itinerary Right

and

our

mission for this deep dive is clear We want to cut through the noise of today's AI developments We are

here to extract the actual signal for founders builders and

operators

Exactly So

let

us

just jump right into this Meta story Social media consultant

Matt Navarro spotted something

It was kind of

flying under the radar on October 28th 2026 Meta launched this mobile

AI styling app Yeah it's called Style

Sketch Right and this caught

my

attention immediately It completely flips the

standard AI script It is not generating

an imaginary space suit for a dog on Mars No not at all It

is deeply grounded in

physical utility It uses the clothes you

actually already own to suggest outfits It builds entire wardrobes This

is a fascinating pivot

for consumer AI I mean for years the

industry has focused entirely

on generative novelty Right the novelty

phase Exactly You

give a prompt and

the system creates something brand new But Style Sketch requires a completely

different foundation Because it needs real data Right

it requires a database of

your

actual physical assets So it connects directly to your

existing digital footprint Specifically it can scan the photos

you

have already posted to Instagram Wow Yeah and it pulls individual

garments

out of those images Then it uses those specific pieces

to construct a virtual closet for you Okay I have to pause right

there

I mean let's unpack this Do

my carefully curated or frankly

completely chaotic Instagram photos

actually reflect the reality

of what is in my closet That is the big question Right

like I might rent a

tuxedo

for a wedding once I take a picture it goes on the grid That

does not mean I own

a tuxedo No it definitely does not And

it

certainly does not mean

it is part of my daily wardrobe rotation So this feels like a massive

leap Going from a social media highlight

reel to functional daily wardrobe advice just seems deeply flawed You are identifying the exact limitation

of relying solely on social media data A feed is just an idealized

projection It is absolutely not an inventory

And Meta clearly recognizes this massive

data

gap They actually built multiple ingestion methods into

the application to solve it Okay so you don't have to just

rely

on the Instagram feed You can actively digitize your wardrobe yourself

So

how does that actually work in practice

Am I just standing there taking photos of all my black t shirts Well essentially yes Okay You can manually

upload images

directly

from

your phone's camera roll Or you can take fresh pictures of

your garments like just laid out on your bed Oh wow That sounds tedious It is

But you

can also record

short videos

You

just pan the camera across your closet

The AI then extracts the individual items from

those videos Wait really Just from a panning video Yeah

You

are

actively training their system on

your physical inventory So you are basically doing the manual labor

of data entry to make their AI useful You are building a personal database for them

Precisely And once that database is populated well

the

orchestration engine takes over Okay so

what

does the orchestration actually do

You

can ask StyleSketch to build combinations around

a single specific item Say you have this vintage leather jacket You

select that jacket in the app Right

The artificial intelligence then builds multiple complete

looks

around that one piece It combines it with the shirts the pants

and the shoes you actually own

Okay that is pretty cool But this brings

us to a wildly ambitious

technical claim they made The application features virtual try ons Yes it

does And Meta states in

their description that

the digital model is matched to your specific body shape They also claim it matches your skin

tone So you are not just seeing a

flat lay of clothes on a screen You are seeing a digital approximation of

yourself

wearing those specific outfits We really need to examine

the mechanics of that

claim carefully Converting two dimensional photographs of clothing into accurate three

dimensional models is an immense technical challenge Because of how clothes fit in real

life Fabric physics are incredibly complex I mean

I

have you wool sweater drapes

completely differently than a silk shirt Yeah that makes sense And lighting changes

the color and the

texture of a garment entirely Right And if the AI gets the

drape wrong or the fit wrong the virtual try on is essentially useless

Exactly Now Meta claims the system

works

seamlessly

However the appearance and fit claims

of

this virtual try on have not been independently verified So no outside testers have

confirmed this yet Right The leap from a

flat

photo to an accurate personalized 3D rendering is just

massive We have to treat that

specific feature with healthy skepticism right now We

need to wait until broader

testing occurs Makes sense Let us move to

the feature

that truly separates this from a standard styling app because the travel integration is fascinating

It really is You input a

destination into the

application say Singapore You enter the exact

length of

your trip You add your planned activities for

each day Right The application then generates a

comprehensive

day by day clothing plan And this is where

we see the shift to complex orchestration The artificial intelligence is

longer just answering a single isolated prompt It is juggling a lot

of things Yes It is synthesizing multiple

independent data streams in real time So break down those data

streams for us What is the system actually

juggling behind the scenes First it must understand the

constraints

of your personal inventory database It can only suggest what

you own Right Second it must parse

your

schedule

and activities A business meeting obviously requires

a completely different outfit than a hiking trip Of course Third and this

is the

most

important part it pulls the real time

local weather forecast for your specific destination That is so smart

Right It maps the

forecast

against your schedule And finally it filters

all of that through the items available in your digitized closet It is

entirely bridging the gap

between your digital presence and physical reality I mean you are essentially outsourcing

the cognitive load of packing Exactly I despise packing for trips

So

this actually solves a tangible

human problem If you were building a consumer

product this is the gold standard Right

It makes artificial intelligence an

invisible utility It is a very powerful

concept but we really must address

the current reality of the rollout Availability is highly restricted at

this moment Right Who actually has this

Well

StyleSketch is available on

both

iOS and Android platforms However it is only operating

in

three countries right now It is restricted entirely to Indonesia Singapore and Thailand Why limit

a

global product

to those three specific countries Are they just you know

testing the waters and markets that index heavily on mobile usage That is the

standard playbook for these types of social utility features Those populations are

highly engaged They are predominantly mobile first So they're good

test subjects Exactly They serve

as excellent contained test beds for new server heavy applications Meta is keeping

their broader strategy very quiet though Really No announcements

No

TechTrench specifically

requested

a comment regarding a wider global rollout schedule Meta just did

not respond to the inquiry Interesting Well if

you

are

an operator or a founder watching this space the implications are

staggering We definitely need

to watch whether Meta expands this beyond those test markets Absolutely But the

real

signal here is the potential for direct e commerce integration I mean

think about it If Meta

knows exactly what is inside your closet they inherently

know exactly what is missing The targeted

advertising potential there is unprecedented Think about

the travel feature we just discussed The AI

knows you are packing for a five day ski trip to Colorado It

scans your inventory It

realizes you do not own a winter coat Oh

wow So

the application can just seamlessly drop a sponsored link for a winter coat directly into your packing list Precisely

The path from identifying a user's need to facilitating a direct sale becomes incredibly

short

It just transforms a

helpful utility app into a highly personalized retail engine Exactly It changes the

entire

paradigm of how social platforms monetize user data That wraps up our

look at Meta's quiet foray into fashion utility Meta's Outfit Planner is

a perfect example of just how hungry artificial intelligence is for our personal data It really is But

what happens when

the platforms holding that data decide they do not

want to share it anymore That brings us to

our next story Reddit is preparing to

make an account the default way to access its site This shift comes as

traditional search traffic falls and AI scraping looms over the entire platform Right And this is

a

fundamental change to the architecture of Reddit The announcement actually came

alongside their third quarter earnings report So what exactly are

they doing Reddit is currently testing

a primarily logged in experience These tests are

beginning in the fourth quarter and they are

rolling this out in select unnamed markets outside of the United States We really have

to look at the data forcing their hand here because the reality

is that the traditional search pipeline is collapsing It is falling

apart Yeah Search referrals

to Reddit fell by over 1 million per week on average throughout October And 1 million

fewer referrals a week is not some statistical

anomaly It is a structural shift

in user behavior We also see this reflected

in

their broader engagement metrics Reddit reported 130 3 million daily

active users worldwide in the third quarter That number is completely flat

compared to the previous quarter There's zero growth The numbers

in the United States

are actually contracting too Daily active users in the

U

S drop from 53 2 million down to 50 7

million That is amazing massive drop It

is If you

rely on

search referrals

for your business this is a massive warning sign

It is a direct result of

generative AI summaries changing how consumers interact with information I mean for a decade the

transaction was very simple Right

you search for something You

had

a hyper specific question

you typed it into a

search engine you clicked a link to

a Reddit thread you read the human discussion

and Reddit served you an

advertisement But now an artificial intelligence

just reads that Reddit discussion in the background

for you Exactly It synthesizes the best answers

from the community It serves that information directly to

you on the main search results page So you get the answer immediately

You never click the link Reddit

never sees your visit They

never show you the ad

The AI platforms are basically extracting

the raw value from Reddit's communities They are bypassing the monetization engine

entirely That makes total sense So this

logged in shift is

really a dual purpose defensive maneuver by Reddit First they are dealing

directly with the scraping

issue

Account gating gives Reddit immediate leverage Right because

it stops unauthorized AI bots from

freely

hoovering up their historical archives Exactly If a massive language model provider wants

the data to train their systems they cannot just scrape it anonymously anymore They

have to pay for it Yes They

have to come to the table

They

have to sign a formal licensing deal

with Reddit You know here is the deep irony of

this whole situation though Steve Huffman the CEO of Reddit he

previously

stated that

his entire life was

defined by the concept of the open internet Oh I remember that Now the champion of

the open internet is

literally pulling up the drawbridge Reddit is basically building a moat They are

sacrificing casual drive by search traffic

to protect their raw material from AI bots It is a profound philosophical shift for

the platform but we must also look at the

second equally important business purpose here Logged in users provide far richer behavioral

signals to the platform Explain how that changes the advertising

math for them Well when you browse Reddit while

logged

out you are essentially a ghost to

their advertising infrastructure They know very little about you

Right You are just a page

view

Exactly But when you are

logged in the entire equation changes Reddit tracks the specific communities

you visit They monitor your daily interests

They analyze your specific voting

patterns So they can target advertisements

with exponentially higher precision Precisely So even if the

total

raw number of users shrinks because of the account

wall the financial value of each remaining

user actually goes up It is a calculated play for

quality over quantity It is a necessary pivot for

a public company trying to satisfy shareholders right now But we really must highlight the

immense caveats surrounding this rollout Because the actual boundaries of this

new walled garden are entirely unsettled right now Very unsettled Reddit claims the logged

out front

page will remain accessible They are not locking the front door completely They also state that forcing a

sign in will not require giving up user anonymity And you can still operate

a burner account with a fake email Furthermore they are keeping their

incognito

mode They are simply rebuilding its underlying architecture But the most important unanswered

question is purely technical Right Will this actually work

It is completely unproven whether

account walls will actually stop determined AI scrapers Because scraping technology adapts incredibly

quickly to new access barriers I mean if there

is a way to automate account creation the bots will find

it That is the arms race

we are watching right now We need to monitor how Reddit defines the exact boundary of what a logged out user can actually see And we will watch how

these fourth quarter tests perform internationally We have to see that

for any massive global rollout occurs Absolutely Meanwhile speaking of artificial

intelligence

attempting

to bypass secure walls let us look at a

major cybersecurity development CrowdStrike has traced a massive

AI driven espionage campaign This is a huge

story

It really is This operation

targeted more than

200 organizations worldwide This is a critical moment

for cybersecurity We are watching AI move

from theoretical threat models into active deployed cyber operations And this

was a China

linked operation CrowdStrike attributes the campaign to an entity they call Vapor Panda The campaign ran for

11 days It began on October

17 And a single controller

used a new AI framework called Artex They targeted government ministries

semiconductor firms and telecommunications

companies And this was across

14 different countries The Artex framework is

really what makes this operation so sophisticated It relied on two

distinct language models Which ones It utilized

DeepSeek and

it also used an unidentified commercial model But the models themselves

are less important than the architecture built around them So let us unpack

how this framework actually operates inside a network CrowdStrike described an agent

based architecture What does that

mean in practical terms It means the

system operates through a structured hierarchy of AI agents There

is a main

agent that acts as the central coordinator This main agent directs

subordinate

planning agents and worker agents They all share a connected

real time record of facts findings and goals So

they

are constantly

communicating with each other Exactly They attempt to

take

an engagement from initial access all the way to moving deeper into a secure network Wait so if I

am

understanding this architecture correctly this is not some

autonomous AI mastermind going rogue right No not at

all

It is much more

like a highly caffeinated intern rattling every single doorknob in the building at lightning speed and a human boss

is standing back holding the floor plan That is an

incredibly precise way to visualize a threat We actually have

the recovered command transcripts to prove the human was holding that map Oh wow Yeah

this was

absolutely

not an unattended autonomous system There was a

human operator actively issuing instructions in Chinese Right and CrowdStrike

noted the timestamps of those commands aligned perfectly with China Standard Time The humans set

the strategic goals The artificial intelligence

executed the tedious parallel tasks And the speed

of that execution is the real danger here The agents explored

multiple different attack paths at the exact same time Just branching out

instantly Exactly If

one

specific path failed the agents

communicated the failure and pivoted immediately to a new strategy CrowdStrike actually

provided a terrifying example of this at a United States pharmaceutical company Can you

explain the specific mechanism of how the AI breached that system The AI

agent found two separate seemingly minor flaws It found

a basic authentication flaw and it

found an XML processing flaw Okay when

we say it combined an authentication flaw with an XML processing flaw think of

it like a coordinated heist The authentication

flaw is like finding a broken lock on the lobby door Right Gets

you

inside the building but you

are still just in the lobby Exactly The

XML

flaw is the second phase of the heist XML is a markup language used to store and transport data Right The

flaw is like handing the internal security guard a forged VIP pass that the system simply does not know how to read properly Oh

I see Because

the internal system gets confused by the formatting it defaults to a catastrophic error It

actually grants access instead of denying it Wow

It

hands the AI a cloud role It

essentially gives the system a master key to the rest of the building Exactly

Then the AI escalated that access to obtain a full enterprise administrator account It

strung these vulnerabilities together seamlessly That

is terrifying What

is truly fascinating here is the evasion tactic utilized during the intrusion The

human operators specifically directed the AI agents to look for computers that did not have CrowdStrike's Falcon security sensors installed So

the humans supplied the evasion strategy Yeah

The

agents simply scanned

the network to find the unprotected targets Yes

The

scale of the processing required to do this is wild It

is massive I

mean in just one intrusion event the

planning agents consumed 1 4 billion input tokens

We really must be very precise when discussing that specific metric though Consuming

1

4 billion input tokens measures processing volume Right

It

measures how much data the model ingested Exactly

Okay

It

does not measure the success rate of the operation It

simply means the models were reading and analyzing a massive amount of internal text like network maps and system configurations during the attack That

brings us to a massive caveat in this entire story There is

a huge gap between being targeted by this system and actually

being breached A

very big gap Over

200 organizations were targeted by VaporPanda However only 12 faced actual credential attacks Right And

only nine of those organizations had confirmed post intrusion activity

Just nine Yes

At

those nine confirmed victims the

attackers successfully deployed a persistent backdoor program called Blood Alchemy But

the vast majority of the 200 targets were merely probed The

AI rattled the doorknobs at 200 buildings but

almost all of the doors were securely locked If

you are a network operator this

emphasizes the importance of basic security hygiene It

absolutely does We

need to

watch whether further investigations by CrowdStrike confirm deeper access at any of those other targeted organizations over the coming weeks Definitely something to

keep an eye on In other news

we just examined

artificial intelligence attempting digital break ins Now let

us look at the intersection of AI and physical security This is

a major shift OpenAI backed

Thrive Holdings has officially agreed to buy the legacy security giant ADT This is a massive 15 3 billion proposed acquisition

ADT is going Thrive

is paying

10 a share and that represents

a 45 6 premium over ADT's closing price right before the announcement Apollo affiliates

are actually providing the necessary debt financing And the

grand strategy here is that ADT will serve as the physical anchor for Thrive's new smart home security business This is

a classic combination of vastly different corporate assets I mean

ADT brings immense physical scale and operational history They

have been

around forever Exactly They have

a massive nationwide installation network They manage 12 000 human employees They possess

150 years of experience in the physical security business Right but

Thrive

brings something entirely different to the table They bring

permanent capital They also

bring a deep structural partnership with OpenAI Right Explain that

specific

partnership How exactly does an artificial intelligence

research lab fit into a private equity holding company Well OpenAI operates

as both an investor in Thrive and a dedicated technology partner This specific partnership

arrangement supplies Thrive's portfolio companies with early access to AI models Okay early access

makes

sense

But more importantly it

provides forward

deployed engineers directly from OpenAI Oh wow

So these

elite AI engineers actually embed within the

legacy business teams at ADT They work

alongside the traditional operations staff to implement new technology That is

the structural thesis yes It

makes

a lot of sense on paper You combine

patient private capital with top tier AI engineering talent And

the goal is to rapidly modernize a legacy physical security company away from the glare of public markets Exactly

They

want to offer drastically better customer support They

want to give first responders highly actionable data instead of just a generic binary alarm bell going off in a dispatch center

That is the

overarching corporate narrative Right I have

to

push back here though They just spent

over 15 billion There is a

massive amount of AI buzz surrounding this deal simply because of the OpenAI connection Of course But

they did

not actually announce a single new AI product today Is this a

genuine technological revolution or is this just private equity wearing an AI trench coat That is the

15 billion question really The AI

integration

currently exists purely in theory Karim Zaki is

a partner at Thrive He talked in

the press release about strengthening technology capabilities over decades Yes That is

a

very long very patient time horizon As of this exact

moment zero new AI security

features actually exist for ADT customers But what about the

existing customer data That is the real

asset ADT has cameras positioned

inside millions of private homes They know exactly when

people leave for work They know when the

house is empty That is incredibly sensitive

behavioral data It is It is

also incredibly

valuable training data for an AI company The companies knew this

would be a major concern They addressed the data

issue directly in the announcement ADT's stringent privacy policies

remain entirely unchanged by this acquisition So they claim Right