Two weeks ago,
Meta just launched:
Muse.
It’s not just a chatbot.
It can:
Send emails.
Plan trips.
Shop.
Fill out forms.
Operate websites.
And even continue completing tasks
after you close the app:
A personal AI agent.
Now, only two weeks later,
Meta is already addressing the next question:
If this agent is really meant to stay with you all day, why do you always have to take out your phone first?
The answer is:
You don’t.
Meta is bringing Muse into AI glasses
At:
Connect 2026,
Meta announced
that Muse will be integrated into
Meta AI Glasses.
This doesn’t mean:
glasses with just another chatbot inside.
Meta’s intention is that:
you don’t need to pull out your phone,
but can directly interact through:
voice.
the glasses themselves.
your surroundings,
and your own agent,
continuously.
For example:
You’re walking along,
a thought comes to mind,
and you simply tell Muse.
You see a product,
you ask instantly.
You’re working on your schedule,
and instruct the agent to
take over the task.
AI begins to shift from being
an app you actively open,
to
a service always by your side.
Meta also created a smaller entry point: Muse Charm
In addition, Meta unveiled:
Muse Charm.
A device:
small enough to fit in your pocket,
about the size of a keychain.
The purpose is simple:
to let you speak to Muse
via real-time voice interaction
without needing to pull out a phone or wear glasses.
Meta has not yet released
full specs,
pricing,
or a launch date.
The official word is
more details will come later this year.
But this direction is noteworthy.
Because it doesn’t even require you to wear glasses
Smart glasses face a big challenge:
not everyone wants to
wear them all day,
nor do all people want to
have a camera constantly pointing from their face.
Muse Charm takes a different approach.
It doesn’t require:
a screen,
bulky headgear,
or even necessarily a camera.
It’s simply:
a portable, voice-first interface
to converse with your agent.
This shows Meta is experimenting with
which devices AI is best integrated into.
The answer may no longer be just
your phone.
The phone is not going away
To be clear,
Meta hasn’t announced the end of the
smartphone era.
You will still need your phone to:
set up accounts,
view content,
confirm sensitive actions,
and handle complex interfaces.
The question is not whether phones will disappear tomorrow,
but whether some tasks
can happen without always taking out your phone first.
For example, if you suddenly think while walking:
“Find me a one-hour slot to review my report tomorrow afternoon.”
Previously, you would:
take out your phone,
unlock it,
open your calendar,
search for time,
and add an event.
Now Meta imagines you can simply say to Muse:
“Find me an open one-hour slot tomorrow afternoon.”
The agent:
knows your calendar,
finds the time,
and depending on permissions,
creates the event,
or asks you to confirm first.
This is:
an ambient agent.
The real change is not just voice assistants returning
We have had before:
Siri,
Google Assistant,
and Alexa,
all able to answer spoken questions.
But Muse is different:
Meta positions it as an
agent.
Not just:
“What time is it?”
“Play music.”
“What’s the weather today?”
Rather,
after receiving a goal,
it handles multiple steps,
using different services,
continuing to act,
sometimes operating in the background.
This is a different level of autonomy
compared to traditional voice assistants.
Meta is also connecting Muse to more real-world services
Another key announcement at Connect 2026 was
continued expansion of Muse’s connectors.
Shopping services now include:
Walmart,
Best Buy,
American Eagle Outfitters,
DICK'S Sporting Goods,
Fanatics,
Gap,
Michael Kors,
Sephora,
Ulta,
Wayfair.
Payment services now include:
Shop Pay,
PayPal.
Travel will soon add:
Expedia.
Groceries:
Instacart.
Work tools include:
Notion,
Granola,
GitHub,
Box.
These connectors are crucial
to Muse truly acting as an agent.
Without connectors, AI can only advise, not act
For example:
“Help me find a jacket.”
A chatbot can:
recommend.
But an agent connected to
shopping services,
payment,
and accounts
can
search,
compare,
add to cart,
and even prepare payments.
The difference is
moving from advice to action.
More interestingly: Muse will have its own email address
Meta also announced that
Muse will have
its own email address.
This may sound small,
but is very significant.
Email used to be
a gateway between people;
now Meta wants
an agent
to have its own
email address to send and receive task information.
You can
communicate with Muse via email.
Other services might also send
data to Muse.
The agent can then
organize,
process,
and track your information for you.
This gives agents
a more independent
digital identity.
This is different from "ChatGPT/Gemini checking my emails"
Those solutions typically have
AI accessing
your email inbox.
Muse having its own email
works the other way around.
The agent itself
becomes a participant in your workflows.
This could enable many new workflows in the future,
such as:
vendors sending
order status
directly to your agent,
airlines sending
itinerary changes,
work tools sending
task updates,
and the agent sorting these
and only alerting you when necessary.
This is the future Meta is truly betting on
Not:
people asking AI
100 questions a day,
but AI that exists alongside you for long periods,
continually receiving:
context,
tasks,
and changes,
taking action when needed.
This is why
glasses,
Charm,
email,
and connectors
must be viewed together.
They are not
four separate features,
but an entire
infrastructure for an agent that works with you all day.
Meta’s new Ray-Ban Meta Audio glasses follow the "always available" path
Meta also announced:
Ray-Ban Meta Audio,
which is
AI glasses that don’t emphasize a camera,
but focus on
open-ear audio
and AI integration.
Weighing only:
43 grams,
Meta says:
single-charge battery life up to
12 hours,
with the charging case providing
an additional 48 hours.
Price starts at:
$349 USD,
shipping begins October 13.
Design matters because the biggest barrier to smart glasses has never been AI strength
But:
whether people are willing to wear them.
If glasses are:
heavy,
ugly,
have poor battery life,
look too tech-heavy,
or feature cameras that make others uncomfortable,
no matter how powerful the agent is,
people won’t really wear them all day.
So Meta now emphasizes:
lighter weight,
more styles,
longer battery,
open-ear audio,
and even camera-free designs.
The tech competition is shifting from:
"What can AI do?"
to:
"Will people actually wear this for 8 hours?"
Meta aims to offer over 100 AI glasses options by year-end
This was another clear message at Connect 2026.
Meta said by the end of this year,
Ray-Ban,
Oakley,
and Meta Glasses
will offer more than
100 different styles and options.
This shows the company doesn’t want
to just make one geeky gadget.
It aims to make AI glasses
a normal product category
within the regular eyewear market.
Just like you don’t buy a "smartphone" generically,
but an iPhone, Galaxy, or Pixel
with different sizes, prices, and features,
Meta wants AI glasses
to become a daily consumer item.
But wearing an agent all day raises tougher safety challenges
Chatbots start talking
only when you open an app.
Wearable agents could
constantly be nearby,
knowing:
where you are,
what you’re doing,
where you’re going,
who you’re talking to,
your calendar,
email,
shopping,
work,
even payment info.
The more convenient the agent,
the more context it requires.
This is the hardest tradeoff
between
convenience
and
privacy.
Reuters recently explored this issue
Before Meta Connect,
Reuters interviewed tech executives,
legal experts, and industry insiders,
focusing on:
AI wearables
facing a
privacy backlash,
especially smart glasses.
The reason is straightforward.
These devices sit on
your face,
but their cameras and microphones
not only collect your data,
but may also capture data of
people next to you who don’t use the device.
Others may not even know they’re being part of the context
You wear glasses,
walk into your office,
and there’s
unreleased product info on a whiteboard,
client data on a nearby screen,
and your colleagues chatting.
If the agent can see,
hear,
and remember,
the issue isn’t just:
whether you agree,
but
whether others agree.
This is why wearable AI
is much harder than mobile apps.
Meta is trying hardware solutions to reduce some problems
For instance,
Meta Glasses with cameras
have an indicator light that turns on during recording.
Meta says:
if the light is covered or disabled,
the camera is turned off.
This design aims to
make others aware when recording is happening.
But the problem isn’t fully solved,
because
"seeing a light"
doesn’t guarantee people understand
what data is collected, for how long,
who sees it,
or how AI uses it.
Reuters notes privacy controversies already impact real use cases
Some places
have begun restricting
camera-equipped smart glasses.
This is why
Meta’s new
Ray-Ban Meta Audio glasses,
which focus more on sound and less on camera,
are worth watching.
This may not be
a downgrade in specs,
but a
product strategy:
collect less data, gain less social resistance.
Muse Charm may follow the same logic
If a user wants to:
check in with their agent anytime,
assign tasks,
or make notes,
they may not want or need a camera every time.
A voice-first pocket device
may be more socially acceptable
than always-on camera glasses.
So Meta isn’t betting on
just one device type,
but trying:
glasses,
audio glasses,
display glasses,
VR glasses,
Charm,
phone,
and desktop,
to see
which will become the true entrance
to the Agent Era.
But more entry points mean more complex authorization
Assuming Muse can:
hear your request via glasses,
check your calendar,
connect to PayPal,
search Walmart for items,
and confirm via its own email,
this workflow spans
five or six services.
What you need to know is:
which steps are read-only?
which steps write data?
which steps trigger payments?
which steps make external commitments?
This goes beyond simple AI privacy settings
In the future, everyone may need
their own
agent permission model.
For example:
Calendar:
can be read.
Adding events:
requires confirmation.
Email:
can draft messages,
sending requires approval.
Shopping:
can compare,
add to cart allowed,
payments require confirmation.
Work documents:
only certain folders can be accessed.
This is far more complex than today's app permissions like:
“Allow camera?”
“Allow notifications?”
Even “human approval” may become too frequent
If the agent asks
“Is this okay?”
for every small action,
users may experience approval fatigue,
ending up clicking “Yes” automatically,
which defeats the purpose.
So a good future agent
won’t just ask for approval on important matters, but will allow setting:
which actions are always allowed,
one-time permissions,
task-specific permissions,
time-limited permissions,
always ask if above a spending limit,
always ask before sending externally.
These
granular permission controls
will determine whether an agent can safely enter daily life.
Meta is already moving Muse in this direction
When Muse launched,
Meta emphasized:
different tasks have
different permission scopes.
Users decide how much
access an agent gets.
Sensitive actions
require approval.
Muse runs on
a secure VM that
isolates the agent’s environment
from other sessions.
So at Connect 2026,
the real challenge isn’t whether Meta thought about security,
but whether
the permission system
can stay simple
as Muse expands beyond the app
to glasses,
Charm,
email,
shopping,
payment,
and work connectors.
Otherwise, the more “frictionless” it is, the easier we forget AI is acting for us
This might be
a very interesting paradox of the Agent Era.
The best user experience is
you don’t have to think about it,
say a command,
and the task is done.
But security requires
you know what’s happening
at key moments,
understand
what AI plans to do,
and have a chance
to stop it.
So the toughest AI product design in the future
may be not model capability,
but
knowing when to let users forget AI and when to bring them back in.
On glasses, this issue grows larger
Phones at least have
screens.
If your agent plans to
send an email,
buy a product,
or change a meeting,
a phone can give you
a complete preview.
But glasses? Charm?
Voice-only?
How do you quickly know
what’s being prepared?
How do you approve?
How do you cancel?
How do you undo?
These user experience questions
may be as important as the AI models themselves.
That’s why today’s news is not just about “Meta launching new hardware”
Looking only at hardware,
you see:
new glasses,
new Charm device,
more styles.
But when you combine:
Muse,
Connectors,
email address,
glasses,
and Charm,
Meta’s strategy is clear:
building an environment where a personal AI agent can always be called upon.
The phone is just
one of the entry points.
The next major AI battle may not be “who has the most app downloads”
but
who is closest to the user.
Google has
Android,
Chrome,
Workspace,
and Gemini.
Apple has
iPhone,
Watch,
AirPods,
and Mac.
Meta has
Facebook,
Instagram,
WhatsApp,
Ray-Ban,
and Muse.
OpenAI
is making ChatGPT
an entry for
work,
search,
shopping,
advertising,
and agents.
As model capabilities converge,
distribution
will become more important.
Wearables may become the next layer of distribution
Phones require
you to take them out.
Glasses are
already worn.
Charms are
already in your pocket.
Earbuds are
already in your ears.
If AI can
always be on your closest device,
usage frequency will be very different.
Meta is clearly
betting early on this future.
But Taiwanese users should not expect "Muse glasses to be widely available in Taiwan" yet
Meta’s announcement focuses on:
products and global expansion plans.
Different glasses styles,
Muse features,
and markets
will roll out separately.
The official new markets listed include:
Singapore,
South Korea,
Mexico,
and the UAE.
Others planned are:
Hong Kong,
Indonesia,
Malaysia,
Philippines,
and Thailand.
Taiwan is currently not on this official new AI glasses market list.
So today it’s worth focusing on:
the direction,
not on
“going to buy Muse glasses at a Taiwan store tomorrow.”
Finally, Meta’s real bet isn’t on glasses alone
It’s on a bigger idea:
AI agents shouldn’t wait until you open them.
They should:
stay with you,
understand your context,
be ready when needed,
complete tasks in the background,
and hand decisions back to you
at crucial points.
If this approach succeeds,
we may no longer say:
“I’m opening AI.”
just like today few people say:
“I’m opening the Internet.”
AI will become
a persistent layer behind devices.
But the most important question
will no longer be:
“How much can AI do for me?”
but:
“For an agent accompanying me all day, how much am I willing to let it know, decide, and on which matters must it always ask me first?”
Today, let’s improve a little with AI.
Learn an AI skill daily.
Save a little time every day.
Enhance your capabilities bit by bit.
SasaDaily, growing with you.
Recommended reading
Why Meta now trusts human contractors to handle AI calls instead of fully automating them
