the localhost:0011 | microsoft ignite
Download MP3I am super excited to be able to
have the local host here in San
Francisco at Microsoft Ignite.
We can go down the line,
and we'll start with Parker.
Yeah.
Parker Johnstone, Agentic Field
CTO at Insight, and I get that
awesome opportunity to go and bleed
together what does AI mean for our
organization, either helping customers
take advantage of capability or
start to look through how do we maybe
start to bleed towards the frontier.
That-- Gotta weave that in
knowing we're at Ignite.
But, on device has been the big
thing for me the last couple months,
so trying to find a way to help and
have folks understand, hey, we can do
it locally, we can do it hybrid, or
we can take advantage of the entire
hyperscaler that we have out there.
Yeah, Will Pocknell.
I'm a director of cybersecurity at
Insight, and, I'm actually part of our
internal cybersecurity team, so I report
directly to our CISO, Jason Radar.
And, I've been at the forefront of this
from an internal security standpoint,
where we have seen and gone through
the transformation of AI in the
cloud, now we're doing it on device.
we were customer zero for Agent 365-
Oh, wow … and a launch partner,
working very closely with Microsoft
on that, and also doing a lot of
testing with Agent 365 for local
inference and local a- agents as well.
So it's, it's, it's fun
being on the f- the frontier.
The frontier.
Yeah.
Jacob, I see you down there.
I think you're actually… You're
probably glad to be sitting,
'cause I think you've been standing
for the last, 72 hours straight.
Maybe 96 hours.
Yeah.
It's been quite a while.
But, hey, I'm Jacob Rhodes, everyone.
happy to be back on.
I do work for Microsoft, but this podcast
is not a Microsoft podcast, so we're here
just to share, and talk, and, share about
what we kn- have learned, and, excited to
pick the brains of Parker and Will today.
Yeah.
Yeah.
And then at the far end, we got
Mr. Chauncey Larson, who helped
organize a lot of what's going on
down on the floor, including Jacob.
How much time are you gonna take
off after you get done here?
At least a week.
Yeah.
Yeah, yeah.
And I did… I don't feel like I
actually helped organize, Jacob.
If anything- … Jacob helped organize me.
But, yeah, same thing, Chauncey Larson.
I also work for Microsoft,
full transparency.
But, super excited to be able
to speak with, Park and Will
h- Parker and Will here today.
and, actually, I just kinda
wanted to do a quick shout-out.
Frank, I didn't know you had
an entire podcast set-up, man.
Right?
Like, this is impressive.
Look at this.
, You own a Sprinter van or a 4Runner.
Hey, we've even got a Surface Hub with us.
Nice.
hey, I don't go anywhere
without my Surface Hub.
I think that's pretty iconic.
So, And then, hey, we've
also got Neil on the line.
Neil, are y- are you…
Where are you, Neil?
I am in Seattle, Washington at the moment.
A member of Surface Engineering.
Frank, your,
there, it's a bit of a s-
a Surface museum on wheels.
I heard Surface headphones
are in play here, the Hub.
full stack.
excited to connect with you all today.
Definitely missing Ignite this year.
but,
kudos to all the presenters, all
the hard work that went into that.
I've been hearing nothing but
great things, despite some
of the facility challenges.
Oh.
… yesterday.
Too soon for that joke.
Glad to see everyone's still dry
and- That hurts persevering through.
I think there was some water slides
that were set up, and there was,
like making the most of it, so.
When Jacob wasn't on his feet,
he was actually in a canoe.
Yeah, yeah.
Yeah, there's video evidence.
I mean, there was pretty epic rain.
I think, we see, we saw like, like manhole
covers being shot out of the ground down
in like the Mission District and different
places 'cause the amount of water that
was coming into San Francisco yesterday.
So yeah, pretty, pretty wild time.
like, you know, leave it up to
Microsoft from Seattle to bring
the rain down into San Francisco.
Maybe the 49ers were playing,
we were trying to drown them
out or something like that.
I don't know.
So… Well, I heard the, DGX workstation
people down there were really like
shielding those things with their bodies.
Yeah.
Supposedly they don't
like them to get wet.
Yeah.
I don't know what that
was all about, but yeah.
We, we were literally threatened with,
PD enforcement to leave the space,
but we were arm in arm around the
DGX stations and- … the Surface RTX
Spark dev boxes, and all this, costly
hardware, but, , just precious, demos
and setup and all that good stuff, so.
Well, I think the first question
I wanna ask everyone is literally
what was the highlight, or what was
the, , and even lowlight, but what
was the highlight of your week so far?
But I know all of you have been spending
so much time on the floor talking to
customers, and maybe just going down
the line, like, , Chauncey, like h-
, starting with you, like what was your
kinda highlight of, your conversations?
And maybe the tone of the customers
and h- what's their tone this
year compared to even last year?
we get to talk about local AI all the
time, and we get to talk about it with
our customers, but it's oftentimes
it's a us to them conversation, where
in this case, it was them coming to us
really wanting to know more about it.
Like, really wanting to understand,
here's what I heard at the
keynote," or, "Here's what I've
heard at, in this breakout session.
Can you tell me more about hybrid
intelligence?" Which I think in a
lot of people's minds is a little
bit of a toggle, which gives us
the opportunity to then open the
conversation to it being a good concept.
so I've, I really have, been excited
to be able to just see customers
starting to, I think, grasp this idea
that like AI can kind of run where
you want it to run and that, there's
a lot of really cool technologies
coming out to kind of support it.
So I think that's, that, that's
my biggest excitement is that
it's starting to click finally.
I'll go next.
I think it's kind of getting a lot of
those models now open weighted, right?
Mm-hmm.
Microsoft is continuing to make
this massive investment in pushing
the concerns of, "Oh, that's a
Chinese model, a Chinese lab.
We don't want it in our data
center." Microsoft is continuing to
be at the forefront of let's open
weight some of these capabilities.
There's a lot of folks out there saying,
"These, these items on our hardware
are secure, performant, able to meet
the needs of our day-to-day workload."
That to me was a big one.
And then you have on stage new coder
models, decision engines, everything
you're gonna need to take that idea
of what's this toggle and where do I
need to route all of this stuff through
to be more secure and performant.
That was my big takeaway.
Yeah.
I mean, I was… got an opportunity
to go down on the floor and, there's
a gentleman that is well-known in the
Surface community, David Alexander.
He's been part of the Surface
team since its beginning.
But David's now, doing some
experiential things down on the floor.
We now call him instead
of DA, his name is DAI.
Small I. So s- small I. So, but
David was actually showing an MAI
model running on an RTX Arc device.
I wasn't actually even aware that
they were available yet, but he was
showing on his, , 128 gig, , unified
memory, this amazing MAI model.
Talk about one stack.
I mean, all those things you just talked
about, Parker, of like being able to
have a solution from a US or, , a non,
, those types of environments where you
can load this model and have it in that
environment right in GitHub Copilot,
, running straight into that environment.
I think that's promising for a lot of
our customers because, , most of these
customers, while I love hacking and
loading different models from GitHub,
most of our customers are not gonna
be wanting to go out to GitHub and
get a, a bunch of different models.
They wanna know, what is it that you
want to be able to put onto this device?
they want that one model
or that one solution.
They don't wanna just tinker
and play like most of us.
, They just wanna know what's gonna
work for them at an enterprise level.
And I wonder from a
security standpoint, they…
Not to take away from your highlight
of your week, but, it seems like,
we're starting to get into a
little bit more standardization.
I still think it's a little Wild
Wild West, but Will, I mean,
what's your take on, what you saw?
Yeah.
I was really excited by the move from
Agent 365 just being in the cloud to being
hybrid and running not just on things
like the RTXs, but also the governance
with, like, Windows 2028 and moving
those foundry capabilities down there.
one thing I didn't realize is, Windows
ML with a build once run anywhere, model.
I was in a session earlier today
and heard about that, and, it was
with Kurt Stephenson and Mark-
Oh, right … on this stuff.
And they made the comment, Windows
ML is like- DirectX, but for AI.
you write it once and it runs anywhere.
You do all the hard stuff.
So I th- I think that's amazing.
the other thing that really jumped out at
me, and this is a bit geeky here, is the,
transparency working with the AI, right?
So when we started all of this
a few years ago, there's thing--
and you had, chain of thought.
You could see all the thinking,
reasoning, and everything.
And now with distillation, all the
big ones are encrypting that, hiding
it away, so we don't know what
it's doing or why it's doing it.
And Microsoft getting out in front of that
and really leading that sort of visibility
and opening up a, a consortium to do that
is huge for me as a security practitioner.
And then along with Agent 365,
getting that visibility and all
the telemetry from not just cloud
models, but on-device models as well.
Right.
Yeah, no, I love it.
Well, that's awesome.
I mean, the biggest takeaway for me so
far this week has just been the people.
Like, these are, experts in their field.
Like I was talking to someone from, I
think it was Harvard, this week, and it
was like he is literally like creating
the academic progress that's helping
drive some of these things forward.
it was really interesting to just
pick people's brains, hear where
they're at on their journey.
, some folks are all about bringing,
, some of these capabilities back
on-prem, back on the endpoint.
and some folks are like, "Hey, we're
just now moving to the cloud. Like, I'm
not gonna run these workloads locally.
I need, , Microsoft or other vendors to
provide more here." So I think there's a
lot of opportunity for some of the people
at this table to, help show that value.
I know Parker, you and Will were
showing some of that this week as
well, so I'm curious how your session
went and, what were some takeaways?
Do you get questions afterwards?
Like- Yeah, we had small but mighty group.
Yeah.
rain impacted a lot of it there,
but I think this week to me was a
takeaway that demos will always win.
Mm. And I always wanna
try to get to that point.
Nowadays anyone can build a PowerPoint.
Mm. But how do we go through
and show the true value?
So Will and I, we tried our best.
I don't think we fully made it to
100% of all of our demos working,
but we put a slide up like, "Here's
a caveat. We're gonna roll the dice."
Yeah.
"You gotta deal, you gotta bear with
us here. We're gonna try to get to
where we need to be." I, I heard
from some colleagues on the, the
Agent 365 team, they had some similar
kind of things with engineer Ian.
But no, it worked out r- well, and we
deliberately went in ambitious where,
Parker was just doing the workloads on
his device and then we actually had our
own intra and Agent 365 and Defender
portal open looking at the local in-
telemetry coming off of his device and
GSA seeing that- Yeah the dynamic risk
scores, all of that just like doing
the side by side so that we can see
what this looks like from an activity
standpoint, but then your cybersecurity
team is not gonna have a heart attack
because they've lost visibility.
And I think the big value there
was the demo moved from surface
to surface, not Microsoft Surface,
but programming surface, right?
Mm-hmm.
We did it in the CLI, we did it
in Visual Studio Code, we did it
in the GitHub Copilot app, and
we also did it on github.com.
Oh, cool.
So we're able to really go anywhere,
and it was both hybrid, local.
Before we knew it, we had everything
covered and Will was blocking and
seeing real time, "You just tried
to call this MCP server that we
have blocked in our organization.
So it was really cool to see that and
show it at Edge For, for those of us
that happen to, miss your session, could
you- We will make sure to put a link in
the comments- See for them to find it.
Y- speaking of this, I feel like-
Yeah some of the things, Jacob, you
were saying was like the customers
are coming up to you with almost like
they're wanting to know more about
how it can solve their problems.
It felt like when we're always talked
about Copilot+ PC, we're telling
people about Copilot+ PC and what it
could do, but now you're almost like
there's a number of customers out
there that are well aware of what this
technology can do, but wanna know a
little bit more about, your take on it.
And I wanna bring Neil into the
conversation of Neil, as you're
talking to customers this week, do
you start to sense that as well?
Do you start to get a feel that,
customers have an understanding
of what these devices are?
This gentleman was saying that he
was spending almost $50,000 a month
on, cloud code, bills, and, he was
looking at actually the DGX station.
I'll say that they were-- he was
not in the NVIDIA's, Silicon booth
saying this to this gentleman.
But it was another Silicon vendor,
but he was more saying, he needs a
solution that can drive, and, bring,
more local AI into that space.
And I was just floored that, somebody
in a random conversation would be
talking about local AI in that way.
So wonder if you have
any input there, Neil.
A- also, what was your highlight, Neil?
We didn't get that yet.
Yeah.
we need to hear.
Like, you actually got to
watch all the sessions, right?
12 hours a day.
Yeah, Neil, what did you think of
our session when you watched it?
I look forward to catching the recap.
Met
a few customers this week, and
I just wanna reflect on how
far we've come as an industry.
I presented on stage, I was
with Snapdragon at the time.
We introduced X2, the
processor with 80 TOPS,
What, what can run locally on the device?
No one's asking that question anymore.
How do I route the right
workloads to the right place?
Highlight for me is Windows Server
2028 coming to fruition, Foundry
scaling across Surface as the
endpoint, Edge servers in Azure.
in one year we went from, "What can run
on my PC?" to, "What should run on the PC?
What should run on Edge servers?
What should run in the cloud?"
Intelligent routing.
And to me, Copilot Fusion continues to be
only a meaningful solution for local
code offload, but really a reference
architecture for cascading models,
right model for right task, and actually
deploying several different models to
accomplish different tasks in parallel.
So those are some of the highlights for
me, and just a reflection on how far we've
come just in the last year as an industry.
Yeah.
Crazy how much has changed Now I
wonder if other people had that
same experience of, like people
coming up and actually almost being
surprised at their level of knowledge.
I always point out like if you look
at the RTX Spark, there literally are
no 128 gig SKUs, available out there.
So obviously people are buying
these, and the geeks that come to
these shows have probably been some
of the people that did buy them-
Yes … and are starting to implement.
And I'm wondering like, if anybody
here has kind of their own story
of being surprised as to how far
ahead some of these customers
are in leading the conversation.
I had a guy actually ask "Why do
you even sell anything less than
128?" It's like, it's literally
the purpose of these devices.
And we had to tell him.
I was like, "No, no, no, like there's
actually other purpose to this device.
There's good creator models, and you
can still run things at 64 and 32."
But he was just like Like he looked
at us like, "No, you're an idiot.
It's 128 gigs or le- or nothing."
like, "Oh, okay, man." it's a, a wide
range of understanding, for sure.
And he had an empty
checkbook for that as well?
he, his was on order.
Oh, perfect.
Like he, he… Then that's
part of his complaint.
He's like, "I want it now, and it's not
here yet." But I mean- It makes sense
… if you have that customer that says, "I'm
spending $50,000 a month on coding for
my 40 developers," and you can offload a
portion of that to open code within some
other model that runs locally and start
to save, and it's doing some degree.
Or with Fusion, being able to find at
least cut it in half- Oh, yeah … where
you're, using, still Claude maybe or
Codex for some things, but you're using
local AI for others, it maybe starts to
make sense as to, Yeah … these devices
will literally pay for themselves.
Even if it is a DGX
station at $100,000, you…
The math is there that, if you're
using it to build a tool that's gonna
make money, then, the… Like, I'll
say mechanics don't really care what
their Snap-on tools cost because
they use them in order to repair
vehicles that, they make money with.
Yeah.
But with tools like this-
Like this … you're actually
making money with these tools.
That's the wild thing is like, I get to
see Surface devices all the time, and
I get to hold them and pick them up.
being able to see like DGX stations
and seeing those things in action, like
that's one of the things I think is pretty
exciting about this week is like they're
right over there and across the way.
Yeah.
And just the pure power and effectively
what looks like a giant heavy
desktop, but it's still just a giant
desktop that like- Yeah … crazy
amount of output on that thing.
And, and like even Strix
Halo device, like- Yeah.
Yeah.
Like, yeah … really capable- So
cool … high RAM- Small … AMD device.
Well, I know I'm known for hauling
around big, heavy devices like the
Surface Hub, but Parker, I think
you've built a reputation for hauling
around your own DGX station over, all
over Texas or something like that.
Yeah.
If you can tell the story.
Back row Forerunner.
Oh, you got a Forerunner too?
very good vehicle.
I also rental.
I also rental.
Okay.
Don't… I don't wanna be in the
Surface Hub Forerunner family.
Yeah, we can't pit those
against each other.
That's, that is the Surface
Hub vehicle of choice-
because this stand fits
perfectly into it as well.
Do you take it with you
when you fly around?
But tell us about like you're
taking the DGX station on the road.
Like, and tell us a little bit
about that device for everybody
that isn't aware of that device.
Yeah, I think it's just a massively
RAM-heavy device, really good
performance, really good capability,
and it started with an idea, right?
We have high spend.
We have customers that have the
same situation at Insight, and
we said, "Let's take a device.
We have good OEM relationships.
Let's go see what we can do."
We got the first one loaded in the
Forerunner, drove it from Austin,
Texas, to Houston, and had this
story of let's try to- Did you stop
and get some sausage along the way?
There's really good- I did feed
it beaver nuggets from Buc-ee's.
That was the only requirement.
Buc-ee's.
So a little jerky in one USB port.
But we got it to the customer site,
set it up, and really had a cool
opportunity to use GitHub Copilot locally.
And I think that's where
this all starts, right?
You have the laptop Surface Ultra
where you can do it, but what do--
That's a one-to-one relationship.
Like a builder class device.
But it's limited to the
individual using it.
Yep.
Now all of a sudden I have this, uh-oh,
my dev team, my pizza box or two pizza box
team is now all of a sudden needing more.
Mm-hmm.
And I can't afford laptops.
I'm not that gentleman that can
just show up and buy whatever.
I need a, a many-to-one device,
and that's where this really
plays and plays very well.
We had four models running on
it, all different complexities,
different sizes, different scales.
Tied all those natively right into
my IDE and was able to hit them.
Sixty, 65 developers.
We haven't hit a threshold
yet- … which is shocking-
Mm-hmm … considering what's going on.
So we're able to immediately see
that ROI from the cost of the device.
We're having the performance throughput.
And this is one thing I've been having on
the floor this week, day in and day out,
is everyone's focused on token savings.
We're to that point now.
My conversation shifted and said,
"You now own the device." Mm-hmm.
Tokens is one thing.
Return on investment from workloads
you've never done before- Mm-hmm.
You can try stuff now.
Yeah, right.
you can actually experiment without fear.
Well, I mean, you were in the belly
of, the oil and gas industry there.
I mean- Yeah I picture like one
of these devices out on a rig in
the middle of the Gulf of Mexico.
Mm-hmm.
Or I see it in the middle of a cruise
ship that is, running the cruise
ship's AI when it's not connected to
Starlink or something along that line.
I think about it sitting in the middle of
an engineering team that can't share their
IP into the cloud because that is their
bread and butter, And I was talking to
another Silicon vendor way back, and they
were s- using that example of, we can…
We missed the first wave of AI because
we couldn't put our data into the cloud.
Yeah.
And now you have a device
that is just as capable.
And I always say these models, these open
weight models are literally only what?
Weeks behind the frontier
models at this point.
And it's ex- like I said, it's exciting
to see MAI, as a Microsoft model
coming out in that space as well.
But maybe you can say, what
were some of the use cases you
saw for a device like that?
Yeah, so development
was first and foremost.
We're spending all of that money, and I
ki- I kinda wanna finish the story there.
But as we went through that
journey, it started to be just
development heavy, day in the life
of a developer, development team.
And then the story shifted, much like
we had on the expo floor here, it
became night in the life of an agent.
I can do all of these things overnight.
I can process documents.
I can do data manipulation.
Just because it's a GPU
doesn't mean you have to do AI.
You can do all the other workloads
you need to be efficient.
And the nice part with that same
situation is we start to layer in all
the Agent 365 capabilities that garner
our devices also now work on that.
So the same policies that we have
day-to-day in our use cases can be
pushed out to those devices as well.
So we can say, "Hey, yep, beefy, a
lot of data in them, lot of tie-ins
to sensitive information. How do we
protect it from DLP?" And it's the
same thing that we use today- Yeah.
Applies.
Yeah.
I mean, that's one of the really unique
selling points with, Agent 365 is that
it's meeting teams where they already are.
Mm-hmm.
You've already got Purview deployed.
You've already got Defender.
You've got Entra, Intune,
all of these things.
People don't have to learn a new tool.
as we've been going through the various
different pr- previews and working through
these things I haven't really had to
talk to our security engineering or, SOC
team about new alerts or new this, that
or the other, 'cause new alerts and new
types of things happen all the time, so
they're just used to seeing things new.
But they didn't have to go
anywhere different to see it.
It was in the same console.
And similarly with Agent 365 now
and the local governance, we can now
treat the agents running on people's
devices as their own identities and
give them the right permissions.
One of the things we showed in our
demo, was, using conditional access
policies to block my GitHub Copilot,
CLI from accessing a website,
and that was the Copilot, not me.
Mm-hmm.
so I could get there just fine in
differentiating that and really drilling
down and making sure that the agents
are doing what you expect them to do.
and we also saw some things with,
things like dynamic access control,
right-sizing and making sure
that we have good alignment with
what the models are trying to do.
Especially now, to your points earlier
about going and downloading these
models from Hugging Face or wherever,
we in security need to make sure we
are governing these agents properly
and these models correctly as well,
and giving good known provenance,
for these agents, and making sure
that only managed agents can run.
So Will, I was gonna ask a question
like, and this is a setup, but it
would be, like why would a customer
not just run Linux on these boxes?
what do you see the value that
Windows brings to, the compliance
and governance standpoint?
Certainly you can wrap a lot around Linux
to make it compliant and governed, but,
from a turnkey or just in simplicity or
like… I don't wanna put words in your
mouth, I'm just wondering like what do you
see out there as to some of the advantages
or, kind of deployment capabilities?
Well, I mean, outside of the Windows
ML thing that we mentioned, being a
really big driver, the other one is
the governance and management of it,
s- particularly around things like
Microsoft Execution Containers, where
we can make sure that the workloads
are running in the right place.
I think up until MXC was a thing, a
lot of people were saying, "Let's go
spin up a Docker VM and run my agent
in there," which for the average person
is okay, but that then turns into
a blind spot for me for IT, because
it's just sitting in this black box.
Now I can have the similar level
of security, if not better, with
MXC and manage all of that through
Intune and make sure that when
things are running in sandboxes, they
are running in visible sandboxes,
so I can have the defense there.
But also make sure that if an agent is
running, it's only writing or reading from
ex- approved parts of the file system.
Yeah.
Well, let's bring Jacob in here.
I think you guys actually
set that up on the floor.
I saw DAI giving the demo of it.
Maybe you can describe what
you were showing on the floor
with execution containers.
Yeah, I mean, there's a number of
applications that take advantage of it.
Pretty much from day one of
availability, a lot of the big
major players were supporting it.
Perplexity is one of the ones
that we highlighted on the floor.
but I've been using it with Unsloth and
with LM Studio and others and, it's just-
Seems from my perspective as not really
the security guy, just seems to work.
Like you set it to work your directory,
and when the agent-- you even
ask the agent to break out of its
directory, like it's working in that
area, it's working in that sandbox.
And so from a user perspective,
which I have a little bit of that
experience with, it just seems
to be a good user experience.
You just choose where you want it, what
data you want it to have access to,
and then those, whether it's cloud,
whether it's local models, it's,
it's working within those boundaries.
Is that right, Will?
Yeah.
Yeah.
In fact, we've actually
seen MXC as an enabler.
So we've been typically pretty shy
about local inference, and we've
very much restricted the people doing
it to people like Parker and others
that we work very closely with.
Just having that a-as that level
of trust and understanding,
knowing where to push, where not.
But I think now that we have this
capability, we can democratize
that a little bit more because
w-we've had the cloud guardrails
we've been able to put in.
Now we can put in the on-device
guardrails as well and give more
people the ability to experiment,
knowing that we've set up safe edges.
Yeah.
Well, what about you, Neil?
I mean, you were out there talking to
actual customers this past week and
dealing-- Well, I think all of you
were, because you were on the floor.
But I'm wondering, like in your experience
of talking about Windows and, execution
containers and other things, like if you
have experience of people starting to
nod their head like, "Yeah, it feels like
this is a, we're willing to accept this,
but where's the governance around it?"
Yeah, the, the conversation has
absolutely shifted that direction.
Now that we've identified targeted use
cases that can run on device, we've
got our right routing mechanisms,
run on the laptop, on an
edge server, on the cloud.
One of my favorite quotes from
Satya's keynote was, "Last year's
models can now run on your laptop.
Last month's models can run on your desk."
so as we usher in this era of local
intelligence, to Will's point, it's not
requiring customers to go and invest in
tools,
but rather take the existing
security policies and change
where those controls are enforced.
And I view execution containers
on device as a huge step forward.
tied into Agent 365, And so I
anticipate the next year to be, all
right, we've got these targeted use
cases, whether it's local coding,
loan analysis running
directly on a laptop Ultra,
workflows running locally on
Intel or Snapdragon devices.
Now the pendulum shifts towards, all
right, we need security to bless this.
Local doesn't equal automatic compliance,
but there's less attack surface.
that checkbox.
So
I, like to express the value of local
as you're in, you're on the fast
track towards security and compliance
and leveraging your existing
Microsoft investment
support not only cloud
workloads, but local now as well.
Did anybody hear of any cool, local
AI uses that you hadn't even heard of
before that people were thinking about or
thinking of crafting or things that they
were building or, at an individual level
or even at a larger enterprise level?
Like, 'cause our customers can do things
with technology that we just can't even
conceive of, and I'm wondering if anybody
has any stories to share on that one
I mean, the first thing that came to
mind was actually not an enterprise use
case, but it was still relevant because
someone was building, with their Hermes
agent locally on a local model, and
they were just talking about how easy
it was in the setup for them to just go
and say, "Oh, I could connect my cloud
provider," or, "Oh, you've got a device."
They had just bought a Laptop, Ultra, and
they've got a device that's capable of it.
Why not just click download on this model?
It downloads Llama CPP, gets
the runtime set up, and then
they're just off to the races.
So they were just tinkering, but
I thought that was interesting.
Like, that's also been my experience,
but it was nice to hear someone
else who, maybe wasn't paid to,
care about this, think that it
was a pretty easy experience.
I remember the Onyx framework was
announced like two years ago- Mm-hmm.
and we were like, "How do we get that
to work? I have no idea. What is that
thing?" And then all of a sudden, like
now you tell Claude Code or Codex,
like about the Onyx framework and in
under like five minutes already created
a stack on your device and you're,
coding up your own, app on the device.
So yeah, I mean, the tools are out there
for people to play with, and there's a
very clear stack that's available for,
these coding agents to take advantage of.
I can appreciate the, democratization
of all this stuff, right?
I just love the fact that I can ask those
questions and it just starts doing it
for me, and I understand it somewhat.
But the coolest use case I've heard
recently was actually the use of robotics.
Like it's crazy how people can like
just spin up robot arms doing crazy
things now because of how easy it
is to train the model themselves,
and they can bring that into their
house and run those things locally.
and it's pretty wild seeing all
that stuff take off right now.
I think to go with that a little bit, I
was down at university recently and one
of the kids was talking about local AI on
a robot to try to like drive it around.
I want to connect to the cloud.
That's just proof that
everyone's now a builder, right?
This isn't a limiting factor of I
have to have a cloud partnership.
I can run these models, SLM, LLM, on
whatever it is to start to learn and
take advantage of everything that's
out there, and I think that's been the
eye-opening pace of change that we're all
seeing is it's no longer a requirement.
Everyone can poke around and have fun.
They don't have to drive
around a Surface Hub.
They can now just try to
do things- They could.
What it means to them, only
if they have a Forerunner.
Right.
Well, I've heard both Neil and you,
Parker, tell the story of like, at
35,000 feet doing coding and, developing
something, and it feels like, that's,
that opens up an entire different world.
I mean, I remember Lennus Tech Tip,
he was more or less saying, "Who would
ever buy an RTX Spark, laptop, unless
you're actually flying in the air and
you're not connected because you'd
probably wanna just tail scale into
your- Yep … server, your RTX 6000,
processor that's on the back end."
but maybe you can also tell a story of
what, the RTX Spark has done for you.
And maybe Neil, you can share the
same story since I think, Neil, it
seems like you're like a consultant
for one of the big, consulting firms.
You're in the air more than you're
probably on the ground sometimes, so.
Yeah.
I, Neil, I guess we'll challenge you.
Watch my presentation.
We'll guess each other's number
of flights- … and go from there.
But I took 125 flights so far this year.
Ooh.
And I think it's one of those
situations to start to talk
through what does that mean?
Everyone loves, Wi-Fi on an airplane.
It's always so reliable.
but now I can do a lot of
work that I need right there.
And I actually have a story
that's not airline related.
It's back to that university trip.
We all went out to lunch, junior student
went out with us and said, "Hey, try
this device. Try this out." We're in
the middle of a restaurant, having some
sushi and other things, and he started
building a website and an application
and then an API, and then mock data.
And he ping, he like taps me
and goes, "This isn't working.
It's not connected to the internet.
I can't get Google to load." And I
said, "Exactly." But he just turned
it around, had the whole app built.
Forty-five minutes, no issue, good
performance, NI Coder one-one flash.
Great opportunity to kind of show
the power of what is coming there.
But it was airplane- Yeah … restaurant,
all the same capability, no
requirement to have tool calls, no
requirement to have outbound access.
Yep.
Knows the environment.
I can add it in and enrich
it with the context.
Same concepts apply.
Very cool.
you win in terms of number of flights.
Yes.
I'll share another example
outside of airports, airlines.
I was fortunate enough to represent,
Surface on the sidelines, now expanding to
Surface everywhere, supporting NFL teams.
I had my curated demos for all the
customers I was hosting, and then they
say, "Oh yeah, by the way, game day
Wi-Fi," you're gonna be lucky if you
have connectivity." I had all of my
demos queued up to run entirely offline.
60,000-plus folks at MetLife Stadium,
no connectivity required, all of my
local AI demos ready to go for customers
that I hosted this past weekend.
again, thinking about those
moments in time, those areas where
connectivity is a luxury, not a given,
can go on.
Good use case there, too, that we're
seeing a lot of use case start to coalesce
around is natural disasters, right?
Hurricane knocks out cellular, knocks
out Wi-Fi, knocks out whatever we need.
How do we meet those families
immediately in a time of crisis
with upload photos, here's a check.
Upload this, here's an opportunity for
you and your family to move forward
and not have those same concerns.
That's one we're starting to see a
lot of as well for these devices and
starts of-- kind of those questions.
Come into the booth, "Hey, have you
thought about this?" There's actually
a, a, a partner that we're showcasing
at the, booth that does it for power
restoration, which is really interesting.
Like, you think about, your local
power authorities, that they have to
go on site and fix things, and they
oftentimes have to go look at, the
issue, go back to the office, read
through this massive manual that is
impossible to, understand to begin with.
And they've taken that manual, shrunk
it down, and now they've gone and taken
a process that's a pain in the butt,
and they've just thrown it on any,
Copilot+ PC, any device that's small.
It's a small model.
it's a RAG model ef-effectively, right?
and they've changed the game overnight
in terms of how people can, access, like,
and fix some of these things on site.
So now you can take a blackout or a
storm out- outage, not to hopefully
it doesn't knock on… Wood.
We're not getting that one here,
but- Right and we can, and just
take care of it immediately.
So, it's cool where everything's going.
Well, I know I've talked about it before,
but I built that little app for my phone
that uses Google's, embedded Gemma,
2, which is like a 450 megabyte model.
And it runs on the phone, but it
has both imaging and, kind of like
structure of file system knowledge.
So you can literally just talk into
your phone and say, "Find me the wiring
diagram for- Yeah … my, inverter in
my Sprinter van." And my Sprinter van,
I could be, way off the grid, like
without Starlink working, and literally
being able to still see and fix, what
I need to, rather than trying to find
that file and trying to find… So
it understands everything that's in
my phone and how to find something,
kind of like Google Photos would.
But Google Photos runs in the cloud.
That's a cloud lookup.
This is taking that onto the device
and being able to run on the Pixel.
Nanda was, the ch- skip level for
both, Jacob and Chauncey, and, he had
come from Google Pixel as the, kind
of the chief marketing officer for
that product, for the Pixel phone.
And I always tell this story that
Pixel's largest, consumer or commercial
sale was to a customer that he was
able to show them that they'd created
a model, Google, a… Who better to
build Transformer to run on a phone?
but they built a model that ran on
the phone that was able to do image
analysis, and then the savings alone
of that imin- image analysis paid
for the phone in just a few months.
But I told that story back to Nanda when
I saw him this week, and he actually let
me know that the customer was Walmart.
So I was like, "Whoa."
So that's pretty cool.
I have no knowledge of this.
No knowledge.
None at all.
I'm actually curious, as we're on a
more industry talk track here, we've
got some thought leaders with us
today, and like we've been hitting
them with some softball questions-
I think, here and there.
Hitting each other with it.
Like, I'm curious, like what are
you guys seeing in the industry?
Mm. Like what are the
win… how are things doing?
It doesn't have to be local AI related.
Yeah.
Like, we're just curious on
like what, what are some things
that we should take away?
I mean, you've got a couple folks
that are working for Microsoft here.
Like, all of us are invested
in the industry together.
Like, I'm curious, Will, I mean, is
there anything to start with for you?
Well, I think the big thing for
us is, just the speed of attacks.
Mm that's a huge thing,
It's not an argument if we're gonna
adopt AI or, or what have you.
It's how fast, what
data can I, throw at it?
How do I defend my
systems against attackers?
Mm. And, one thing we're really
happy to see go GA, is Perception.
Mm. Mm. So we have been working with
the Perception team for several months
now, on that, doing some testing.
And, we're running it in our
environment and using it, and
we've just been blown away.
For the novices, what is Perception?
I'm hesitating 'cause I had that
question earlier and I tried to condense
it and I didn't do a very good job.
Wait for Neil to finish watching our
presentation- … and he can report back.
Yeah.
So Perception is agentic cybersecurity
in the Defender platform.
To start with, what GA now is very
focused around the agentic SOC, so
helping you do the investigations, alert
triage, creating detections, doing threat
hunts, pen testing, things like that.
But it's going to evolve beyond that
to more, with helping you do things
like deploy Purview or, detecting a
risk for your agents and responding
to that, It's really designed to
not replace your SOC, but give
them more capabilities and powers.
So where an investigation may
two-- may take two hours, you
could do it in 15 minutes.
Wow.
And the level of quality we've
seen come out of it- Mm-hmm … has
been absolutely amazing.
I-- When I was doing a deep dive
last week of some metrics, it was
reported 99.7% agreement between the
Perception agents and our human agents.
Hmm.
And it was just a number up on a, a, a
PowerPoint presentation, and they were
just going through, and I, "Hang on.
is that a right number?
is that something you just
pulled together?" "No, no.
This is based off of your incidents
and investigations, both current and
historical that we ran through." because,
we all know AI hallucinates, et cetera.
Sure.
but I need the confidence that if
I'm gonna use a tool like this,
it's gonna give me quality data,
and that number speaks for itself.
Wow.
That 0.3% was all of my actions-
… that just can't be contained, so.
It's just like, "What is this?"
I think- We trust Parker.
Right.
Yeah, yeah, yeah.
It's carved out.
I think we're still, whether it's AI
or non-AI related, still trying to
see people figure out what it means
to move forward, and I'm gonna put
that in a lot of different ways.
Does it move forward with AI?
Does it move forward with
application development?
Does it move forward with the next thing?
MXC is a great one.
Please dig into that.
That is a massive opportunity.
But also try to find a way to build in
very limited use cases in your business
to go through and start to play, and I
think that's been the other part too.
A lot of people are
hesitant to get started.
A lot of people are
concerned to get started.
What does it mean for us to
build that out and go from there?
So dig in, find those use cases,
carve out your environment,
and go and be successful.
That's what we're seeing
in the industry today.
Right.
And I think that's the only way
that this group can go forward.
That's awesome because the, the number
of people that I talk to even today,
like one of them was, m- may or may
not be a customer, it's a gentleman
who walked in the, the room just now.
they were talking, about how they're
just like racking up Mac Minis,
and they're just playing, and
they're testing, and they're trying
to get some RTX Spark devices.
And like what I was inspired
by was just the fact that
they're like exploring- Yeah.
… and they're allowing their
teams the freedom to test and
figure out what's right for it.
So it's great that you're
seeing something similar.
I think if you don't do
that, shadow IT comes back.
Yeah.
And that's my big concern, right?
We can secure it, and we can have
carve-outs, and we can be successful,
especially with the laptops, RTX
dev devices But if you don't give
that playground, I'm just gonna
go do it on my personal laptop.
That laptop now uploading all my
documents, I'm downloading my emails,
I'm doing all of these things that
Will's not gonna be happy I'm doing.
Yep.
So how do we go forward from
there and be succ- be successful?
Well, on that point, we don't
see a lot of these security tools
as breaking mechanisms to stop
productivity and innovation.
We see them as accelerators because,
we talked about this a little bit
earlier, having Agent 365, having these
capabilities on managing devices and
the models running there, it gives
us more confidence to experiment more
because we have that level of control
that we otherwise wouldn't have.
Just really quick to wrap that
up, and I got another question.
Local inference also is free, right?
I paid for the device, but
I can experiment safely.
That's the other big thing to me.
Foundry Local was a big start for that
last year, maybe a year and a half ago.
I can be successful in how I'm going
and driving that towards adoption.
And then as we move into production,
hey, you have the same evaluations,
the same things that I need to go
and run for guardrails that I've just
ran locally on my device as well.
It's moving from, half billion parameter
models to 140 billion parameter models-
Yep I mean, I said that I was gonna
get-- hit you with harder questions, but
you guys are Microsoft experts clearly.
know how to read the
room here, some folks.
But that's awesome.
I love to hear the
insights from both of you.
So while you guys are
talking, and I'm gonna put Mr.
Mike McGinley, on the mic 'cause
we're just gonna roll with this live.
But Mike actually came in the room.
Mike, why don't you come on
around and we'll, we'll bring
you on camera for just a second.
Because Mike, a former, a member of the
local host, always welcome to come on.
Mike, you're on camera.
And, but hey, I wanted to say you
represent some of the largest financial
services customers in the United States.
I'd love to hear a
quick, rundown from you.
Like, how are these customers
accepting AI into their workplace?
'Cause you in particular, I mean, we were
all at a dinner, so we can probably talk
about it 'cause it was, talked about live.
But Flagstar being one of your customers,
they were talking about the use of local
AI, and they've actually put a little
commercial together that went out, in
one of their sessions this past week.
So Mike, any stories you
can kinda share with us?
Sure thing, Frank.
Frank was, referencing a recent,
Surface win with Flagstar Bank.
one of the key takeaways from there
was, he didn't look at the Surface
device With a price tag on it.
He looked at it as a tool to
reinvent their branch of the future.
he's built all of these AI models, and
he's running them locally on the Surface
devices that they've recently purchased.
And he's got some, great stories
about how he intends to, reinvent
the bank or the branch of the future.
So it's just been an incredible
amount of momentum, especially
with the newest, laptop Ultras that
recently became available in October.
I mean, I love that, and I think I'm
gonna lean on Will for, like, when
you're selling local AI or AI to a
financial services company that has
to go through some of the biggest
compliance hoops, that are out there.
I'm wondering, like, how do you talk to
them, to be able to assure them that these
tools that are out there or, that there's
environments that they can start to trust.
So I'm wondering from a security
standpoint, what they think about there.
definitely a lot of concerns there.
it's interesting, though, you have more
of the governance in the cloud, so you
have kind of that comfort level, but now
we've come back the other direction there.
but a lot of the times we show it to them.
So we do a lot of demos where they're not
canned environments or anything like…
This is our internal environment, and
we'll just, open up the, the surfaces
and start working on things and then
show them almost a version of what
we did in our session the other day
of seeing the telemetry, seeing the
controls, and doing a deep dive into it
so that they can understand what they
can control and what they can't, and
then make that a risk-based assessment.
I mean, that's honestly a big repeat
because Mike can give the story here, too.
But, we helped build them a proof
of concept application that showed
the day in the life, of a banker.
we even used a local PhiSilica model that
made an MCP call into a Dynamics database
that helped you prep for your next
customer that you're about to meet with.
not only were they able to look across
that Dynamics to say, "This customer
has this amount of-- in their checking
account, but they also do this five
twenty-seven plan, or they do this,
this investment banking, component."
And we showed that actually to the
Flagstar exec, and he came back and
said, "I love this, but could you also
tell me the products that the customer
doesn't have, so that when we do
meet with them, we can provide that?"
And so sure enough, the PhiSilica model,
tiny little model, like what seven
billion parameter model, was able to
work across all of that data and then
come back with a report that said, you're
about to meet with, Elena Martinez.
she has two kids.
They have a five twenty-nine plan, but
they don't have an investment plan and
maybe this is something you should work
on them to recommend and maybe some of
the language or the encouragement that
they would use in order to do that.
definitely from a bank, especially in
that lending, you need to have visibility.
You need to be able to report on
and attest to the fair lending
practices and things like that.
And having that visibility is critical
to be able to do that, to show your
compliance officer and the teams that
you're being compliant and that you Have
the receipts for what that agent did
and why it made this recommendation.
So that if someone does push back on
it, you can show them the clear details.
And that also goes back to the
explainability point that I made
at the, the top of the s- the
session, is we need that visibility
to be able to explain ourselves.
It can't be a black box.
Yeah.
Yeah, I think really quick on
that, attestation, huge thing that
Microsoft is working on, right?
We can have attestation around the model.
We can go and do what we need.
That's a massive value.
One feature, I don't know, maybe
we're gonna have a wish list here
soon, but one feature I would love
to see is like an artifactory of
models hosted inside of Foundry.
Mm-hmm.
This is what my team is blessing.
I don't want that teammate
going to Hugging Face.
Please don't type in your Hugging Face
key somewhere Let's give them a list
of what's prescribed and give them the
capability of these six models can run
on your device based on your permissions.
Here's a stable version, here's a
release version, and go from there.
That's something I'm hoping comes
soon so that we can get off of just
download, download, download, and
pray that it's the right model.
Well, maybe we should get back
to the, Ignite show floor and
talk to the Foundry team and- Ah.
-add that to their wish list.
Yeah.
And I'm not walking
back down that pit, so.
All right.
Well, before we go, I wanna ask Mike,
anything else you saw that was fun or
cool on the show floor this past week?
Well- I know you were
working up in the store area.
Yeah.
Well, I… Like I s- I shared with
you earlier, Frank, out of all the
exciting things that happened this
week here at Ignite, I think the most
exciting thing was running into you- Ah.
Mr. Buck Holtz.
no, I gotta tell you, I, I was on
booth duty for the last couple of days.
I was, actually up at the Surface
store booth and, I g- I, again, I can't
tell you about, all the excitement
of every person that came in there.
we would've, we would've just- Sold
a lot of 'em … did better than we,
we ever could've managed, imagined.
But yes, there was a, a ton
of, momentum built around the
latest Surface Laptop Ultra.
And, we've had, a, a ton of foot track,
ton of foot traffic with, customers
walking out with 32 gig and 48 gig devices
and, it was just awesome all around.
Well, thanks, Mike.
Yeah, you bet, Frank.
Thanks for making the cameo.
Thanks.
Thanks for having me.
And I do see a couple more of
those chocolate-covered almonds.
The chocolate almonds.
Is it cool, is it cool if I, No,
of course … go break that in?
And I also have to check
what's in your fridge too.
Yeah, yeah.
I do wanna go back to one thing.
I think that was the call out of that
device of the future, and we've talked
about it, cost, and Chauncey, maybe you
can chime in as you go through this, is I
d- had to defend myself from the trolls.
That involved explaining what a troll
was to a four and a seven-year-old.
When I posted on LinkedIn about
flying the devices around and how
we're using them, cost, everyone
wants to talk about it, right?
Those devices, how do we go about it?
We joked about it earlier, blank check,
but it's a $60 a month difference
between a MacBook that has 64 gigs
of RAM and going to this device.
If you can prove $60 of value a
month over three years, which is a
typical depreciation cycle, you've
paid for the device difference.
A 64 gig MacBook is not
gonna run 128 gig capability.
I think that's where I just
wanna call that out as well.
What we've seen in the
marketplace is, "Oh, Buy me Max.
Buy me Max." It's not the solution.
I actually, would The sticker
shock, I mean, 'cause we're so used
to selling entry premium devices-
Mm-hmm … but still not $6,000 devices.
Even for me, it was initially,
ooh, that's like, what is that?
I just didn't know what
that, meant to people.
But being at this event, I think more than
anything really solidified the fact that
actually that's a steal in- Yeah And every
customer I talked to when I was like,
"Yeah, it's 5,999," they're like Okay.
Like it was just like a, "Okay, that's
fine." They're like, "That's cheaper
than a tool." Like, "This is, yeah,
this is phenomenal." Like, and they
really were just like, "Okay, well, how
do I get one?" and I think the trolls
that are, we're running into the most
are the f- not the right customer.
Right.
Yeah.
And I think that's a unique…
That's what it became, but also, do
you feel the $6,000 price tag has
become more acceptable as GB10s are
now eight plus thousand dollars?
It's a tool, right?
that's what I tell my team is like-
Right … if you're not in the weeds
of some of these, how much this stuff
is actually costing right now- Right.
Yeah.
You wouldn't use this.
Well, we have another podcast member,
Robert, he would always famously say to
me like, "I will never sell service to a
customer that doesn't have Microsoft 365."
But now I think he's taken on the
mantra of "I'm not gonna sell, RTX
Spark or AI devices to customers
that just don't understand what the
capabilities of those devices are." Yeah.
I wonder, Neil, like if you have,
customers that, you're dealing with
from the price perspective, and, how you
overcome that, or, what you show as the
value of these devices into the space.
Yeah, absolutely.
I think we have some great
examples in the room.
Parker, I'm not gonna hold you to
sharing the exact number, Ultra may have
already paid for itself based upon- Yeah.
…
Copilot bill was previously.
think it's, it's targeting
those AI power users, right?
we do have token economics calculators.
We can show a workload cloud optimized,
and we can show that shift with not the
entire coding environment running locally,
but again, using like Copilot Fusion
to use local models when applicable.
So
it's often starting with
what is the workload.
the
ROI calculator to back
up that justification,
yet to meet a development team that
doesn't wanna test RTX Spark, Really
quick on that, I have paid for the device-
…
in two ways.
Generated memes, number one.
Oh, good.
I could use memes.
I could just generate
a meme straight away.
Worked fantastic.
That's worth its weight in gold.
I don't care what the device costs.
And tokens from development.
And I think the Copilot Fusion callout
that you have there, we were GitHub
Partners of the Year this year.
Massive impact.
Congratulations.
Thank you.
I have sweatshirts to prove it.
Wow.
I have a title belt that
weighs like 10 pounds.
Title belt.
it's a belt, like a huge- It's a
literal belt … wrestling belt.
Like it's massive.
That's cool.
It's really cool.
That's awesome.
it weighed way more than I thought
it was going to when I grabbed it.
Like, whoa.
All right, we're coming back.
but with that, we have massive partnership
in that ecosystem, and we see our numbers,
and people have talked about Copilot
and Copilot Fusion, usage-based billing.
Copilot Fusion has made an impact.
We're seeing the usage-based billing
numbers come through across the
various numbers, and it's down.
Costs are down, which is fantastic.
Not good for maybe CSP numbers and
everything else, but great for where
people- I think they'll do okay.
They'll be okay.
Great for where- … great for where
the development needs to be, right?
My budget is now going
to the end of the month.
It's not running out on the 20th.
And then we dug into those numbers.
It's a blend both of Copilot Fusion
as well as on-device inference, right?
Mm. So picking the right
model is fantastic.
I know, Will, we were featured on a slide
earlier today that, Micoder 1.1 Flash
is, what, 70% on the suite benchmark.
Yeah.
It's like 70.2% on device
on suite benchmark testing.
like 72 if it's run in the cloud.
So that's even closer than we
were anticipating it to be there.
So Copilot Fusion, Copilot in general,
is now helping get to that point.
Just wanted to make that call-out of,
yes, I've paid for it, however… Yeah.
All right, gentlemen.
Well, I know that it is, 5:30 here-
Yep … local, and for some sad
reason, Jacob and Chauncey still have
to go back and work two more hours
down on the floor until 7:30 tonight.
at least.
Minimum.
At least.
Although tear-down seems… It
takes… I always find it takes
a week to build a booth, and it
takes, an hour to tear it down.
Motivated.
But- Yeah.
Yeah, motivated people, yeah.
Well, last… I've been there where
the, literally the people, the, the
crew is, working around you to tear
stuff down, and you're like, "I'm
trying to get out of here, too,"
But, maybe we just go down the line,
just to get, anything, any last
word, any last, kind of appreciation
of, being here at Microsoft Ignite?
Chauncey Yeah.
I'll start.
appreciation, Parker and Will,
thank you so much for joining
us- Absolutely … guys.
I've probably learned more from you
guys than my, entire time here at Ignite
just because you've been on the floor.
So thank you so much for, spending
the time, but also just, appreciation
in general just to have the
opportunity to be here and, at
Ignite, work alongside you every day.
appreciation, comfortable
shoes and an umbrella.
Oh, man.
Yeah.
I think that's a pretty
easy find in there.
Umbrella.
Yep.
but in all seriousness, I think it's
a focus around people willing to push.
Mm. Microsoft willing to make the
investments to do both hardware
and services in the cloud to
help advance the industry.
That's my appreciation,
pretty s- plain and simple.
Yep.
Look forward to trying it out
and wasting more of tokens.
Yeah.
Will?
I think for me is I've been working
with a lot of folks, on the backend
in product- Mm … and engineering,
and getting to meet some of them
this week has been fantastic.
And also just some chance encounters.
Parker and I met someone on the
Windows Security team earlier talking
about some of the things she was
doing and what we were seeing, from
our customers and in the future.
And to echo Parker's point, really
being willing to lean in on where the
frontier is and work with us based
off of what we're seeing to drive the
direction and really anticipate what
our customers are, don't even know they
want, but will need six months from now.
That's amazing.
And Jacob?
Yeah.
For me, I mean, it's all about learning.
I've been learning on this journey, and
a big part of the reason why I love being
on this podcast is it forces me to learn
things- Mm-hmm … from you all, from
guests we have on, like Parker and Will.
and part of that is sharing
the things that I get to learn.
And so I've had a few folks actually
this week at Ignite come up and say,
" Hey, I'm listening to Local Hosts."
Like, "It's so cool to meet you in
person." And that's, Mind-blowing.
I mean, this is obviously, the actual
type of audience that may actually
listen, but it's still really
cool to have some folks that are
learning on this journey with us, so.
Yeah.
Are you an influencer now?
No.
No.
People recognized you.
No, no.
You are an influencer though-
Yeah, you are … because of
that Laptop Ultra presentation.
and like part of- 30, 30,000 people.
Yeah, I didn't expect that one.
Yeah.
And then Neil.
Yeah, a note of gratitude.
It is the season of Thanksgiving.
Wanna give a big shout-out to Frank
for the vision, the execution,
bringing Local Host podcast to life.
a joy to have been, be
part of this journey.
look forward to this hour every week.
to our special guests, Parker and Will,
thanks for the continued partnership.
I think we were hamming
well into Sunday evening.
Parker's point about pushing each
other to be better, you absolutely
frontier, through and through, right?
pushing the boundaries of what's
possible, not just with Surface
hardware, but the broader stack.
So excited to see E7 and Surface
featured together on stage.
Can't wait to catch all the recaps, but
thank you all for the opportunity to
join you here- for the podcast today.
you
know where you can see it?
In our presentation.
And I think I heard Neil work in a Neil
I am or a Will.i.am joke in there again.
I don't think he did it deliberately,
but thank you for calling that out.
Yeah.
It's actually an inside joke,
or not even an inside joke.
If you listen to Local Host, Neil's,
Neil was hobnobbing with will.i.am,
and he's, he… It's in his brain now.
He's a DJ.
he's teaching a course at ASU.
He got a grant from NVIDIA, and
all of his 200 students have Sparks
create digital twins of their creative
personas, so- Sounds like a future-
… I'm glad we could do our presentation
guest host on the- There we go.
There we go … Local Host.
So gentlemen- One last point of thanks.
Yeah.
I'm gonna give a thanks to you, Neil.
Yeah.
You've been a fantastic partner,
working with you and, Andrew, your
colleague, getting the devices for
Parker and I so we can do these things.
Yeah.
So thank you for everything you've done.
Very grateful.
Absolutely.
All right, gentlemen.
Thanks.
Thank you much.
That's a wrap.
127.0.0.1. There's no place like home, and
this is the Local Host, so bye for now.
