প্রতিলিপি 🔗
WEBVTT
NOTE
Transcription provided by Podhome.fm
Created: 06/20/2026 14:09:38
Duration: 2098.808
Channels: 1
1
00:00:07.759 --> 00:00:08.879
Maybe you can
2
00:00:09.120 --> 00:00:12.480
tell the audience who you are, what you do, what you like.
3
00:00:14.135 --> 00:00:15.655
Yes, so I'm
4
00:00:15.655 --> 00:00:17.895
a solution engineer at Microsoft.
5
00:00:18.935 --> 00:00:20.455
What I do, I do
6
00:00:21.895 --> 00:00:22.615
mostly
7
00:00:23.415 --> 00:00:25.335
everything gets related to
8
00:00:26.230 --> 00:00:27.830
deploying applications
9
00:00:27.830 --> 00:00:30.390
to Azure. So that's our
10
00:00:31.430 --> 00:00:33.590
serverless products, right? So that means
11
00:00:34.070 --> 00:00:37.989
from Azure Functions to API Management to AKS,
12
00:00:38.230 --> 00:00:38.870
that sort of thing.
13
00:00:41.065 --> 00:00:42.745
And also focusing on now,
14
00:00:43.145 --> 00:00:45.065
as we all are doing on
15
00:00:45.065 --> 00:00:48.665
on AI, of course. So also Microsoft Foundry is also part of my
16
00:00:49.305 --> 00:00:51.705
some of the responsibilities that I have.
17
00:00:52.345 --> 00:00:59.480
But not only how you host the applications, also how you can build them using, in our case, GitHub Copilot
18
00:00:59.640 --> 00:01:02.280
in combination with Visual Studio Code
19
00:01:02.440 --> 00:01:04.360
mainly, but also nowadays
20
00:01:04.360 --> 00:01:08.280
a lot in the GitHub Copilot CLI mode, so from a terminal.
21
00:01:12.064 --> 00:01:20.545
Lately I'm really occupied with especially spec driven development, that's one of the passions that I have. So last year I got in touch with
22
00:01:20.865 --> 00:01:23.664
our internal GBB
23
00:01:23.100 --> 00:01:23.980
organization
24
00:01:24.220 --> 00:01:25.020
on
25
00:01:25.820 --> 00:01:27.420
all spec driven development.
26
00:01:27.820 --> 00:01:39.185
And that's where I talk a lot about nowadays with customers on how they can implement that in their own software development life cycle. So if it's brownfield or greenfield applications
27
00:01:39.585 --> 00:01:41.025
and how they can do that.
28
00:01:41.665 --> 00:01:44.625
On a personal note, I live in Aemness with
29
00:01:45.585 --> 00:01:48.065
my wife and children. I have chickens,
30
00:01:48.065 --> 00:01:49.505
I have guinea pigs, I have a dog.
31
00:01:50.220 --> 00:01:51.899
So it's a zoo at my home.
32
00:01:52.460 --> 00:01:55.579
And that's the way how I like to rewind a
33
00:01:55.979 --> 00:01:57.740
little bit from the tech side and
34
00:01:58.939 --> 00:02:10.335
and do also some things in yeah. With my family, but also in nature with with the animals around me. So, yeah. I have a nice anecdote about chickens, by the way. So that's that's
35
00:02:10.335 --> 00:02:16.175
the one that's that's really nice. If you wanna hear, then I will do it. With this girl. This girl. Okay. So
36
00:02:17.135 --> 00:02:19.055
it's a it's a fun fact. Right? So
37
00:02:19.640 --> 00:02:22.920
a chicken has a
Load full transcript
really low heart rate. So
38
00:02:23.880 --> 00:02:30.040
what I do in the weekend sometimes is just grab a chicken and the chicken that are really like they're
39
00:02:30.600 --> 00:02:34.345
used to be picked up, but they have a really low heart heartbeat,
40
00:02:34.504 --> 00:02:39.225
right? So their heart rate is much, much lower than from us humans normally.
41
00:02:40.665 --> 00:02:43.305
And when you have the chicken close to you,
42
00:02:43.784 --> 00:02:47.305
the heart rate of the animal gets a little bit transferred to you. So
43
00:02:48.880 --> 00:02:52.240
your heart rate will go down as well. So if you're a little bit stressed
44
00:02:52.880 --> 00:02:53.440
and
45
00:02:53.760 --> 00:03:01.600
well, I'm not that much stressed anymore because I have chickens, of course, but I just grab a chicken and I hold it under my arm for about maybe
46
00:03:01.600 --> 00:03:09.995
an hour and my heart rate goes down as well. So at one point the chicken even goes to sleep because it's just underneath
47
00:03:09.995 --> 00:03:16.715
my arm. Yeah, it just goes to sleep, but it's really like it's really calming you down to a sense that it's really amazing.
48
00:03:17.280 --> 00:03:18.320
So I
49
00:03:18.640 --> 00:03:28.319
can I can recommend everybody get some chickens and cuddle your chicken in the weekends to lower your stress levels? That that's that's that's a cure that works for me.
50
00:03:29.925 --> 00:03:34.645
Wow. I did not know that. And that makes so much sense because there's
51
00:03:34.645 --> 00:03:39.845
science about heart coherence where the heart actually has like a force field, an electromagnetic
52
00:03:39.845 --> 00:03:41.444
field that can influence
53
00:03:41.900 --> 00:03:42.620
other
54
00:03:42.700 --> 00:03:43.580
electromagnetic
55
00:03:43.580 --> 00:03:55.900
fields like your heart, for instance. Yeah. That's so cool. Yeah. It's reciprocal. And it really works. It's it's it's it's a it's a really yeah. You need to yeah. You need to try it out. So come in my home, my buddy, and I will let you cuddle my chicken.
56
00:03:56.715 --> 00:03:58.315
That sounds great.
57
00:03:58.795 --> 00:04:01.835
Cool. Alright. I will. I will. Thank you.
58
00:04:02.955 --> 00:04:04.475
I think that's a cool fact.
59
00:04:04.875 --> 00:04:13.140
So and and just before that, you spoke about all sorts of things, AI. Right? You touched from GitHub Copilot to Foundry
60
00:04:13.140 --> 00:04:15.940
to Spectre of Development. That's a lot of stuff.
61
00:04:16.420 --> 00:04:23.460
Yeah. Maybe we can focus on, like, what do you use daily? How do you how do you work and how does AI fit in there?
62
00:04:24.419 --> 00:04:25.540
Yeah. So
63
00:04:26.015 --> 00:04:30.335
to be honest, I think this is something that really changed in the last,
64
00:04:30.575 --> 00:04:32.575
I would say, five to six months
65
00:04:32.975 --> 00:04:35.695
from to where I am right now. So
66
00:04:36.015 --> 00:04:39.250
this is a really, I would say, a fast changing
67
00:04:40.210 --> 00:04:41.730
environment that we're in.
68
00:04:42.130 --> 00:04:42.930
And I
69
00:04:43.170 --> 00:04:51.250
keep it keeps reminding me when I talk to customers is that we are sometimes, as from Microsoft's perspective, we are sometimes in a bubble
70
00:04:51.665 --> 00:04:55.105
because we are mostly using the frontier
71
00:04:55.825 --> 00:04:58.065
yeah, I would say the frontier solutions.
72
00:04:59.425 --> 00:05:01.025
And it's really hard to
73
00:05:01.345 --> 00:05:04.305
grasp and to imagine what you can already do with it now,
74
00:05:06.820 --> 00:05:09.780
especially when you're outside of that bubble, right? So that's what
75
00:05:10.340 --> 00:05:12.340
I see a lot with my customers. But
76
00:05:12.980 --> 00:05:14.660
for me, I use
77
00:05:15.860 --> 00:05:24.555
I'm to be honest, I minimise most of the applications that I have running on my computer, so on my laptop. I only use a few nowadays.
78
00:05:26.395 --> 00:05:26.955
And
79
00:05:27.275 --> 00:05:28.075
that's
80
00:05:29.515 --> 00:05:32.235
for me, that's Microsoft three sixty five Copilot,
81
00:05:32.315 --> 00:05:36.315
where I basically already draft my emails or already
82
00:05:36.395 --> 00:05:37.355
use it to,
83
00:05:38.440 --> 00:05:40.520
well, to search internal
84
00:05:40.520 --> 00:05:50.919
documentation or to find information about things that are relevant to me that I need to look up. So I already use my Microsoft three sixty five Copilot a lot already.
85
00:05:51.159 --> 00:05:59.825
I use it on a daily basis. But when it comes to more the development perspective and especially when it comes to the things that I do for Azure in combination with
86
00:06:00.705 --> 00:06:06.700
GitHub Copilot and Visual Studio Code, the development parts and the hosting parts, is that I
87
00:06:06.940 --> 00:06:07.740
nowadays
88
00:06:07.740 --> 00:06:11.660
only sit in the CLI. So only in the terminal
89
00:06:12.060 --> 00:06:15.900
using GitHub Copilot CLI and connecting it with
90
00:06:16.060 --> 00:06:18.140
Azure via MCP servers,
91
00:06:19.345 --> 00:06:26.545
creating skills or getting skills for agents to be more effective and to help me achieve some of the things that I need to do.
92
00:06:30.625 --> 00:06:38.530
And using spec driven development basically to set up some of the things that I want to create. So I have a lot of ideas and I have a lot of things that I like to build.
93
00:06:39.090 --> 00:06:45.889
And we are now at the stage where basically everything is possible, right? So you can now create anything that
94
00:06:46.129 --> 00:06:46.770
you can think of.
95
00:06:47.425 --> 00:06:48.065
So
96
00:06:49.265 --> 00:06:52.465
and I think that's the nice thing of this
97
00:06:52.785 --> 00:07:00.305
age where we're because I started programming when I was really young. I started in QBasic when I was seven years old. So I
98
00:07:00.465 --> 00:07:02.465
was there from the beginning
on, I would say.
99
00:07:03.480 --> 00:07:07.160
And I've been a developer and hooked on to development ever since.
100
00:07:09.160 --> 00:07:18.120
And I really lost being in the terminal and being sorry, being in Visual Studio Code for example, or being in Visual Studio to create my applications.
101
00:07:20.034 --> 00:07:22.595
But that that one changed like in a
102
00:07:22.995 --> 00:07:35.710
six or seven months from now where I mostly transfer now to the to the to the CLI and create the things that I that I want to create. And so basically the only things that I have to I have open at the moment is just
103
00:07:36.190 --> 00:07:40.110
my browser, my GitHub, my terminal, my GitHub Copilot CLI
104
00:07:40.110 --> 00:07:40.830
and
105
00:07:41.070 --> 00:07:43.310
Microsoft three sixty five Copilot.
106
00:07:44.884 --> 00:07:50.405
And I'm not and I try not to be as much in our internal systems as they can be confusing
107
00:07:50.405 --> 00:07:52.885
and frustrating sometimes as well. So
108
00:07:53.604 --> 00:07:56.165
I like to then prompt I to prompt
109
00:07:56.405 --> 00:07:57.764
then
110
00:07:57.260 --> 00:08:04.700
better than I like to go into an application and look for the things that I need to do. So that's my change. That's how I use it nowadays.
111
00:08:05.740 --> 00:08:10.314
It's weird how it changes, right? So I do the same. I
112
00:08:10.314 --> 00:08:20.794
I actively develop also production software. Like, I have a side hustle, which is a podcast hosting platform. I do that, like, in the evening, which is production stuff that actual users use.
113
00:08:21.034 --> 00:08:29.670
But, like, six months ago, I used to do Visual Studio. Actually go in and then see, I want this feature. Let me just think about it and build it actually.
114
00:08:29.990 --> 00:08:33.670
And now I just basically one shot most things.
115
00:08:33.910 --> 00:08:34.870
Don't tell my customer.
116
00:08:37.635 --> 00:08:38.275
Yeah.
117
00:08:38.595 --> 00:08:39.795
But we
118
00:08:39.715 --> 00:08:41.635
in a time that we
119
00:08:41.955 --> 00:08:42.835
Yeah, sorry.
120
00:08:43.395 --> 00:08:46.595
Yeah. And then I just test it, it works,
121
00:08:46.915 --> 00:08:51.259
and then I verify that it works, and then I just check it in and boom, the point, that's it.
122
00:08:52.300 --> 00:08:53.579
And that's
123
00:08:55.500 --> 00:08:57.579
the thing that I think a lot of customers
124
00:08:58.220 --> 00:09:01.259
and mostly outside of the bubble, they
125
00:09:02.060 --> 00:09:06.105
cannot grasp it. You can achieve so much in such
126
00:09:06.425 --> 00:09:06.904
short
127
00:09:07.865 --> 00:09:15.225
amount of time and create the things that you could not create before. Because, for example, I'm not a UX designer,
128
00:09:15.704 --> 00:09:27.550
but I'm able now to give my agent the skills to be an excellent UX designer because I can just get get that skill from somebody who is and who created that skill. And then you can see
129
00:09:27.870 --> 00:09:29.950
that it creates the things that you couldn't
130
00:09:30.774 --> 00:09:31.575
imagine,
131
00:09:31.575 --> 00:09:35.735
but you have the idea. So that's how it works. And to be honest,
132
00:09:36.135 --> 00:09:37.735
I get the one shot farts,
133
00:09:37.895 --> 00:09:48.100
But when you use spec driven development, it becomes even better, because then you're gonna make it more concrete and specify it in a certain way that it becomes more production ready.
134
00:09:48.900 --> 00:09:59.595
Yeah. So maybe let's talk about it. Also talk about spec driven development with customers that are interested in it, that they are using it to co pilot, but they want it to be more structured,
135
00:09:59.595 --> 00:10:01.195
more more guardrails,
136
00:10:01.275 --> 00:10:02.555
production ready. So
137
00:10:03.355 --> 00:10:10.440
I teach them, you know, this is for instance, spec kits. These are the steps you do to to, you know, do a feature for instance.
138
00:10:10.920 --> 00:10:15.800
How do you use that? And what what thing do you use? Do you use spec kits? Something else?
139
00:10:16.120 --> 00:10:16.759
How does Yeah. That
140
00:10:17.639 --> 00:10:18.680
That's a good question.
141
00:10:19.160 --> 00:10:24.195
So so basically, spec driven development is not something that we from Microsoft own, right? It's methodology It's
142
00:10:25.155 --> 00:10:27.475
that basically everybody can use.
143
00:10:28.755 --> 00:10:29.155
And
144
00:10:29.795 --> 00:10:35.795
you can see that some of the frameworks that popped up as like, oh, would say open source frameworks
145
00:10:36.470 --> 00:10:45.190
around spec driven development. Just as you mentioned, SpecKit is one of them. But then I've noticed that not all customers like the approach of the,
146
00:10:45.750 --> 00:10:47.590
I would say, SpecKit,
147
00:10:47.590 --> 00:10:57.005
for example. They sometimes find it a little bit too much, a little bit bloated. There's a lot of things in there you need to get into that structure specifically.
148
00:10:57.485 --> 00:11:01.165
So I, for example, also use OpenSpec as
149
00:11:01.885 --> 00:11:02.605
a counter
150
00:11:03.940 --> 00:11:06.180
specification mechanism to
151
00:11:07.220 --> 00:11:11.300
basically spec my features and also spec the things that I want to create.
152
00:11:12.420 --> 00:11:12.900
So,
153
00:11:14.180 --> 00:11:16.820
for example, in comparison with SpecKits,
154
00:11:17.125 --> 00:11:23.605
OpenSpec is much, much lean and mean. So it basically only has like three operations which are
155
00:11:24.165 --> 00:11:29.445
three prompts that you can use, which is basically creating the change proposal,
156
00:11:30.320 --> 00:11:39.680
applying the change proposal and then archive it, which is which is pretty simple from a from a standpoint. But the nice thing is that when you
157
00:11:40.240 --> 00:11:48.595
when you create that and you archive it, it will it will create like
this generic spec for the whole project. So basically,
158
00:11:49.155 --> 00:11:54.915
you have created a spec to create a certain like a podcast, a podcast application
159
00:11:55.490 --> 00:12:03.490
and you want to add that so when you archive it, it has the contents of what the project is, which you created basically,
160
00:12:03.970 --> 00:12:05.329
the podcast application.
161
00:12:05.410 --> 00:12:10.850
And then, for example, when you add a new feature on top of it, it will just open up a new feature spec.
162
00:12:11.405 --> 00:12:18.925
And when you apply it and then archive it, it will just embed it into the, I would say, the generic spec
163
00:12:19.485 --> 00:12:21.325
to make it consistent
164
00:12:21.565 --> 00:12:24.045
across the whole project. So you have
165
00:12:24.480 --> 00:12:30.320
feature requests in a sort of a spec way and you have separate specs that just explain
166
00:12:30.480 --> 00:12:34.320
the complete project that you're working on. So that's it's different
167
00:12:35.520 --> 00:12:43.144
well, spec end uses it in a similar way, only SpecInt uses more the, I would say, the branching
168
00:12:43.225 --> 00:12:44.505
mechanism,
169
00:12:44.745 --> 00:12:49.545
where OpenSpec doesn't use branching in that case. So it's much more free, it's much more open.
170
00:12:50.345 --> 00:12:53.625
I like to use like the GitHub Copilot
171
00:12:53.625 --> 00:12:56.240
plan mode as well. It's
172
00:12:56.240 --> 00:13:14.385
also a matter of creating specs in a way that it makes sense to use to make the outcome to be more of a higher quality. So yeah, I use that as well. So basically for me these are the three things. I also heard from some of my colleagues that use superpowers a lot,
173
00:13:15.345 --> 00:13:16.625
which is also nice.
174
00:13:17.985 --> 00:13:20.305
Superpowers is basically also a
175
00:13:20.465 --> 00:13:21.185
set of
176
00:13:23.660 --> 00:13:24.860
custom agents,
177
00:13:24.940 --> 00:13:29.420
skills and prompts that you can use to basically do brainstorming
178
00:13:29.420 --> 00:13:30.140
or
179
00:13:30.700 --> 00:13:32.140
creating specifications
180
00:13:32.140 --> 00:13:34.700
and applying that. It's a little bit more extensive,
181
00:13:35.105 --> 00:13:37.425
but it has really nice features
182
00:13:37.505 --> 00:13:39.425
specific for certain use cases.
183
00:13:40.225 --> 00:13:43.185
So for example, the custom agent on brainstorming,
184
00:13:43.185 --> 00:13:58.190
that's a nice one, right? I think these are this is how we now nowadays, when we are working with AI, we want to have we want to work with the specialists that can do that task for me, right? So when I want to do specifications or brainstorming, I would like to have
185
00:13:58.590 --> 00:14:02.990
an agent that helps me brainstorm on the idea I have to get it
186
00:14:03.365 --> 00:14:04.645
as crisp as possible
187
00:14:05.524 --> 00:14:08.565
and then generate the spec. So then different methodologies.
188
00:14:08.565 --> 00:14:15.204
I use different ones because I think it's good to know what is happening, but you can I see companies also create their own
189
00:14:15.940 --> 00:14:20.180
own spec driven approach or pipelines or whatever?
190
00:14:22.500 --> 00:14:23.140
And
191
00:14:23.620 --> 00:14:24.740
what I always
192
00:14:24.980 --> 00:14:39.165
wondered so I do a lot of GitHub Copilot workshops. Like one day I talk about everything that you can do, including spectrum development and all the features and things that you can do to customize, custom instruction skills, custom agents, all that type of stuff.
193
00:14:39.565 --> 00:14:42.045
And I always wonder to myself,
194
00:14:42.630 --> 00:14:45.190
what am I doing? Is this really necessary?
195
00:14:45.910 --> 00:14:48.790
I can also just talk to the agent,
196
00:14:48.950 --> 00:14:53.990
just type in the box. And usually it does a pretty good job at whatever I wanted to do.
197
00:14:54.550 --> 00:14:57.835
All of these things, all of these marketing files, skills,
198
00:14:57.835 --> 00:14:58.635
instructions,
199
00:14:59.035 --> 00:15:00.075
superpowers,
200
00:15:00.075 --> 00:15:00.875
plugins,
201
00:15:01.275 --> 00:15:01.995
specs,
202
00:15:02.235 --> 00:15:06.475
is it really needed? Especially when you use the Frontier models, MPT55,
203
00:15:06.475 --> 00:15:07.755
OPUS 4.7.
204
00:15:08.155 --> 00:15:11.035
What do you think? Yeah, it's a good question. I think
205
00:15:12.190 --> 00:15:20.350
I see models getting better. I see the code. So the software, the agent itself. So the like GitHub Copilot or Cloth Code,
206
00:15:21.390 --> 00:15:23.150
software around the LM. Yeah,
207
00:15:23.950 --> 00:15:26.430
the harness gets better and better, to be honest. So
208
00:15:27.375 --> 00:15:33.455
I see that the maybe the need of this will in the future will be much, much less. I don't know.
209
00:15:34.175 --> 00:15:34.655
But
210
00:15:36.015 --> 00:15:37.455
when I I
211
00:15:37.455 --> 00:15:46.110
want the agent to do it in a certain way that I would prefer it. Right? And that's the whole thing. So that's why I use Custom Agent. That's the reason
212
00:15:46.589 --> 00:15:52.270
why I use some of the skills is that, for example, when I ask it to create this,
213
00:15:52.589 --> 00:15:54.589
I don't know, a real time voice agent,
214
00:15:55.285 --> 00:16:05.205
I would like to use the Foundry tools. I know the stack, I know the architecture that I want basically. And that's something that you want to enforce. And basically I think with
215
00:16:05.765 --> 00:16:13.400
customers this is the same thing is that they already have a tech stack that they're using and they want to be they want to go
216
00:16:13.400 --> 00:16:17.640
into the same stack as well. So that's why I think it still makes sense to
217
00:16:18.120 --> 00:16:23.575
use custom instructions and skills and custom agents, etc. Yeah.
218
00:16:25.334 --> 00:16:25.735
I
219
00:16:26.135 --> 00:16:26.774
agree.
220
00:16:27.014 --> 00:16:37.400
Especially for now, and we're getting memory. GitHub Copilot has memory that can also live in the cloud later on. I think it will learn from what a customer says
221
00:16:37.960 --> 00:16:41.000
and what he does. So yeah, let's see.
222
00:16:41.800 --> 00:16:43.480
Often find myself thinking,
223
00:16:43.960 --> 00:16:50.014
how hard is this? We're just typing in this box and I'm doing a whole course on it the whole day. How
224
00:16:50.014 --> 00:16:52.255
you type better into the box, basically.
225
00:16:52.654 --> 00:16:53.615
Very strange.
226
00:16:54.175 --> 00:17:02.095
Yeah, no, I know. But I get that, right? So to be honest, I'm doing the same thing. So I'm explaining to customers like, okay, how can
227
00:17:03.720 --> 00:17:06.440
they use it in a better way? And the thing is that
228
00:17:07.240 --> 00:17:11.480
when we look at a year or two years ago when we used GitHub Copilot
229
00:17:11.800 --> 00:17:13.640
and we did a one shot,
230
00:17:15.560 --> 00:17:17.400
the answer or the output was
231
00:17:17.785 --> 00:17:22.105
way off of what the things that we wanted to create. I agree that
232
00:17:22.585 --> 00:17:35.380
this is why we came up with things like skills and MCP servers and tools to make it more sophisticated on a certain part. But the LLM models also evolved at that time and pace,
233
00:17:35.860 --> 00:17:38.660
also very quickly. So I agree that
234
00:17:38.900 --> 00:17:41.700
there should be like a healthy balance in it.
235
00:17:46.765 --> 00:17:52.284
Oh, you dropped there for a second for me. Oh, I can still hear you. You're still here for me.
236
00:17:52.845 --> 00:17:58.445
Oh, that's good. That's good. Yeah. So, yeah. Let's see. You know, this stuff keeps evolving.
237
00:17:59.240 --> 00:18:04.600
I'm sure we're super close to another batch of new models, who knows? It's been a month or so.
238
00:18:05.480 --> 00:18:06.040
Yeah.
239
00:18:06.840 --> 00:18:07.640
We'll see.
240
00:18:08.040 --> 00:18:11.559
Yeah, so I think this is going so quickly, it's really
241
00:18:12.520 --> 00:18:13.880
mind blowing to be honest.
242
00:18:16.055 --> 00:18:24.375
So yeah, and it's always interesting to hear about like how, especially when you talk to customers, like how they perceive this, right? They
243
00:18:25.495 --> 00:18:30.090
see also of course, they see like Microsoft investing a lot in AI,
244
00:18:30.090 --> 00:18:31.210
of course. And they
245
00:18:31.930 --> 00:18:39.130
say like, okay, but you're saying it is so great because you need to get that return on investment, right? So you need to sell it as well.
246
00:18:40.015 --> 00:18:41.615
Of course, that's
247
00:18:41.615 --> 00:18:43.775
also part, but it's
248
00:18:43.775 --> 00:18:47.134
not the one that drives me the most. It's the thing that
249
00:18:47.934 --> 00:18:53.294
it really disrupts. So what it changes in my day to day work that really excites
250
00:18:53.294 --> 00:18:53.455
me.
251
00:18:54.720 --> 00:18:55.360
And
252
00:18:55.680 --> 00:18:58.080
yeah, this is something where I see like,
253
00:18:58.480 --> 00:19:07.280
yeah, we need to bridge that gap a little bit and create that perspective that it's really becoming, especially on the development side, it's getting
254
00:19:07.765 --> 00:19:09.284
that good at this
255
00:19:09.525 --> 00:19:10.164
moment.
256
00:19:10.645 --> 00:19:13.044
Do I see it taking over our jobs?
257
00:19:13.285 --> 00:19:20.565
No, because I still think that we are the ones driving the car, right? We are the ones driving the we are steering.
258
00:19:21.045 --> 00:19:25.370
And I think that will be eventually that will be still the case.
259
00:19:25.930 --> 00:19:32.250
Only what I see now is that I I am much more reviewing my code. I'm much more
260
00:19:34.090 --> 00:19:34.570
sometimes
261
00:19:35.125 --> 00:19:38.085
deliberately testing a lot more myself
262
00:19:38.085 --> 00:19:39.845
because I sometimes
263
00:19:39.924 --> 00:19:47.205
do not the fact that I when I can do like I use Playwright, MCP Server a lot to do the testing for me
264
00:19:48.085 --> 00:19:49.445
and it does a really good job.
265
00:19:50.059 --> 00:19:51.019
But still,
266
00:19:51.580 --> 00:19:53.100
when I tested myself,
267
00:19:53.340 --> 00:19:54.620
there are also always
268
00:19:54.620 --> 00:20:08.345
things that it didn't cover or still some things that it didn't get right. So this is, I think, still something we need to be mindful of, that we are almost there, but it it isn't a 100% proof. So we still need to be,
269
00:20:08.745 --> 00:20:18.059
yeah, still driving need to be driving the car and and validate if it's good or not before we gonna release it or before we gonna put it into production.
270
00:20:19.020 --> 00:20:20.779
Yeah. Now. Two liter.
271
00:20:21.900 --> 00:20:24.139
Right. For that that's yes.
272
00:20:24.460 --> 00:20:27.580
That that we always end our sentences with for now.
273
00:20:28.059 --> 00:20:30.700
For now. Yeah. Because you don't know what the future will all do.
274
00:20:32.395 --> 00:20:32.955
So
275
00:20:33.275 --> 00:20:39.914
GitHub Copilot, right? Copilot three sixty five, that's all stuff that is happening on your machine for your personal productivity,
276
00:20:39.915 --> 00:20:40.475
basically.
277
00:20:40.955 --> 00:20:41.914
Yeah. Right.
278
00:20:42.315 --> 00:20:44.875
And you also said something about Foundry.
279
00:20:45.060 --> 00:20:47.940
You also talk about Foundry with customers.
280
00:20:48.340 --> 00:20:52.660
Maybe can you explain what Foundry is? How is it different? When would you use that?
281
00:20:53.140 --> 00:20:56.820
Yeah, so Foundry is basically our
282
00:20:58.465 --> 00:21:04.864
serverless service within Azure that we provide to customers to spin up and provision their own models.
283
00:21:05.825 --> 00:21:07.985
So, basically these are the LLMs
284
00:21:07.985 -->
00:21:15.440
that, for example, a coding agent uses. So, if you look at my GitHub Copilot CLI agent, it uses models.
285
00:21:15.600 --> 00:21:21.200
You can get those models from GitHub itself, but you can also get those models from Foundry.
286
00:21:22.960 --> 00:21:25.120
And the thing is that within Foundry,
287
00:21:25.825 --> 00:21:26.304
we
288
00:21:26.865 --> 00:21:27.424
have
289
00:21:28.065 --> 00:21:30.065
a system where we can provision
290
00:21:30.065 --> 00:21:33.665
any kind of model from any kind of third party provider.
291
00:21:33.825 --> 00:21:40.259
So we have Gemini in there, we have Cloud, we have OpenAI models, we have any kind of model that
292
00:21:40.500 --> 00:21:45.220
is out there, we have it in Foundry. So these are like nowadays
293
00:21:45.220 --> 00:21:47.140
there are maybe 15,000
294
00:21:47.140 --> 00:21:50.179
models already in the Foundry portal, so it's a lot, right?
295
00:21:51.635 --> 00:21:57.154
But sometimes really dedicated for some of the use cases, which I think is really important because
296
00:21:57.875 --> 00:22:07.600
every use case could use a different model for when you're going to build something. So for example, I don't need a frontier model for everything. I could use,
297
00:22:08.160 --> 00:22:21.195
for example, mini models or even nano models to use for testing or for doing like really basic stuff or summary or whatever. So this is something where you can think about like, okay,
298
00:22:22.235 --> 00:22:24.874
these are the models that you can provision in Foundry.
299
00:22:25.515 --> 00:22:31.835
And where would you use Foundry for? It's to create or basically make your own current applications intelligent,
300
00:22:32.280 --> 00:22:38.520
more intelligent than they are right now. So if you want to create your own agent or if you want to create your own,
301
00:22:38.840 --> 00:22:39.720
I don't know,
302
00:22:41.080 --> 00:22:42.440
agentic solution,
303
00:22:43.320 --> 00:22:47.960
then you can use Foundry to basically use as a base for your model provider
304
00:22:48.355 --> 00:22:49.474
and also the
305
00:22:49.715 --> 00:22:56.114
agent service even if you want to host your agents in Foundry as well. And then you can create there your agentic
306
00:22:56.835 --> 00:22:58.754
multi agent solutions
307
00:22:59.475 --> 00:23:07.609
using Foundry and using maybe your own code base or using it with your own application or embedding it into your own application
308
00:23:07.850 --> 00:23:08.729
or whenever
309
00:23:09.610 --> 00:23:10.409
is needed.
310
00:23:10.889 --> 00:23:16.409
And we saw this with the for example, the like a few years ago when you saw like JetGPT coming up,
311
00:23:17.345 --> 00:23:19.985
people really love that they can just talk to
312
00:23:20.225 --> 00:23:22.145
an agent that really
313
00:23:22.545 --> 00:23:25.265
gives you that, I would say, that more personal vibe.
314
00:23:26.225 --> 00:23:28.465
And this is something that you can create yourself.
315
00:23:28.785 --> 00:23:32.970
And this is where we use Foundry for, to basically provide the model
316
00:23:33.130 --> 00:23:34.090
interaction,
317
00:23:34.650 --> 00:23:36.330
so you can create your own
318
00:23:36.730 --> 00:23:38.170
chat application.
319
00:23:38.650 --> 00:23:44.970
Maybe a chat application that talks to your data, or a chat application that uses your application's information,
320
00:23:45.050 --> 00:23:51.235
or does things for you on your behalf in that application. So these are the scenarios that we see
321
00:23:52.035 --> 00:23:54.674
other customers creating with Foundry.
322
00:23:55.715 --> 00:23:59.235
Yeah. So a lot more control models,
323
00:23:59.659 --> 00:24:02.059
running agents, guilty agents,
324
00:24:02.779 --> 00:24:06.940
including data as well, right? Rack and stuff like that. Evaluation
325
00:24:06.940 --> 00:24:10.619
as well, like that you can see the performance of your models,
326
00:24:10.940 --> 00:24:17.405
that you can tune it, that you can There are all kinds of mechanisms in there that can make it even better.
327
00:24:18.045 --> 00:24:23.645
And also that you can tune the model in a certain way. So you can tune your application
328
00:24:23.965 --> 00:24:27.400
on how it uses the model. Yeah. Yeah.
329
00:24:27.880 --> 00:24:36.360
And so I hear all these things, right? I always try to see what to use when. So you also like GitHub Copilot SDK.
330
00:24:37.000 --> 00:24:38.040
Where does that fit in?
331
00:24:38.865 --> 00:24:45.664
Basically GitHub Copilot SDK is the way on how to create your own, I would say your own coding agent.
332
00:24:46.065 --> 00:24:55.539
And we talked about that harness, right? So that basically the GitHub Copilot SDK already has that harness in it. So that's basically the harness of the agent
333
00:24:57.140 --> 00:24:59.299
which has capabilities like
334
00:25:00.980 --> 00:25:07.995
shared memory or using sub agents and all these kinds of mechanisms that are already part of the agent,
335
00:25:08.155 --> 00:25:10.315
that you can build your own agent with it.
336
00:25:10.955 --> 00:25:13.835
And basically get a model from Foundry,
337
00:25:13.835 --> 00:25:17.035
like you can get Cloud Opus 4.7
338
00:25:17.195 --> 00:25:23.960
from Foundry and then embed it into your GitHub Copilot SDK to create your own development agent.
339
00:25:24.440 --> 00:25:25.479
So you're
340
00:25:25.799 --> 00:25:29.479
basically free then to create it yourself, even if you want.
341
00:25:30.280 --> 00:25:30.999
Yeah.
342
00:25:31.320 --> 00:25:32.519
Okay.
343
00:25:32.679 --> 00:25:41.984
That's very cool. So in the future, we will have lots and lots of different agents running in an enterprise. GitHub Copilot powered ones.
344
00:25:42.544 --> 00:25:43.904
Microsoft's agent
345
00:25:44.304 --> 00:25:48.544
framework powered ones, other ones, third party ones, all sorts of stuff that runs around.
346
00:25:49.260 --> 00:25:52.460
How do
we keep track of all that stuff and make sure that nothing
347
00:25:53.020 --> 00:25:56.620
strange happens? Like, you know, somebody deletes a database or something.
348
00:25:58.380 --> 00:26:01.340
Yeah. Think so. I think this is where I think
349
00:26:02.654 --> 00:26:07.774
where we see that Foundry comes in as well, right? So you can build that security layer on.
350
00:26:07.934 --> 00:26:11.695
So if you use Foundry as your, I would say your centralized
351
00:26:12.495 --> 00:26:13.854
model provider,
352
00:26:13.855 --> 00:26:18.559
you can build security measures in on top of there to already
353
00:26:18.960 --> 00:26:22.960
make sure that nothing really critical is happening as well.
354
00:26:24.880 --> 00:26:27.120
To be honest, I see different approaches here.
355
00:26:27.520 --> 00:26:36.665
One of the interesting part is that I see, for example, when it comes to development, a lot of companies are also running it in sandbox environments. So basically, isolated environment,
356
00:26:36.665 --> 00:26:37.065
just
357
00:26:37.544 --> 00:26:49.390
using a coding agent to do a certain task, but having all the capabilities that they need from that sandbox that they can use. Because when you're giving an AI agent on your machine a lot of access,
358
00:26:50.590 --> 00:26:55.230
it can do much, much more. And we saw that, for example, with OpenClaw,
359
00:26:55.310 --> 00:27:04.705
is that it can do proactively also a lot of things for your machine as well. And you maybe don't want that, right? So this is where you want to a little bit create
360
00:27:04.705 --> 00:27:16.690
like your own corporate harness around the agents or basically your your yeah, I would say maybe your your agents, but also the models that you use to basically secure them a little bit off.
361
00:27:19.010 --> 00:27:29.075
I think, yeah, so and I think using like something like like like Foundry where you can monitor on what it does or what it's capable to do, limiting
362
00:27:29.155 --> 00:27:30.115
it with,
363
00:27:30.435 --> 00:27:31.395
for example,
364
00:27:31.715 --> 00:27:33.875
MCP access, so you can
365
00:27:34.435 --> 00:27:37.235
limit the access of or basically
366
00:27:37.475 --> 00:27:41.875
prescript the things that an agent can do to a really minimal extent.
367
00:27:42.750 --> 00:27:45.070
So it isn't able to do everything,
368
00:27:45.310 --> 00:27:47.309
but it can just do enough
369
00:27:47.470 --> 00:27:48.990
of what it needs to do.
370
00:27:50.670 --> 00:27:53.070
So yeah, I think that this is
371
00:27:53.390 --> 00:27:53.950
the way
372
00:27:55.054 --> 00:27:56.654
we mostly do it,
373
00:27:57.054 --> 00:27:57.774
I would say,
374
00:27:58.655 --> 00:28:01.374
on an implementation side. I
375
00:28:02.174 --> 00:28:04.735
don't know if it really answers your question, but
376
00:28:06.655 --> 00:28:07.294
I know. What
377
00:28:08.174 --> 00:28:10.174
are you specifically looking for? I don't know.
378
00:28:10.909 -->
00:28:12.669
Yeah, we it. No, it's
379
00:28:14.669 --> 00:28:15.870
I'm not talking about like
380
00:28:16.909 --> 00:28:17.549
Yeah?
381
00:28:18.909 --> 00:28:20.109
I'm just wondering what
382
00:28:20.350 --> 00:28:28.405
this will look like, right? Because this is a whole new world. We never had this. All these new things that run on their own that have incredible capabilities.
383
00:28:29.045 --> 00:28:32.725
So how can we control that type of stuff without huge
384
00:28:33.125 --> 00:28:47.419
security incidents or stuff like that? So Yeah. I I think that it's it's it's just it for me, it's not determined yet. I see that from Microsoft perspective, we have Agents three sixty five. Right? Like this unified console plane where we can
385
00:28:48.139 --> 00:28:54.139
add all our agents and have a have an overview of what each agent is doing. So you can embed your
386
00:28:54.485 --> 00:29:03.365
your foundry agents in there, you can embed your your your local agents in there that you can see what they are doing and how they are progressing, and you can monitor to that.
387
00:29:03.685 --> 00:29:05.845
But I'm I'm really curious to see
388
00:29:06.420 --> 00:29:14.900
if that will be the solution for the for the end one. And I I don't know. I'm a little bit I'm I'm a little bit unsure there what's what the best approach is, to be honest.
389
00:29:16.420 --> 00:29:21.625
Well, thing is this is new for everybody. So I think we're finding it out in real time really.
390
00:29:22.825 --> 00:29:23.465
There
391
00:29:23.705 --> 00:29:27.305
are some things that can happen that we do offer as Microsoft.
392
00:29:30.265 --> 00:29:31.065
We
393
00:29:31.065 --> 00:29:31.705
can
394
00:29:32.370 --> 00:29:34.690
kind of end on this note. So
395
00:29:35.570 --> 00:29:37.810
we're now in May
396
00:29:37.810 --> 00:29:39.010
2026,
397
00:29:40.050 --> 00:29:47.644
and around this time ish, when this podcast comes out, GitHub Copilot will have a new billing mechanism,
398
00:29:47.804 --> 00:29:54.044
which means that we basically pay by token. OWL's calls slightly different. How can people
399
00:29:54.524 --> 00:29:56.205
use
400
00:29:56.205 --> 00:29:57.085
their tokens
401
00:29:57.270 --> 00:30:01.990
efficiently, effectively? Are there any tips or tricks to reduce your token count,
402
00:30:02.070 --> 00:30:03.270
to use more What
403
00:30:04.390 --> 00:30:05.350
can we do?
404
00:30:05.670 --> 00:30:08.230
Good question, good question. So I think
405
00:30:12.475 --> 00:30:14.154
it comes to token management,
406
00:30:14.555 --> 00:30:15.195
I think
407
00:30:15.915 --> 00:30:18.154
it's very important that you try
408
00:30:19.515 --> 00:30:20.715
to limit the
409
00:30:21.675 --> 00:30:26.650
number of tokens that you spend, right? So you keep the context minimal as possible.
410
00:30:26.890 --> 00:30:34.330
Also using the appropriate model for the appropriate task. I think that's also already a way to limit your token usage.
411
00:30:35.050 --> 00:30:36.809
Personally, I really like
412
00:30:36.970 --> 00:30:38.809
the AI fluency
413
00:30:38.890 --> 00:30:40.490
plugin from RobBoss.
414
00:30:41.605 --> 00:30:51.524
This is something I use on a daily basis. I already implemented it, for example, in my comments line that I always see the number of tokens I used for the last thirty days.
415
00:30:52.325 --> 00:30:54.885
This already gives me some insights
416
00:30:55.690 --> 00:30:58.730
on the spends that I have when it comes to
417
00:31:00.010 --> 00:31:01.610
tokens that I use.
418
00:31:02.410 --> 00:31:03.049
So,
419
00:31:04.010 --> 00:31:10.010
I think from one point, I think it's having insights on what you're using and what you're spending basically.
420
00:31:10.525 --> 00:31:16.205
But also on the other hand, I think it's very important to look at what's actually going to the LLM.
421
00:31:16.525 --> 00:31:20.044
So if you, for example, in Visual Studio Code you can use,
422
00:31:20.285 --> 00:31:22.124
I think it's called chat debug,
423
00:31:22.285 --> 00:31:23.245
you can just
424
00:31:23.485 --> 00:31:24.045
you can
425
00:31:24.640 --> 00:31:32.559
collect, so when you go to your common palette, do chat debug, you can can sketch out what is actually being sent to
426
00:31:33.120 --> 00:31:34.239
the LLM.
427
00:31:34.880 --> 00:31:37.919
And here what you sometimes see, and especially I do this
428
00:31:38.105 --> 00:31:48.985
when I'm asking a question and the output is pretty much not what I requested. Right? So when it's really it's when it's totally off, but I think you can use it in many, many ways.
429
00:31:50.280 --> 00:31:57.960
But then I would just go into the trace of what is actually being sent to the model. And then you see that, for example, there's a lot of maybe
430
00:31:58.520 --> 00:31:59.320
MCP
431
00:31:59.320 --> 00:32:01.160
baggage that's
432
00:32:01.160 --> 00:32:08.414
not really interested for your question, but it still goes to the model, right? And you need to be conscious of what you're sending there.
433
00:32:09.455 --> 00:32:13.215
And you can also there limit a lot of the context
434
00:32:13.295 --> 00:32:15.375
in a better way, the
435
00:32:15.775 --> 00:32:22.029
quality of the answer would be much better from the LLM. So for me, it's having insight in your tokens,
436
00:32:22.429 --> 00:32:25.950
having insight in what's actually being sent to the LLM.
437
00:32:27.470 --> 00:32:29.869
And I think from a loss perspective is the
438
00:32:30.510 --> 00:32:35.165
model. Think about which model to use for which kinds of purposes. Because
439
00:32:35.485 --> 00:32:41.885
Haiku, for example, is really good in when it's just want to summarize or doing a little bit of minimal brainstory.
440
00:32:41.885 --> 00:32:46.765
You don't need then a Claude Opus model to do that. So be conscious about
441
00:32:47.419 --> 00:32:53.580
what you're using in in what yeah, in which case. And also the AI fluency
442
00:32:54.059 --> 00:32:56.700
plugin, I think it's really nice because it gives you
443
00:32:57.100 --> 00:32:57.899
also
444
00:32:58.220 --> 00:33:02.779
a little bit more context on how you use GitHub Copilot, not only from a token
445
00:33:03.715 --> 00:33:05.874
usage, but also gives you stats
446
00:33:06.355 --> 00:33:10.914
in how you use it and how even how optimized your
447
00:33:12.355 --> 00:33:24.769
requests are in combination with your context window. So it already gives you some insights on, hey, am I an AI expert or not? And I think this is a nice
448
00:33:25.730 --> 00:33:26.369
tip.
449
00:33:27.409 --> 00:33:28.129
Yeah.
450
00:33:28.289 --> 00:33:29.010
Okay.
451
00:33:29.295 --> 00:33:38.655
So those are good tips to use, you know, start exploring, make sure that you optimize your tokens so that you don't run out at the end of the month, right? Because that would be super spent.
452
00:33:38.815 --> 00:33:39.215
Yeah.
453
00:33:40.175 --> 00:33:43.919
You can't work anymore to close your laptop. Well,
454
00:33:43.919 --> 00:33:49.759
you can always request more tokens, right? So it's bumping up the limit, right? That's
455
00:33:51.040 --> 00:33:52.480
I'm coming from Microsoft, Marie.
456
00:33:54.875 --> 00:34:00.155
Well, no, I know. So we are we were blessed with currently unlimited tokens, which is lovely. Yeah.
457
00:34:00.795 --> 00:34:09.710
That's that's that's a true blessing. I would I said to I said internally, I said this is even better than the Xbox Ultimate Boss, James Boss.
458
00:34:11.230 --> 00:34:13.230
Use it much, much more than than
459
00:34:13.550 --> 00:34:15.790
I use my Xbox right now. So, yeah.
460
00:34:17.230 --> 00:34:17.870
Yeah.
461
00:34:18.430 --> 00:34:24.415
Alright. So on that note, let's let's end it. Thank you very much for this. You're welcome.
462
00:34:25.375 --> 00:34:26.175
Let's
463
00:34:26.175 --> 00:34:27.855
maybe reconnect in
464
00:34:28.015 --> 00:34:30.975
six months, a year. Let's see what the world looks like then.
465
00:34:31.375 --> 00:34:32.255
Good idea.
466
00:34:32.415 --> 00:34:35.535
Good idea. See if we if we got it right on some points, right?
467
00:34:36.369 --> 00:34:44.050
Yeah. And and and after you have cuddled by chicken, so then you know if your heart if your heart rate is is lower or not.
468
00:34:44.530 --> 00:34:45.010
Yeah.
469
00:34:45.730 --> 00:34:46.930
Alright. We'll find it.
470
00:34:47.803 --> 00:34:50.283
You very much. You're welcome. See you.