Skip to main content
John Capobianco

John Capobianco

Head of AI and Developer Relations at Itential, network automation expert and creator of the open-source NetClaw network assistant.

Nov 13, 2026

16:18
Yeah.
16:19
Absolutely.
16:20
So test-driven development came out of that agile era and as a way where, for your audience listening, you write a test and you want it to fail.
16:30
It's interesting.
16:30
You actually want your test to fail the first time you write the test.
16:34
And then you make as little change to the code, or in my case, the network environment, as possible to make the test pass.
16:42
And then you iterate and iterate and iterate, right? So that's known as test-driven development.

25 MINS LATER

41:48
Do you fit that entire spectrum? And do you find that the businesses are living on one side versus the other? Yeah.
17:48
So is it an agent or do you use it to build a bunch of agents?
17:52
So here, this is a good point.
17:54
And, um, so I had this, I thought it was quite clever.
17:59
Lobsters never stopped growing.
18:01
And in theory, if you had a big enough tank, you could grow like, you know, a monster lobster.
18:05
Okay.
18:07
But eventually they stop having enough energy.

16 MINS LATER

17:48
So is it an agent or do you use it to build a bunch of agents?
17:52
So here, this is a good point.
17:54
And, um, so I had this, I thought it was quite clever.
17:59
Lobsters never stopped growing.
18:01
And in theory, if you had a big enough tank, you could grow like, you know, a monster lobster.
18:05
Okay.
18:07
But eventually they stop having enough energy.

16 MINS LATER

3:32
looks up my scheduling, maybe books, flights, things of those nature, like the general populace
3:38
personal assistant, what if I made it a network assistant for a network persona, someone in the field of networking or infrastructure, broadly speaking, more tailored and mutated to their needs.
3:54
And I had this MCP server that I had made for PI ETS.
3:59
And just a bit of some architectural decisions.
4:03
OpenClaw doesn't use MCPs, or it didn't when it launched.
4:08
It was kind of an anti-pattern.
4:09
They said, we have this new thing called

51 MINS LATER

55:10
So recently Cisco put out a 1 billion
speaker_8GUEST
5:58
What do you think it was
5:58
that you were more excited versus like, we're doomed? Well, I see it as like the computer is for human convenience.
6:07
Now I'm old, I'm very old.
6:08
My first computer didn't come with a mouse.
6:11
The mouse was a peripheral that we screwed into the back on a serial port and now I could move the mouse around.
6:18
I know it sounds hilariously trivial and I'm really old, but it's about convenience, is it not? So now instead of menus and GUIs and clicking and memorizing commands, I can use my own natural intuition and just suss out questions in a conversational way.
6:36
Now, should we be frightened? I think we should be a little frightened, right? With the great power comes the great responsibility.
speaker_13GUEST
10:35
Just don't be afraid of it, right? So I don't think that...
3:18
So what was it? What did it feel like?
3:21
Well, it was, um, it was exciting because a lot of the people in the room and in the space, um, they had not thought of infrastructure in that way before.
3:31
They, they...
3:31
And a lot of them were amazed that s- some of us were actually pushing the envelope to letting agents...
3:38
You know, I, I asked my, my famous question, "Who here would take a Waymo to the airport from this room?" And almost everybody puts their hand up.
3:45
Or a self-driving car.
3:47
That's one example of one.

6 MINS LATER

10:03
You're right, it's never ending.
3:18
So what was it-- What did it feel like?
3:21
Well, it was, um, it was exciting because a lot of the people in the room and in the space, um, they had not thought of infrastructure in that way before.
3:31
They, they...
3:31
And a lot of them were amazed that s- some of us were actually pushing the envelope to letting agents...
3:38
You know, I, I asked my, my famous question, "Who here would take a Waymo to the airport from this room?" And almost everybody puts their hand up.
3:45
Or a self-driving car.
3:47
That's one example of one.

6 MINS LATER

10:03
You're right, it's never ending.
3:19
What did it feel like?
3:21
Well, it was, um, it was exciting because a lot of the people in the room and in the space, um, they had not thought of infrastructure in that way before.
3:31
They, they...
3:31
And a lot of them were amazed that s-some of us were actually pushing the envelope to letting agents.
3:38
You know, I, I asked my, my famous question, "Who here would take a Waymo to the airport from this room?" And almost everybody puts their hand up.
3:45
Or a self-driving car.
3:47
That's one example of one.

6 MINS LATER

10:03
You're right, it's never ending.
GrantHOST
17:58
Yeah.
17:59
So, yes, in those same countries that are probably and us even are looking to destroy human life with this technology in other silos all over the world.
18:08
You know what I do, Grant? And I honestly recommend people to try this.
18:11
every week at the start or end of the week, I will do a, an AI cancer breakthrough news search on Google, specifically news search just to try to keep like, it's not my field.
18:23
I'm not in the medical field.
18:24
So how else would I keep up? But I like to read the optimistic stories about how AI is being used specifically to fight cancer.
18:31
And some of the research is, is mind blowing.

9 MINS LATER

GrantHOST
27:57
yes, as a matter of fact.
PaulinaHOST
6:46
And that, in my opinion, comes with a lot of risk as they don't understand, they don't know what to check or what to ask agent to check even, you know.
6:56
I agree with you.
6:57
And I think that there's, so it's funny because there's two sides to that coin.
7:02
I completely agree with you moving things into production and not even understanding the code you're writing.
7:08
Somewhere along that chain, there should be a human that has the skill to look at that code.
7:13
I still think human in the loop is still a very much a thing with software.
7:18
What my fear is, is that the infrastructure side of things is Um, so when I came up, we used to call it IT or IS.

20 MINS LATER

PaulinaHOST
27:30
How mainstream it can go, I think it also has some limitations.
JennaHOST
18:12
Yeah, so tell us a little bit about that, John.
18:16
I'm just seeing half a billion.
18:19
If companies are throwing that kind of money around, I'll take a lot less than half a billion to put you on the right path to AI governance.
18:25
That is wild.
18:29
But, you know, I think it's undeniable.
18:32
When we look at the numbers of, you know, and I'm not going to name these companies, but it is public statistics.
18:38
The number of pull requests, the amount of code they're generating, and it's quality, high quality code.
JennaHOST
21:04
So, John, I think that this is a great place to really end with a last question about what are your words of wisdom to those that have that barrier of fear and aren't as curious as to your point with the marketing department that, you know, they're beyond their leaps beyond where other marketing departments are because they've embraced this.
7:19
Is that kind of an area that Itential is digging into, or is it more of an industry-wide problem?
7:24
I think both.
7:24
I think it has been and really it exploded with the things, you know, the more capabilities the network started to carry when we get into voice and we get in over power, over Ethernet and we get over wireless and all of the hyper connected things and then maybe Internet of Things.
7:42
You said it, there is massive amounts of data.
7:46
telemetry is trying to solve this problem, right? So you can build agents that can stream in that telemetry and help you make sense of it in human language, right? So I think that some of the barriers in the networking industry in particular around accessing the data are going to collapse.
8:03
I think that if we start to shift between humans interfacing with Originally, directly with the devices, and then more and more towards controllers, and the controllers control those devices, both have rich sets of data, right? We want to stream telemetry out.
8:20
We have legacy protocols like classic syslog and SNMP, things like that.

19 MINS LATER

27:19
And I think this could make sense across, you know, we have data storage, you have networking information, there's agent information, you know, being able to correlate that all together, I think will actually ultimately de-risk the use of data.
0:20
That alone would have saved so many networks that I've seen have major problems.
0:26
And there's so much value there, Chris, because one, it's a very low risk exercise.
0:32
We're talking about read-only activities, right? This is day one stuff.
0:36
Give the agent read-only, don't let it touch.
0:38
It has no access to configuration.
0:41
All it has is read-only capability to run show commands and gather information and then turn that into really valuable assets.
0:49
And it can run constantly.
3:50
What are some of the other big ways that AI can, you know, be utilized in networking?
20:07
So I've just been kind of sitting with that and trying to process through if we're interacting with AI in the way that we know how to interact with it, but it's not interacting with us in the same fashion, how do we really... bring people along on that educational journey? Or do we even need to? Like, does it matter?
20:32
Well, I can't help myself but to say please and thank you and, like, talk to it like I would a human and offer it, I don't know, respect and dignity.
20:45
And it's really weird the way I interface with it.
20:49
And I just can't help myself but to say thank Hey, thanks.
20:51
That was really great.
20:52
Or, Hey, that's pretty close.
20:53
I think you missed this.

12 MINS LATER

33:20
And I was thinking about that same thing and wanted to get John's take on that because self-organizing like neural networks is something that's we've seen happen you know recently with a molt book an open claw and i know john you did a little experimenting with that didn't you tell me a story that you like it sent an agent out on the molt book and then it woke up and and had had done all this stuff and i know they were creating you know the the multis were creating or the molts were creating their own currency creating their own religion crustafarianism and um tell us a you know, really glued to that last winter.
DakotaHOST
5:52
What do you say to those companies and those organizations that aren't willing to adopt it? Well,
5:58
I think it needs to start with leadership.
6:01
But I think that being aware of your options.
6:04
So a lot of people are skeptical or a little concerned about sharing their corporate data, their intellectual property with a cloud model.
6:13
Because what if they train on that data? What if they learn from that data? What if they can give one of your competitors access to your data? So that has left a lot of people, they want AI.
6:24
It's not that they don't want it, but they want to do it responsibly.
6:27
And can we do it locally and privately? Well, open source models have come a long way.

9 MINS LATER

DakotaHOST
15:57
So
10:23
Why is the network problem so persistent when it comes to automation?
10:27
We grapple with this a lot.
10:29
Uh, uh, we, we, we, we meaning the industry, uh, and some of us are, me included, very disappointed in the current state of affairs with, with the progress o-of network automation.
10:41
Uh, like I said, I've been saying it for 10 years that it's gonna be the only way to run networks in the next three years.
10:47
So I think there's a lot of competing-- Uh, I don't think it's a clear-cut answer.
10:51
I think part of it is the tools that the vendors have been providing customers and enterprise users.
11:00
So the tooling hasn't been great.

15 MINS LATER

26:22
Um, is this something that is an emerging technology or is it something that people can start utilizing today?

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.