John Capobianco, Selector.AI
Welcome to the first episode of Jason's Industry Insights!
Be kind. It's a learning curve, especially with sound quality and editing.
In this episode, I speak with John Capobianco from Selector.ai, exploring the innovative applications of AI in network management. We discuss how Selector.AI enhances root cause analysis, predictive capabilities, and the integration of AI into existing systems.
John shares insights on real-world applications, customer success stories, and the future of autonomous networks, emphasizing the importance of data management and the measurable ROI of AI solutions.
00:00 Introduction to Selector.ai and Its Mission
04:39 The Role of AI in Network Management
09:53 Expanding Beyond Service Providers
15:02 Streamlining Root Cause Analysis
19:46 The Future of Autonomous Networks
25:33 Engaging with the Community and Learning from Users
Transcript
Hello and welcome to Jason's industry insights, the podcast. Each week I'll spend time digging deeper into one of the stories in the newsletter that I think is kind of interesting to find out a bit more. This week, I wanted to spend time getting to know more about a company named selector.ai, network observability. That's what they're all about. But I'll let my guests really give you the real insights and talk more about it. With me today is John Cappabianco. John, what's your title? I should have had this before from you, but you know, in,
John Capobianco (:No, so thank
you, Jason, for having me here. I think this is a wonderful endeavor. The more voices out there in the community, the better. And it gives another whole new perspective on the industry. So I'm actually a product marketing evangelist. I joined Selecter in August, early August. I was formerly with Cisco as a technical leader in artificial intelligence. And prior to Cisco, I was a senior network architect for the Parliament of Canada, the House of Commons. I, someone...
Jason Presement (:Absolutely.
John Capobianco (:Prominent in the industry, Jeremy Schumann sent me a video of Selectors work about a year ago because he knew that I was pursuing artificial intelligence and building tools and exploring AI from a practitioner's point of view. And I saw this video and it was really what I was trying to achieve but at a much larger scale, at an actual commercial enterprise scale for large service providers.
a natural language interface into your infrastructure, a network language model where we can fine tune an open source model with network specific data and SQL to English language data. It's like you said, observability, but it really is an AI ops tool. So we're going to collect all kinds of heterogeneous data and use machine learning to do natural baselines and log mining.
for named entity recognition, right? And we're going to use all of this massive amount of data in the data lake and cluster, you know, cluster things based on attributes and draw correlations and root cause analysis.
Jason Presement (:So let's get
into that in a sec, but first I have to ask you, what kind of dog do you have? that's a golden retriever. It sounds like something that's a little tiny dog. I have a golden retriever. It doesn't sound like that.
John Capobianco (:So that's the Golden Retriever. Like I said, that's a Golden Retriever. She's
very unique in that regard, yes. I hope she's not too audible.
Jason Presement (:Everyone has dogs, cats, you know, it's fine anyway.
John Capobianco (:Well, it's like I
was saying earlier, we had a big snowstorm and my wife's literally my wife is out snow blowing right now. So the dog is a little distracted by her being outside the window. that's enough.
Jason Presement (:You'll have to have a
separate podcast on how you got your wife to snow blow because it's separate conversation. so, SelectRAI, so I mean, I, know, the years I've been in networking and you've been in networking and we met each other earlier in the week or last week when we chatted about this, but network management has always been a thing, right? And there've been managers of managers and everyone's had their domain management. And I know there's a lot of companies out there now who are talking about AI with network management, but it seems to be a lot just putting
John Capobianco (:you
Jason Presement (:large language models in front saying, hey, tell me what's going on or something like that. But you guys are taking it a step further. mean, there's network management, there's network observability, but what you're, what you seem to be doing is extracting data or collecting data, whether it's structured, unstructured, SNMP, whatever the case may be from everything, whether it's desktops, network elements, everything, the network, pulling it together. And then through simple sort of queries, being able to
on either understand what's going on or be notified really what had happened with, with really meta information or root cause type of things. Is that, does that sort of sum it up? Is that correct?
John Capobianco (:Yeah, it really does. With a hyper focus on the network, the founders of the company are from Juniper. They used to lead routing and switching at Juniper. It really is from a network perspective out, but you nailed it. It is not just that it has a natural language interface in one direction. You mentioned the alerting. Those root cause, we all probably have a folder in our email.
Jason Presement (:Right. Yep.
John Capobianco (:know, call the alerts that we add filters and just dump all this stuff into, right? And then try to make sense of it. But, you know, the smart alerts actually identify the root cause and then associated events, right? So here's the actual root cause that the AI is detected. And here's the subsequent related events.
Right this interface that went down caused BGP, ISIS, OSPF interface down. Right we actually make sense of that in a nice human language message in one direction that then you can interrogate and follow up with your own natural language in the other direction.
Jason Presement (:So how do you capture that? what's it built on? And how do you, I know in one of the other podcasts, which I'll talk about later, the heavy networking one from Packet Pushers, you talk about sort of there's this, the different circles or the concentric circles that get built to put this information in place and really, you know, not only inform it with something that understands networking, but then understands your network as well and really what's going on.
John Capobianco (:Right, so I think there's two components and there's sort of a dividing line between the machine learning aspects, right? And we ingest that data and create the natural baselining and do the anomaly detection. But then there's the generative AI for the natural language interface. So we do start with, know, organizations don't just want an interface into chat GPT, right? There's regulations, there's restrictions. They don't want to share that data.
Jason Presement (:Mm-hmm.
John Capobianco (:So we take the llama three model, the metas llama three model and add a, add a first add an outer band of fine tuning. And it's not really a band. It's just for you can visualize it in your mind, right? Like so that we have the core, the llama three model. Then we add a layer of training, a fine tuning of SQL to natural language. So that that's where the company's title comes from selector is in a select statement.
Jason Presement (:Interesting. All right.
John Capobianco (:Because
in that data lake, and once we've normalized all that data, we do transform it into SQL. But we don't want to have to force network engineers to become DBAs all of a sudden, So that natural language interface really democratizes complex SQL queries.
under the hood that we do display. We do show these queries in the dashboard after they've been executed so you can understand and validate the call. And then they turn into smart dashboards, honeycomb views, graphical views, timeline views based on the query results. So the actual outer band of fine tuning that we add is that network state information. So.
Jason Presement (:Okay.
John Capobianco (:And additionally, the metadata, it's not just syslogs and SMP, it's streaming telemetry, it's REST API data, it ties into source of records or source of truths like Netbox or Notobot, or there's others out there, CMDBs. So, you know, the information is actually...
contextualize to give you the insights that you're looking for, right? It's not just another dashboard with another alarm, another, you know, glass of pain, right?
Jason Presement (:Well, so on that glass of pain or that's actually an interesting one. So just on that actually. So would you see yourselves as AI of AI the same as you would have a manager of manager? Meaning that if I've got systems in place already, how do you are you getting rid of those systems? Are you integrating with those systems? Are you pushing information back into them to help them do their job better? Or or do you see yourselves really as this manager of managers or or?
AI of AI for the infrastructure.
John Capobianco (:Well, I think on a couple of levels, I feel like artificial intelligence is an augmentation to our current skill sets, right? And Selectors has a similar approach where we don't, if you read any of our materials, we're not claiming to be there to displace or replace any existing technology. If you are a solar wind shop and you've been using solar winds for two decades, the last thing you want to hear is we can replace your solar winds, right? Or if you're heavily invested in Splunk.
Jason Presement (:Mm-hmm.
Right.
Mm-hmm.
John Capobianco (:or other ecosystems or your own custom DIY at home. You know, we want to be an augmentation and be able to ingest data from those platforms and make sense of it, right? So that we have less pillars, less, right? An application team, a network team, an infrastructure team, the database team, right? If we can get all that data into a data lake, we can really use this new revolutionary technology to cluster similar events.
Jason Presement (:Yep. Yep.
Mm-hmm.
John Capobianco (:do correlations and say, listen, here's where the problem lies. Here's exactly where the problem lies. I've seen configuration management issues, where a configuration by a human is the root cause. This selector easily identifies that. Here's the configuration change. Here's all the problems that it caused based on the TACACS or radius data. Here's the user that actually made the configuration change.
Jason Presement (:So, and will you go out and actually effect change or today is it collecting data and displaying and basically saying here's where the problem is, go take a look at it, maybe part of root cause analysis or are you actually going out and fixing the problem in some cases where you've got the ability to do that or at least spin off scripts and things and whatnot? Yeah.
John Capobianco (:Right, some, yeah, exactly. Some
customers are, let's say, closing the loop at the end of this smart alert. Well, we know what the problem is, is the next natural step not to just remediate it. We hand off to tools like Itentual or IP Fabric or many other NetBrain, other tools that do have the automation built in. I don't want to just limit it to those three. There's a lot. But...
Jason Presement (:Mm-hmm. Mm-hmm. Mm-hmm.
Mm-hmm.
Sure, there's a lot, yeah.
John Capobianco (:But we're really trying to focus on bringing that root cause identification and then working with solutions with partnering up with our customers to say, do you want to close this loop? Are there ansible playbooks you want to run or Terraform jobs you want to run or any type of how far do they want to take it? I think that
Jason Presement (:Right, they have a deeper level of integration, Mm-hmm.
John Capobianco (:A level of trust needs to be established before we then say, now let's actually solve these problems. But I think over time that that will happen more and more where we're gonna close the loop on this automation. The big service providers are really interested in that aspect.
Jason Presement (:Yeah.
but you're not just service providers ready mean you really your your you're dealing with anybody who's got an enterprise any network it doesn't really make a difference anybody who's got any infrastructure at all
John Capobianco (:No?
That's right. We're expanding into financial services, healthcare services, manufacturing. Retail is a big one, especially with a lot of branch remote offices because of our topology views. That's been very interesting and exciting. No, it really is more than just service providers. I think they wanted to start somewhere and I think that that was the, it was comfortable to them coming from a Juniper background to go into the service provider space. But
Jason Presement (:Yeah.
Mm-hmm.
Right. It was comfortable, right? Yeah. Right.
John Capobianco (:You know, more and more that people actually see the results of this. Like I'm really excited for Jeremy Schumann's and Autocon 2 video to be released, right? I'm really excited for people to see that. If you actually want to see what a customer is doing with it, I suggest you seek out that video when it comes out.
Jason Presement (:I'll
see if I can get a link to put in this with the notes when I put this together. Yeah.
John Capobianco (:Yeah, hopefully it'll be out before this before this goes. But
there are also like the networking tech field day. If people actually are interested on the selectors YouTube, there's a networking tech field day. It's like an hour long deep dive into the product with real demos at the end. Yeah.
Jason Presement (:Mm-hmm.
That's cool, okay.
And you guys have active customers, right? This isn't something you're planning on deploying. This is something that is actually deployed with customers, right? That's an important point, yeah.
John Capobianco (:no, no, that's a very good point. Yeah, this
is not proof of concept or someday down the road or imagine a world where we are actually solving problems with customers. Artificial intelligence, there's a lot of that, which I think is healthy. I think dreaming of the possible and coming up with new tools and utilities. We have a free for the community.
Jason Presement (:Right. There's so much of that.
Mm-hmm.
John Capobianco (:We have this packet copilot where you can upload PCAP files and chat with them in natural language. That's free and private. We don't scrape any data. That's on, yeah, it's on packet, packet copilot dot selector dot AI. And we actually have five or six example packets for people to download if they don't have their own. Yeah, so, and you know, that's, we put that out for a few different reasons. We wanted to give back to the community. We had this technology and this capability and it's not really something
Jason Presement (:That's on your website. That's available from
play with.
John Capobianco (:that maybe down the road in a few months we may try to incorporate it into the product. But for now we wanted it to be something to get network engineers excited about artificial intelligence and give them something tangible. Why don't you chat with a packet and start thinking about your prompts? How would you actually speak to a packet? Because it's a skill, right? Right.
Jason Presement (:Right.
Right. That's interesting. Right.
It's garbage in, garbage out. I've learned
that with just prompting in some of the tools that I use about, you ask it a stupid question, you get a stupid answer. Actually, not a stupid question. You ask it an incomplete question, you get an incomplete answer. It's more on that line. So in the deployments that you have, and I understand you guys have a fair amount across different industries, any examples of use cases where
John Capobianco (:Right, right.
Jason Presement (:Customers have seen really impactful outcomes from what you've done or just examples of how folks have used it to solve big problems that otherwise would have taken them forever.
John Capobianco (:Well, there are many use cases in that regard, but I know in the service provider space, in the carrier space, the transport networks, those are very costly links when they go down. So we've actually been able to give the smart alert to these ISPs to tell them, there's a transport problem on this link due to these conditions and the impact it's going to have or has had.
That has been transformative.
Jason Presement (:So, so it's preempt, so that's interesting. So are you, for things like capacity planning or network planning or even preemptive things, so you're actually able to look and detect something that preemptively could cause a problem and notify, hey, something looks like it's going bad. Get on it now. Do something about it before it actually takes down something terrible.
John Capobianco (:Yes, there is predictive aspects of this, I don't know, like an optic temperature is climbing. We can see, in this many days or this much time, we're projecting that to fail, right? And you can proactively take care of that. Now that at scale is quite interesting and solves quite a few problems preemptively. Be able to schedule a maintenance window and fix that optic or fix that transceiver or whatever. A hand of it just crashing and burning on you, right? Quite literally.
Jason Presement (:Mm-hmm.
So unfair question maybe, but you remember the Rogers Outage a couple of years ago, right? So if they would have had selector.ai implemented in their network at the time, could they have just typed in what the frick just happened before? Right.
John Capobianco (:Yes.
Well, I would hope that it would just come to them and say, hey, here's, you know, this is what happened.
Someone made a change on this router and here's all the impacting issues. It really is. It's a good example.
Jason Presement (:But that's an example, right? But I mean, is that an example of
where something could have said, here's, I mean, I don't, know, I think we all have our ideas of what happened there and everything, but it's a matter of why wasn't it resolved more quickly and, you know, won't get into that. But would that, what you're doing have cut out hours and hours and hours or even a day of time to repair that fault?
John Capobianco (:That's
think we're
getting closer to the real impact of the tool for most of our customers is that mean time to let's say innocence or at least mean time to detection. I know when I used to run large scale networks, the finding the root cause was 96, 97 % of the effort for me most of the time. Once you figure out it's a Configure 8, you roll back to Configure, you apply a new Config or you replace the part or you divert traffic or you call your ISP.
Jason Presement (:Right. Yeah.
Right.
John Capobianco (:Right? Once you know, is really quite simple in some cases to resolve that problem. But how amongst potentially dozens or more dashboards and silos and collectors of logs and telemetry and all the information, right? We're expected to be experts in every form of language and structure of data as humans, right? We just, even large teams of humans,
Jason Presement (:Mm-hmm. Mm-hmm.
Mm-hmm.
John Capobianco (:You know, there's one of our organization that ingests about a terabyte of data every 24 hours, a terabyte. So how many people would you need realistically to go through that and then say, actually, here's the root cause and I've connected. That's, you know, one source of, that's just the raw data, right? That's not what the machine learning is crunched and correlated and graphed and made knowledge graphs of and put it into the vector store.
Jason Presement (:Wow. Yep.
to go through that data, yeah.
Right.
Mm-hmm.
John Capobianco (:It's just humans trying to absorb that much data as best as we can, right?
Jason Presement (:So you're, you've got me going on a different task here, tangent about just the data itself and where it's stored. We'll come back to that in a sec. But in terms of when you guys talk to customers, the return on investment for this has got to be, there's a tangible piece, there's an intangible piece, but what do you typically base it on? Is it that, you know, the intangibles of customer reputation and things like that, or is there a real downtime piece that you go after, or how do you position it with customers?
John Capobianco (:Yeah, well, there's
certain like the reduction in tickets. I think it's a measurable standard that we were happy to work with customers to figure out how many incidents or events or tickets or whatever. and even that alone, it's funny at a certain scale that starts to become expensive. If every single alert and every single alarm triggers some cloud ticket and you're paying per ticket. Well, now we can create a single ticket with all those events in that single ticket. Right? So we're down from.
Jason Presement (:Right.
Mm-hmm. Mm-hmm.
John Capobianco (:however many tickets per event, per incident to just a single root cause ticket that's being created, right? So the reduction in tickets, the reduction, the signal to noise ratio becomes a lot clearer, right? You're not, it's not just a fire hose of alerts and alarms and events that people are trying to figure out and connect the dots. The dots have been connected for you, right?
Jason Presement (:analysis, yeah.
And you know
exactly what you have to go and fix, go and fix it. Yeah. Okay. And back to the data piece. is this, is it housed in the cloud? Is it onsite? Is it, is it hybrid? How do you guys architect the?
John Capobianco (:Right.
Yeah,
I don't want to, we sort of it's whatever fits the customer. We do have a cloud that we can host and that is probably the path of least resistance for customers is for us to host it in our cloud, provided they can, we can work out the security and the privacy and the data. It's all about the data. You talked about the data, right? So if we can figure that out, we'll happily host it in our cloud or a customer's cloud. If they have a cloud of choice.
Jason Presement (:Yep. Yep.
John Capobianco (:or even on-prem for customers provided they have some GPUs. And it's not a lot of GPUs, but if they want the natural language processing and the network language model on-prem, GPUs are going to be preferred, right?
Jason Presement (:It'll all come down to their individual governance and whatever they have in place to get things done, right?
John Capobianco (:Right, right. But
I don't want anyone to think that if, you know, we are sort of, we try to be cloud first, but we're happy to work with on-prem because of governance. And we do have customers that totally run this on their own stack on-prem.
Jason Presement (:I can, I can imagine there are many who are concerned about putting things in places where they don't have control over it. But that's, that's, but that's just the nature of the nature of the industry, right? Where you are. So how long does it actually take before it becomes, I'll say smart, right? So how long does it have to learn and be trained on the network before it can actually action something that might happen?
John Capobianco (:Right, right.
Right, so I think the initial fine tuning, the initial phases, you know, takes a little bit longer, a few hours as opposed to a few minutes. It's not, no, no, it's not, it's not, right, so we're not actually training a large language model, let's say we're fine tuning an existing model. So the Delta and the time really is big and we can do this in hours to, know.
Jason Presement (:okay. So it's a or days or whatever it is, it's right.
John Capobianco (:to a 15 minute cycle when changes occur. It's naturally always learning from the data on the cycle that the customer specifies. So it understands configuration changes in a positive way that someone actually made a change that affected things that didn't necessarily mean there's an outage, right? Otherwise we would have real problems, right?
Jason Presement (:Yeah, I guess it's a, actually hold on, I'll take that part out because I completely lost my train of thought there. Where was I gonna go? So it gets, hold on, I'm thinking here, it gets smart, it understands what's going on. do you think there's a time, do you think there's a time where the networks become fully autonomous, where you've got tools like yours that really understand what's going on, the feedback loop is there, it understands how to fix the network.
And at some point in the future, people just rely on it to get done what it needs to get done without human intervention.
John Capobianco (:Well, I think anything is possible based on, and I based that on just my two year journey with AI. Had you told me two years ago, a little more than two years ago, it was November 30th when Chat GPT came out, the things that we would be able to do in just two years with that initial technology, I never would have believed you, right? I really would have kind of said, really, I don't know. I don't think so. So for me to dismiss,
Jason Presement (:Mm-hmm.
Mm-hmm.
John Capobianco (:a fully autonomous network system, given the trajectory of what I've seen, I think we're going to start closing more loops. And yes, I think with this rise of agentic AI and little agents, we may interface more with agents than we do with underlying systems. I think there's going to be a nice abstraction layer there of agents.
Jason Presement (:Yeah.
Mm-hmm. Mm-hmm.
Interesting. Yeah.
John Capobianco (:that do most of the work for us based on natural language prompts. And then from there, once those agents all get together and get connected, there's an interesting thing. Look up Minecraft AI agents. They program thousands of Minecraft agents. They were like building religions and creating social structures and coming up with currency and...
Jason Presement (:Doesn't
that scare you a little bit though? know, unchecked and without guardrails, right? Just in governance, just, you know, your network takes over the world, becomes a sentient being and all of a sudden, you know, nobody can do anything anymore.
John Capobianco (:It's, it is really...
Well, I don't know. I don't
know. It's exciting. think, I mean, I think we're at a point where these networks and systems on top of networks are getting to the point of complexity that humans need in artificial intelligence really to understand what's going on as a whole, holistically. think individuals, I might be an identity services engine expert, right? That might be what I bring to the table, but...
Is there, I don't know, it's hard to find someone that understands the holistic entire system that you've built, like a doctor understands a human body. Although I guess doctors are specialized too, right?
Jason Presement (:Mm-hmm.
Well,
yeah, exactly. mean, there, you know, you've got your, your, your GP who's a generalist and they fire me out to the specialist when, they need the help and that somebody's got to find root cause at some point. Right. And like everyone always says, go get another opinion. So it's interesting on that side of things. Well, it's, it'll be, it'll be interesting to see where all this goes. And, I know that one of the things I wanted to come back to just as we talk about AI and, I think a lot of people, not a lot of people, but I think a fair amount of people are familiar with,
John Capobianco (:Right, right.
Right, right.
Jason Presement (:what's a call retrieval augmented generation so the whole rag side of things are are you guys sort of taking that a little bit further with some fine tuning is that you
John Capobianco (:So it's, yes, we've actually taken retrieval augmented generation and combined it with fine tuning to do raft. Retrieval augmented fine tuning. So this might be hard to follow for some people. I know it's a little hard for me too. But with retrieval augmented fine tuning, you're using AI to generate the data set that then you train the AI with. Right? So then,
Jason Presement (:Okay. Okay.
John Capobianco (:retrieval augmented portion you're going to say something like here's what I've done at least please generate 150 questions about the routing table and the rag has the routing table in the vector store so it generates a data set of 150 questions that then we use to fine-tune the the model with
Jason Presement (:Okay, so it's, and that's an ongoing process, right? That's a continuous process overall. Okay, so where do you guys, where do you see the selector product heading in terms of the new functionality? I'm not saying, hey, disclose roadmap, but where do you see it evolving to in terms of new applications, new functionality, things, problems that people have asked you to solve? Because that's probably where a lot of it's coming from, right?
John Capobianco (:Great.
Right. So
we're getting a lot of very interesting problems and I don't think that there's any limits on what we can do or can't do from an infrastructure point of view. We're excited. I'm excited. And what I think is going to happen is it's going to become a more commonly known name in the industry. It's going to become a more sought after tool from organizations and from network engineers and operators and CTOs, cetera, as they see the capability. I think
Jason Presement (:Mm-hmm.
John Capobianco (:It's still early for us, but I, I really do see it being, you know, you talked about going back. I'm trying to think of the first network management system I used. was either, LMS from Cisco land management server, or it was what's up gold from Ipswitch.
Jason Presement (:man.
I remember that one actually. That's funny. Alright, so it's been a while.
John Capobianco (:Okay, so that's, yeah, that's, going back to about the what's up gold days, early 2000s, right? Anyway, I
really would like to see selector adopted and just because it actually is implementing artificial intelligence in a way that I don't see very often. It's actually a practical augmentation of existing skills and solving real problems. Root cause analysis, natural language interface.
Jason Presement (:Mm-hmm.
John Capobianco (:meantime to detection and innocence and even recovery.
Jason Presement (:So is it for every, I mean, it sounds like it's for everyone, right? It sounds like it's for anyone who has a network. Do you guys sort of scale down for the small folks or are you thinking about a managed service sort of approach to be able to maybe handle the guys that don't have the infrastructure from a human resources perspective, perhaps, to take on a full implementation or something like a Rogers or a Bell or whoever would take on?
John Capobianco (:It really is. It really is.
Right, that's hard for me to say. That's a better question for my leadership, but I don't think that we would turn down opportunities to help solve problems, right?
Jason Presement (:Opportunities? Right.
Fair point. Fair answer. All right. Anything you want to, I'm trying to keep these podcasts sort of short so people can, again, take the dog for a walk and listen to something and learn something at the same time. Anything you want to throw in there that we haven't discussed that you think would be interesting to the tens of people that read my newsletter every week?
John Capobianco (:Yeah!
Well, no,
just, like I said, we do have a free tool in the Packet Copilot. I really do encourage you if you're in networking or even not networking. If you're interested in Jason's content, I think you're going to find that quite interesting. And share your results with us. We're really excited about the prompts. Like I haven't thought of everything and there's some emergent behavior such as using multilingual.
Jason Presement (:Mm-hmm.
John Capobianco (:Someone started chatting with them in Spanish and I didn't, I never thought anyone would ever use Spanish to chat with it and it worked. It responded in Spanish, right? Someone else asked to visualize the packets and it made a little ASCII diagram of the packet flow for them, right? I never thought it would do that.
Jason Presement (:That's fair. Yep. Yep.
Yeah, interesting.
Mm-hmm.
So you're learning, yeah, you're learning with your customers, right? You
know what you've asked it to do, but there's always gonna be somebody, there's always a corner case and those are the ones that ends up causing the problems. There's always gonna be a core, I mean, problems in the network perspective, not in the solution perspective, let me clarify that. But it's those corner cases, I think, that really help you guys be able to, like you just said, come back with these anecdotes around how people have used it and problems they've solved with it and how they've done it.
John Capobianco (:Yes, yes.
Yeah. Someone
asked about SIP and RTP traffic, if it could do voice and multimedia and it worked with SIP and RTP traffic.
Jason Presement (:And it'll do video now, like video stuff as well. Hey, why isn't channel 13 working or something like that? To the extent that, you know, that that data is available, I guess. That's it's there. Okay. so just before I ask you where folks can find out more. So packet pushers, you did episode seven 58 with them. so packet pushers, heavy networking podcast. I link it in issue 48 of my newsletter.
John Capobianco (:you
If it's in the packet capture, that's right.
Jason Presement (:which is coming out this week, but this podcast won't be coincident with that. So if you want the link to that, it's an issue 48 of Jason's Industry Insights. It's issue 758 of Heavy Networking Podcasts under Packet Pushers, where you go into much more detail, right? I've just touched on some stuff. There's much more detail in that. And John, if they want to get ahold of you or somebody at Packet Pushers, at Selector, what's the best way to do that?
John Capobianco (:Sure, so me individually, I'm always on my LinkedIn and my social networks. I've just joined Blue Sky, which I think is a wonderful social network platform. I'm sort of migrating there. Not full time, like I'm gonna still keep my Twitter active. So if you're a Twitter fan, I'm on there as well. selector.ai is our website. The packet copilot.selector.ai is the packet chat. If you want a demo, like if you're...
If you are interested in this seriously as a demo for your organization, reach out to me and I'll make sure that I get you in touch with the right people geographically and based on what kind of demo you're looking for. We are looking for use cases and problems to solve. Yeah.
Jason Presement (:to solve, yeah. Okay.
tening to this, it's probably: John Capobianco (:Alright.
Happy New Year. Welcome to the year of AI agents. I'm going to leave
people with that. Try to look into, I'm serious, look into AI agents. They're fun and simple to write. If you have ChatGPT, ask it to help you make an AI agent for a weather app. In 25 minutes, you'll have an AI agent in Python for a weather application. No problem.
Jason Presement (:Interesting. That's interesting, actually.
That's, know what, there's so much, think that people it's, this, you don't know what you don't know. And sometimes, yeah, absolutely. Absolutely. No, but even, mean that like just the AI agents thing and everything else, it's just, you just, there's so much, right. And how far do you go? And, and it's, it's this constant need for people to want to find out more. Right. And, dig. So to your point, right. Keep asking people, keep looking, following people online and chasing people around to see what they're doing.
John Capobianco (:Well that's why we're having these conversations though, right Jason?
Right, right.
Well, I'll tell you what, let me come back in March and we'll use the bellwether to see how close I was on this AI agents thing. We'll give it three or four months and yeah.
Jason Presement (:Well, there you go. We'll give it three or four months. Excellent.
All right, John, I really appreciate you helping me out with the podcast and representing SelectAron here. And thanks for your time. Really appreciate it.
John Capobianco (:Hey Jason, this was a lot of fun and I wish you all the best. I think more voices the better and I've been reading your newsletter. It's great stuff. All right. All right, take care.
Jason Presement (:Thanks.
Perfect. Thank you. Appreciate that. It's a labor of love for sure. All the best. Thanks.
