Welcome And October News Setup

Chris 0:12

Hello everyone and welcome back to another episode of the Cables2Clouds podcast. Uh today on the show, we are gonna have the next installment of our monthly news recap uh going over uh some of the major news articles that we've had come out in the last month, uh mainly around October in general. So uh joining me is my good pal, um always smiling, always a happy face, Tim McConaughey. Um and let's just roll right into it. There's a lot that's happened over the last uh last month. I'm sure uh, you know, if you're if you're working in operations or anything like that, you've probably been a very busy person. So let's let's hop right into it and see if we can cover some of this stuff.

Anthropic S-1 And IPO Reality

Chris 0:52

So uh first thing, um, which is, you know, if you've unless you've you know been completely under a rock, you've probably heard about this in some assemblance. But so anthropic um has been kind of the the first to kind of motion towards their um uh their public IPO. Um and the first step of that is issuing a basically a public S1 report, which is kind of a registration statement that uh the company is usually legally required to do um prior to um doing that IPO. So in this report, um there's a lot of a lot of things. Uh I'm not a person that necessarily has the intelligence to go through an S1 report and make an accurate um kind of depiction of a of what a company is you know succeeding at or failing at or what have you. But um from what I've gathered, it sounds like there's a lot in there, and we're kind of uh in this this new era where um I don't know, there's a lot of things happening in here that you usually don't see, um, I would say, and there's a lot of things happening in um where you see them operating at kind of a kind of a tremendous loss, um not even profitable at this point. Um, you know, if you kind of think back to the um the days of old with kind of these 5x performers in terms of you know software companies that have gone to market and and you know achieved great heights, such as Microsoft. I think it was reported that pretty much every single one of those companies prior to filing an S1 has been profitable. So this is kind of a complete opposite step in that direction. Um, so they've reported an 81 billion net loss um for the first half of 2026. Um again, 16 billion in revenue with a huge compute and uh commitment of 161 billion to Broadcom, 120 billion to SpaceX, 114 billion to Google. So it's it's obvious that they've already got a lot of commitments in place, which is um, you know, I don't know if you're doing the math, but that's a that's a lot that they have to live up to over the uh the next uh few years. Um so yeah, I'll I'll take a pause there. Tim, how do you how do you feel? Um you getting in on day one or what?

Tim 3:10

Uh yeah. So the reason to the S1 is to basically let potential investors know everything that you know it's legally required so that you don't uh essentially dump worthless trash stock on the on the market, or at least if you do, that people will know what it is that it is. Uh yeah, I anthropic I guess is brave for doing this first, but I expect any of the other frontiers, if they put out an their own version of an S S1, it'll probably look just as ridiculous. Um 160, 120, what is it? Like six, what are you looking at? Like 600 billion in com in commitments with at an 800 billion net loss like already just from operating. Like I mean, it's just oh my god. Yeah. I have nothing to say except how ridiculous that sounds.

Chris 4:05

Yeah, like like I said, that there's much smarter people that I've uh I follow on the internet that have kind of uh looked into this. It seems like everyone's kind of said that like it whether whatever story you want to build about anthropic, there is actually a page in this S1 that probably supports where you're going. Or you want to say there's no way they'll succeed, there's there's a page and a chart for you. And if there's um if you expect them to, you know, you know, go all the way to the moon, then there's there's pages for you, right? So um I think you can kind of build whatever story you want. Um

Gartner Warns Of AI Spend Risk

Chris 4:38

it's uh it's hard to read a read the writing on the wall at this point, but um, on the back of this related, we also saw an article come out from Gartner warning that there is about 3.5 trillion in AI infrastructure spending that could be at risk by 2030. So um they have uh indicated that spending could face impairment by 2030 with roughly 3.6 trillion of the forecasted AI application revenue falling about $5 trillion short of the $8.5 trillion that is needed. This is this is the most this is the most times I've ever said trillion on this podcast, without a doubt. Uh it's probably not even said once prior to this. And we're talking about trillions of dollars, which is just um not not increments that we've ever really had to think about in in terms, unless you're talking about like global GDP, which is which is fucking insane. Um and uh one interesting comment that also came in this in this Gartner report is that they're predicting that more than 80% of these neo cloud providers, uh, which there's a there's a number of them now. Um if if they're not household names to you, probably a few of them will be in in the upcoming years. But this uh this this is predicting that 80% of those will stop operating independently by 2030 as well, which uh that doesn't necessarily mean that they're going to fail, but that does mean that it's likely that they'll be part of a merger, an acquisition, what have you, which which makes sense. They're they're moving at kind of these velocities because they can kind of do whatever they want. Um and you know, I know I know Tim, you're kind of working with this a little bit in your day-to-day now. Do you have any any commentary here?

Tim 6:18

Yeah, so I am working uh internally with some neo cloud providers. We're working on some partnership stuff, go to market, et cetera. Just just technical prove-outs for a lot of the stuff. I can't get into it, right? Obviously, for for for clear reasons, but I can say working with a lot of these neo clouds that you know, a lot of the neo clouds operate kind of on their own, recognizance, right? There's no real standard, there's no real it's just think of like uh uh a GPU farm that maybe it used to be a Bitcoin farm and then now it's being used for AI, and they just kind of like built wrapped it in some kind of shell, right? Some some providers uh drop ship racks of gears to places, some some providers have like pops where you can come get it, like for example, Latitude Megaport, right? That's kind of their thing. Uh Megaport acquired Latitude last year, I think it was. Or two years ago now. I think it was last year, yeah. Yeah, yeah. Uh so that type of thing, right? So but the point the important thing to understand is that these none of these ni neo clouds. I know we we grouped them together and say neo cloud, and what we're really talking about is like GPU farms, basically. Um, but they don't there's no standard for neo cloud, right? Like there's no it's not like AWS and and Azure and Google Cloud that kind of have like a similar offering. No, it's complete Wild West, right? So um it'll be interesting to see if these neo clouds uh continue to operate independently or if they just get snapped up just for well hardware if nothing else. I'm not I'm not sure, right? So yeah, it's interesting.

Chris 7:53

Yeah,

Data Center NDAs And Community Pushback

Chris 7:54

agreed. Um maybe also just kind of in relation to uh kind of data centers. I know it's kind of a uh a daunting topic that not a lot of people like to hear about, especially in this day and age. Uh we were just talking before we hit record about how it was funny how data centers become this like taboo word um, you know, prior to the the midterm elections that just wrapped up in in the US um at the time of us recording this. Data centers obviously played a huge role in that, which which makes sense. But uh, you know, being on the technical side of this, it's funny how um, you know, you can talk to um you know your your grandmother and she can be talking about how you know data centers are bad when you know in the all actuality, it's yes, 90% of the data centers that are getting built today are probably bad, but the term data center is not necessarily a bad thing. Like it's it's uh a lot of companies run their own data center, whether that's in their own facilities or not. So it's it's just funny that that's become a taboo word. But um we do have an article here as well from Network World saying that the hyperscalers are allowing more scrutiny over controversial data center plans in that they're loosening the concept of the NDAs around the the build-out of these data centers. So um Amazon's data center commitment uh will disclose more energy and water data and consider limiting NDAs for official reviewing plans. Meta is also reviewing the NDA use while Microsoft and Google have already moved away from it completely. So opposition pressure is obviously mounting uh to these kind of being these fully non-disclosed um details about about what's going to get consumed from the communities that they're going into. So Tim, any comments there?

Tim 9:41

Yeah, I mean we've talked about you know, since it's been in the news, right? So oh well over a year or two, um, about all you know how these these companies will they want to build a hyperscale data center, which really is what we're talking about when when people like in the news, when you keep hearing data center, data center, that's what we're really talking about. We're talking about those 10,000 square foot like monstrosities that guzzle rivers or the water and and all of that, right? So focusing on that, you know, there was this kind of subterfuge or stealth, stealth mode thing where these companies would would want to build these these hyperscale data centers, and you know, they'd go to like the city council and essentially pay them off, or you know, I won't say I guess I can't say that for sure because uh necessarily some of those haven't been proven to be that way, but but you know, you know, right? So, but then these these uh communities would suddenly find out that hey, you know, 10 feet out of the city limits, we've got this, or maybe even within the city limits, right? Uh we have this new hyperscale data center being built, and it's got tax breaks, it's gonna eat up, you know, God knows how much water and power. And people started really uh rebelling against this, as they should, right? So I think now what's happening is we the pendulum is actually swinging enough, and that you know, never let it be told that activism doesn't do anything, right? Um, communities are now preemptively like putting in like moratoriums on data centers. Like we're not even gonna talk to people from data center companies for a period of time. So these data center companies now they're realizing that hey, we can't just kind of like slide under the radar. Like if if we don't uh disclose stuff, right? Like we're just gonna get locked out. They're gonna ice all more and more of these of these locations are just gonna ice us out. They won't even talk to us, right? And that's how it, and honestly, that's great. It should how it should be if they're not gonna disclose. So this I think is just them trying to uh manage, right? Ma manage the the PR and the the all of that around it. I I they're certainly not doing it out of the kindness of their heart, right? So I do think it's a good thing though.

Chris 11:55

I don't know, man. No, I'm just kidding. Yeah, it's definitely not one. I guess I guess they're gonna have to put these things on the fucking moon. I guess uh Yeah, yeah.

Tim 12:02

Elon's gonna put 'em in orbit or whatever. Now that uh absolutely. Um okay. So yeah, so so short answer is uh we're gonna have to wait and see on this stuff. I think I I still have to I still stand by what I've said from the beginning, which is that um we can build the data centers, we can't give them enough power, and we can't fill them with enough racks of silicon. So why are we in such a hurry to build all these data centers? Yeah. Right. Anyway, all right, uh moving on.

eMudra Certificate Fiasco Explained

Tim 12:37

Two other huge uh stories. And when we were talking about operations, uh when when Chris mentioned operations at the beginning of the episode, I'm sure most of everybody knew exactly where we were going with that, right? So what we'll just do one at a time, right? These are so now so you know, we're late to the game, right? Because we're when we're recording. The first one is uh this Indian uh CA uh certificate authority called E Mudra, I think maybe how you say it, eMudra or E Mudra. Yeah. E Mudra. So they were they basically became an intermediate uh certificate authority, which means that you know, a root certificate authority entrusted them, gave them, you know, to sign certificates for it for on their behalf. And what that means is that when you get a signed certificate from this intermediate CA, uh it presents, you know, you attach it to a web server or whatever, you you attach it to something, and then when uh we'll just stick with the web server example, when you know you go to a website, your browser has a a slew of automatically trusted certificate authorities in it, like pre-programmed, that's why you don't have to do anything. And you know, eMudra uh was an intermediate CDA for one of them. Uh the problem is that eMudra here, their intermediate CA, when you build this stuff, it's all cryptographic, right? So you're you're you're doing algorithms and you're doing cryptography. And I'm assuming by accident or something, uh they generated a cryptographic uh key with a num like number the number was what four? Like it had one single number.

Chris 14:16

There was like seven of them. Oh, yeah, it was like four, eleven, four, four four was the most comical one, I would say. But then there was like one. Yeah, exactly. Yeah, 237 or something.

Tim 14:29

Right, which of course could immediately without even bring you know, a person probably could have reverse engineered this without even the help of a computer potentially, uh, reverse engineered this. And the problem then is if you can reverse engineer or spoof the certificate, then you can become uh do like a man in the middle attack or something where you can tell other people, I am that browser, or sorry, I am that web server, and your browser will automatically trust it. But really, you know, you're stealing credentials or something like this. You're masquerading as a real web server, right? So this is obviously a huge deal. And you know, we think about things like certificate revocation lists and just the life cycle of certificates, you know, it's get this life cycle keeps getting shorter and shorter because the uh cryptography uh cracking abilities of computers keeps getting faster and faster, which is you know where PQC post-quantum cryptography comes into it. But but not to get too far off of it. The point is these keys were, I mean, compromised is a is a generous word. Uh and so uh and so what happened is like overnight, all of these issues, all of these certificates that were issued are basically like revoked and and useless because uh they are otherwise completely spoofable. So this became a huge problem. Lots of it, lots of things on the internet stopped working because these certificates were no longer good. Um and I believe Emuja actually has now exited the uh business, the the certificate authority business. They basically put out a yeah, they put out a they put out a uh a statement basically saying, Oh, we this was never really big business for us anyway, we're exiting, like kind of like you know, we're gonna take our ball and go home. Like, yeah, no, no, please, please get out of here. Yeah, yeah. Um and and it's it's it's something that's so comical, it's like you know, space balls level, right? Like a one, two, three, four. I have the same combination of my luggage, right? Like it's it's it's so comical that nobody looked at it, right? Like that's how because it's been out for years now, right? It's been this way for years, and it's one of those things that's so unimaginable that nobody really thought to try. Like to try to like see, hey, can I reverse engineer these keys? Like, you know.

Chris 16:50

I think they did come out and say that it was something like it was it was a a mishap in the the automatic rekey kind of process that they had where something had accidentally got rekeyed without, you know, obviously obviously there's no validation there. There was no validation to make sure that it was it was uh done properly, and that's where these kind of seven certificates uh made it out into the wild. But I mean, obviously to your point, there's a lot of websites out there that you know had this in their kind of uh certificate trust chain. So there was many websites that would you know you'd be browsing to show to secure and and trusted by the root CA, but that that was you know could have been completely false.

Tim 17:30

Steal your credentials, etc. etc. Right?

Chris 17:34

I mean, thankfully thankfully in terms of the I think in you know, kind of general like internet browsing, I think it was reacted to relatively quick. So obviously the the the the browsers, the major browsers, you know, being you know, Google, Mozilla, uh Apple with Safari, things like that, they they were able to actually get this in and update out to the public very quick. Um now if you haven't um updated your fucking browser since this has happened, you probably want to do that pretty pretty soon. Um but so they I feel like they took care of that pretty quick, which was nice. But uh the thing that's um also a bit concerning is that um I think a lot of these certificates were also used for like software signing, um or actually like you know, like uh solidity of actual software. So if there's if there's dependencies out there that um where you know specific software was signed with one of these certificates, and obviously I think that could be that could be compromised as well. So um I actually haven't heard much on that since. I don't know if there's additional fallout or additional kind of research going into that, but uh it'll be interesting where that comes out on the other side.

Tim 18:43

Yeah, I I'm I am curious because you would I mean if in theory how the CRL, the certificate revocation list should work is that that should just cover it anytime a certificate then gets checked, you know, as part of the software signing or or checking for software signing as part of like installation, for example, it should flag it and stop it, right? I now in real life, I don't know, right? Especially for something that's off the internet or unable to check the CRL, right? So that's a very good point.

Chris 19:13

That that's that's kind of what I'm getting at, is the things that are kind of air gapped, uh uh SCADA, like you know, infrastructure.

Tim 19:21

Uh yeah, absolutely. Yeah. All right, uh moving on.

PHP Unauth Exploit And Patch Urgency

Tim 19:26

The the second big one, uh big story out there, is the PHP vulnerability. And uh so they they released a uh they're calling it P2B uh the P2 PHP to shell is the kind of vulnerability. Um it was released now. I'm trying to think when was it released? It was released just a few weeks ago, like two let's see, patches, let's see, patches were released on October 17th. I think it was just last uh October 17th, yeah. I was like, I think it's about about two weeks now. But basically PHP, which has been around for many, many, many years, and and honestly, to the point where most a lot of people have moved away from using it for anything. Uh the software is is got a uh has has a vulnerability that basically just you know straight to root, like you know, like do not pass go. It's uh it's not the highest on the C VSS scale, it's a 9.2. But the important thing is that the exploit can be done without authentication on vulnerable servers, yeah. And you're then you're in, right? Like that's it's it's one of those. Um now why this is such a big deal is that even though PHP is old, it's also kind of ubiquitous, which is to say it's again everywhere on the internet pretty much has something with PHP in it. Yeah, if you if you run WordPress, like it's right. Yeah, like anything Lamp Stack, yeah, anything Lamp Stack, uh, etc. Which which was like you said, like blogs, anything anything pretty much on the internet, including vendor software as well, right? So vendors some vendors that have uh webs, web uh interfaces, you know, they're using PHP, like which means if you can exploit that uh shell, you know, depending on how of course how that web interface is integrated with the product, you could potentially bypass that the authentication and now you can own uh routers or firewalls or or whatever, what have you. So this, I mean, they did come out with a patch patches pretty quick. Our vendors are of course releasing patches for their own own products, etc. Um, but you know, SysO definitely advises that you take care of that as uh sooner rather than later. And I think there was one specifically for Palo. There's one that has recently been uh for Palo Alto exploited, right?

Chris 21:54

Yeah, there was an actual like exploited um piece related. To PHP on like some Palo firewalls, yeah. I think I think it was a relatively known number or a low number, I should say. It was like, you know, maybe like maybe 25 to 50 firewalls or something like that.

Tim 22:09

But I mean Palo's not alone, right? Lots of vendors have that have PHP in their uh web. I'm I I you know, for example, I think uh even our own companies probably have at least a couple products that uh were listed in the in here somewhere. You can look at the CVEs and check. Um yeah, so this one was big. Not so again, because because it was a large unauth because it offers unauthentic unauthenticated exploit, and because it's so ubiquitous, even though it's an old, you know, PHP is old, it's just everywhere, right? So that's two big operational problems with stuff that's just all over the internet pretty much at the same uh essentially at the same time.

Chris 22:48

So yeah. One thing

Disclosure Drama And Chainguard Fallout

Chris 22:50

that I think is actually more interesting than the vulnerability um is obviously a vulnerability is uh kind of uh you know uh the very common to sort of have something critical released on a weekly basis now with these frontier models. But um, so when this was discovered, um I don't remember the exact details. I'm probably gonna butcher this, but uh, you know, bear with me here. So I think there was there was actually the way it was discovered is someone was looking for using the frontier models to try to discover uh this specific type of vulnerability in a different language. I can't remember exactly what it was, but I remember the model apparently came back and said, like, well, I can't find that vulnerability with this, but I actually did find it in PHP, and then it and then it started getting disclosed. And as far as what I what uh what I remember is I believe a lot of the PHP maintainers um kind of basically created this kind of like um embargo, right? So a lot of a lot of the major vendors that were that utilize PHP as a as a uh kind of an open source project, things like that, use the in their in their um products or you know, offer, you know, like if we think about all the major Linux distributions, things like that, a lot of those folks were notified as well. And it's basically saying, like, hey, here's the vulnerability, here's a kind of a private consultation, so you guys can patch ahead of time, you can fix this, and then once it goes out to the general public, you know, a lot of these things already have patches available so they can immediately be done. Um, and one of the companies that was on that list was ChainGuard. Um, so if you're not familiar with Chainguard, they're kind of a software, I guess, supply chain company. So basically they can, I think it in layman's terms, they basically you can you can run Chain Guard images of all the major distributions of uh software and they kind of own that SBOM. Um and and and as far as I know, they own the the uh uh the piece about patching all of those open source projects that might be tied to that so that you don't have to do that yourself. So if you get something from ChainGuard, it's kind of the it's it has very few CVEs, you can it, you know, it can get be patched within hours, things like that. It's it's kind of uh what their MO was. And apparently ChainGuard kind of jumped the gun and released this. There it I guess someone had somehow the vulnerability had leaked into the public and someone found out about it. And ChainGuard, I guess, in in kind of depending on who you talk to, some people think this is a very selfish act, some people think that they should have waited. Um, apparently, like six hours before the major release went out, they put this something out to their customers saying, like, hey, a patch is available. And apparently within that six hours, it uh it took maybe less than an hour for someone to look at the diff of what they changed, reverse engineer it, and then they had the exploit basically figured out for a PHP, and that's why it was in the wild for a while before people could even patch. Um, so the uh it seems like there's some discourse going on amongst the community about like you know, if they've is like is Chain Guard gonna get removed from any of these notification lists ahead of time because they've done this now. Like, um, and you know, if they're put on kind of a blacklist, you know, I mean, how do you know if you're on a list until the next thing happens and you figure out you weren't notified ahead of time, right? So um, yeah, really, really interesting political story related to this one as well about where things are going to go. So um yeah, uh let's let's let's keep an eye out for the next one and see if uh see if Chain Guard's in the know or not. All right. Uh

Cisco Bets On AI Code Refactors

Chris 26:32

one that we'll squeeze in here, it's just a little bit too relevant to not cover. Uh, I don't think Tim's gonna have too much commentary on this one because it does relate to his employer. So um I'm I'm assuming he's going to plead the fifth on this one. Um so there has been uh released by my multiple publications out on the internet. Uh it seems like this is something that Cisco themselves have gone to the press with. Um so there's been a story that's uh come out and said that they've basically given a 27-year-old, I don't know why they keep calling out specifically it's a 27-year-old. I guess that's the uh uh their name is Patty. Uh their real name is Padamabon Krish Krishnamurti. Uh I probably butchered that name. I'm sorry, Patty, if you're listening. But um, so basically they they've essentially given him a blank check in terms of tokens uh related to the frontier AI models. Um and he he's basically committed that he can he can refactor 98% of Cisco's entire legacy code within 24 months. Um so it um I'm assuming this means most of the things kind of under that hood being Cisco iOS XE and you know, maybe any other versions of that. Um there's not a lot of details about what I what is actually happening on the back end, you know, who will who knows? I mean, based if we're going based on the way Reddit works, he's probably rewriting the whole thing in fucking Rust or something, I guess. Um but it's like look, I I don't think this is my personal take. I don't think that it is um I don't think that it's weird that Cisco is even going to make an attempt to do this. I think if you can use these frontier models to do such powerful things as you know recompiling Nintendo Switch games to run on a MacBook, uh, I think you can use it to refactor, you know, probably some legacy code that's been in the market for you know 30 plus years. So I I don't think that's a that's a a weird thing to take on. What I do think is weird is that they've they've specifically called out that it's one guy and he's 27 years old and he's gonna refactor the entire code within 24 months, and he's been given a blank check. Like all these little details from a marketing perspective are just weird. Um, which I don't I don't know exactly how to read that. Uh I mean I I I I think under the hood there's no way it's just one person. It's probably, you know, a team of, you know, maybe maybe it's a you know, to use the the the Andy Jassy thing, that maybe it's a one pizza team doing this or whatever, but it's um it's just strange that this has kind of been put out into the ether in this in this manner. Um but uh you know what I wish them well. I hope I hope it is successful. I hope it really kind of uh reinvigorates the the the code to a degree that is uh um uh I mean it like we said at the beginning of the show, if you're working in operations, you've probably been doing a lot of patching over the last few months, which um, you know, I I think we all, you know, from the vendor side, we all of us have to kind of up our standard and and get to a um kind of a a higher plateau so that you know the the actual customers on the other end are not affected by this. And um, I'm sure this is this is one of those steps. So I'm not I'm not saying this is a bad thing. Uh I just think it's weird the way that it was put into the press. So if you want to read more about that, have a look at the notes and that'll be in there. All right. Tim, you want to go on to the next part?

Tim 30:09

Yeah. Uh let's see. All right. So

AI Agents Inflate Telemetry Bills

Tim 30:13

this this is interesting. We're getting we're gonna we're gonna wrap it pretty soon here, but uh I thought this was a really interesting article. So AI agents are actually uh driving up the cost of telemetry and complicating network observability from a cost perspective. So this is this is interesting. Uh it's something you kind of make sense as soon as you start to actually stop and think about it, right? So the more uh agentic operations that that are out there, the more the data is moving. And you know, depending on where that data is, uh that could end up costing money. So um you know, Splunk did a survey of about 2,000 people, uh 2,000 IT professionals, and uh found that, you know, 57% spend more on observability than on monitoring their infrastructure, and 38% have reduced coverage or telemetry retention, creating blind spots. So what's interesting about that is that um it doesn't so this this particular thing doesn't mention the exact they weren't able to figure out the exact amount the AI agents are are kind of making uh the share is it? Uh sorry, my screen is actually blinking at me at the moment. So it doesn't actually get into like exactly how much the the share of AI agentic operations are uh causing here. But you know, the point is that the observ the cost of observability is actually exceeding uh the cost of the infrastructure that's actually being monitoring monitored, that's what it says, which which I find interesting as a uh a statement, right? So I'm aft assuming they were talking about like operationally, like opex style. Like they're paying you know, observability in opex is costing more than the op ex of actually running the infrastructure. Um another 60% said unexpected billing spikes create more operational strain than consistently high costs. So so basically like the spike the spike is worse than it just costing a lot, basically, because at least you can you can plan for it, which is I think what most people find as well. And that uh 55% cannot accurately control their AI expenses, which God, I think we've all understand that one, right? Because the more we do it agentically, uh the more it spirals out. Well, like basically the more you hand over to AI, you you don't at that point you've handed it over, right? You don't get to choose uh what the agent will do to accomplish the goal anymore. Um, so it's kind of interesting to see this idea that you know telemetry and agentic operations actually uh has a built-in cost associated with it, right? Because, you know, you're talking about retention of data, you're talking about move moving data, not just data, not data to do the business, but data to observe the business being done, right? And then make predictions. So um, you know, agents also dumb generate telemetry about their workflow, about tool calls. You know, think about it. Like if you ever sat there and watched an agent work, right? And and you told it to essentially keep a log of it, you'll see like, you know, agent doing this, agent doing this, agent. Like they think about think about now, think about the cost associated with the agent essentially reporting what it's doing, not just the cost of the agent doing the thing, right? But you can't not have that because you need to know what's happening. So it's kind of an interesting conundrum that we're we're finding ourselves as we move towards uh agentic operations and just kind of using agents more and more to manage or or uh diagnose, troubleshoot, uh you know, whatever, change our infrastructure. So yeah. What's your thought there?

Chris 34:05

It's it's an interesting one. I like because I now, you know, kind of after after reading and going through this, like I never even really thought about it, but yeah, I think once we started seeing this rise of um, you know, kind of agentic usage across the board and what that meant for kind of the natural patterns of using um not just IP, not just the internet, but kind of IT infrastructure in general. Um I'm I'm starting to I'm I'm sure this a lot of these I don't think this is unique to Splunk by any means. I think this is just observability companies in general, because obviously observability is um kind of a always a pay-per-use uh paper use. Yeah, it's consumption model. Consumption model. Yeah, exactly. So like I'm sure every single one of those companies, once they started seeing these trends, their eyes kind of lit up, right? Uh there's a lot to be gained from this. But I think, yeah, if if you're kind of on the if you're kind of on the practitioner side or on the on the operations side as a customer, I think you've got to kind of start weighing things like how much observability is actually valuable. Um, I understand if you're feeding that through another AI, maybe that is where the all the correlation happens, but then like it just it like this is just gonna exponentially compound, right? Like who, like, I don't know, like who pays for all this shit? Like we've we've you every single like vertical related to IT into in terms of AI, the who pays for this shit conversation is always there. And this one, uh, now that you know kind of we've cracked open the can of worms, it's like, holy shit, like this is gonna get worse.

Tim 35:48

Well, I mean, and what you're hoping, right? What you're hoping is what you've always hoped, which is via observability and and all of the oh my god, all the agentic operations that come from observability, uh, can we avoid or mitigate the outage that's costing us a million dollars every minute that it's down, right? Like because if it does, then hey, maybe that's maybe we're stroking a check for the right reasons, right? We're gonna spend ten thousand dollars and save a hundred thousand dollars, right? Although I as we know it's never quite that simple, right? It's never that easy to draw that, you know, correlation uh even with a even with an observability. It's not that easy to draw the correlation. Uh so I think that's going to be a question, though. I think it has been a question, but I think it just has to continue to be a question and become a larger question as operational budgets are going to continue to increase as we increase our capabilities of observability, not only observability, but what we do with the stuff we have like what we do with the data, the correlation, all you know, all that shit that comes out of doing the observability.

Chris 36:54

Well, I think, yeah, one thing it it does obviously it relates to operations and uptime and things like that, but I think where this will really boil down is just fucking governance. Like uh how how much are you how much telemetry are you obligated by your local government to to keep track of? Um, and there's probably gonna be a lot of people getting to the point where like they're just like, look, I'd rather eat the fine thanks. Yeah. Like it's people, it's gonna get real fucking grim uh in terms of splitting hairs on on what costs uh what and and what's worth it. Um so yeah, it could get ugly.

Tim 37:34

No, that's a that's a very good point. So uh yeah, so kind of an interesting thing. I hadn't really given it a lot of thought because you know, certainly as network engineers and with our our, you know, remembering all the the the 3 a.m. calls where you had to wake up and try to figure out what was wrong, right? Every we just think, man, more if I'd only had that level of observability, and if I could have asked Claude what I could ask, yeah, class asked if I could have asked Claude what was wrong and it could have answered me because it had that level of of detail, how amazing would that have been? But yeah, I mean there's like you're paying for the retention, you're paying for the data movement, you're paying for the correlation, like the AI part of it. Like, man, that's yeah, it compounds, like you said. So it's uh gonna be a really interesting balancing act that I'm you know, and as certainly as vendors, you know, ours included, uh, add more capabilities around 8 genetic operations because again, the goal is to avoid downtime and to maximize your utilization and ability to use the product. Um, yeah, like that's somebody has to bear that cost to uh of all that extra stuff. Yeah.

Chris 38:40

Who's holding the bag at the end of this is the is the question. All

Wrap Up And Share Request

Chris 38:46

right. Uh with that, I think we'll go ahead and wrap. So thanks for uh thanks for tuning in. If you made it all the way to the end, we really appreciate you sticking with us. And uh, if you can, please share this, um, whether it be with a friend or on social media, what have you. Um share it with your enemies. Share with your enemies. Actually, yeah, share it most with your enemies. Share it like 20 times with your enemies as as much as you can on every platform. Um but yeah, so this is this has been the news for the month of October. Um surprisingly, before we hit record on this, uh there was a lot we actually had to cut. So if you want to look at the the weekly, or sorry, the monthly news document, uh, there's some more kind of tidbits in there if you want to read more. But um, with that, we'll go ahead and wrap it for this month, and uh, we will catch you next time.

Tim 39:32

Yep. Take us.