Shimon Tolts (Copperhelm): Hello. Okay, Shimon Tolts, co founder, CEO of Copperhelm, Claude Ambassador, that's recent. CNCF
Joe: nice.
Shimon Tolts (Copperhelm): Ambassador, AWS Hero, that's the gist of it.
Joe: Cool. Well, Shimon, I I really appreciate you coming on the podcast. You know, I I this thing has been in the making for a bit and I'm really excited for our conversation today. I think it's gonna be, you know, really relevant and timely for how everything is just getting so crazy for professionals in security, you know.
Shimon Tolts (Copperhelm): Absolutely happy to be here. Thank you very much.
Joe: Yeah, absolutely. So why don't we start with telling, your background. How how did you get into the space? What made you want to go down the security path? What what does that look like for you?
Shimon Tolts (Copperhelm): Yeah, so the gist of it. I I I'm like this PC kid. I started when I was like twelve years old. I installed my first Linux, ran to the living room, and then my mother was like, What? Like yeah, we'll find the cure or something. I don't know what you did. And yeah. Fast forward I worked for McAfee, I worked for Unity as a general manager of infrastructure and really got fascinated with The intersection of infrastructure and security. I'm also very, very active within the community. in 2012, I've s me and some other folks established an AWS user group. We have more than 12,000 members and I've been awarded as an AWS hero. And now I do the same for Claude. And I'm a Claude ambassador and I do videos and so on. So really it chose me. It's not like I chose it and I really, really love it. And I think it's not by chance that we focus on the area of cloud security.
Joe: Yeah. Yeah. when I started to get into IT, cloud was really in its infancy. And I remember, you know, reading up on it and saying, like, okay, well, this is where the future is going. Like, I need to learn this thing, you know, just cloud overall, right? And started with AWS and started to dive into it, saw how flexible it actually is, and I I I I've been I've been stuck with it ever since, for good or bad, right? but cloud security now really poses a a I don't know if it's a unique a unique issue per se, but it's amplified an issue that already existed within the cloud because really cloud security previously was kind of a mess. To some extent. you had to get your your guardrails in place. You had to really build things out properly. But once you did, it was very doable. It was very manageable moving forward, right? Well, now with AI becoming a bigger topic and a bigger attack plane is really what it is. th those boundaries, those guardrails that you once had are being pushed to their limits. They're being, stretched a little bit farther than maybe what they were designed to do in the first place. And so now as cloud security professionals, we're in a situation where the environments are are so big that we kind of need them to fix and remediate themselves rather than rather than a security person being in the middle, slowing down the process, having a change board and all this other stuff, you know, like it just doesn't really work like that anymore.
Shimon Tolts (Copperhelm): Absolutely, if we look at you know at the at the history, right? So cloud security was built around prioritization, which was the general thing, and like you would work your way from it was started with just severity. And then they were like, Okay, but maybe it's like C VSS nine, but maybe it's not really being used. So they added EPSS, so like how how popular is this? And then exploit intelligence, is it being used and so on. and we really were used to to this state of mind of prioritization as a strategy. And and what we assumed that we had a window. And we had a window in which we could actually fix the things. So you would get it, you do some prioritization, you work with your engineering teams, and and you really go with the the window. And the backlog sort of like reflected like what defenders could postpone because Attackers are unlikely to pursue it, right? So it's like if I have some low hanging, low C VSS and so on, it will take time. And I think that everything changed in 2026 this year. It's crazy to think that at the beginning of the year, still most people were writing code like this, and and this window collapsed. And I think that there are like several Several causes to it. number one, obviously the capabilities the AI can develop exploits, and I'm proud to say that I did it using JLM and Kimi and other like mainly Chinese models that have no guardrails. So basically, every CV has an exploit today.
Joe: And it's like immediately exploitable too. there's no delay. it when it's found and identified, there's automatically an exploit, basically being attached to it and already running across the internet. that's the that's the crazy thing. You know, because like I remember I don't know, I I I remember when what was it? It was like the blue wasn't blue origin, but it was the tool set from the NSA, right? And everyone started freaking out saying like, this is gonna be weaponized and everything else. Right. And it eventually was weaponized, but it wasn't weaponized overnight. I mean, it took like a couple of months for it to really be weaponized, and then we saw some advanced attacks with it, and the rest of the security stack. caught up and adjusted, right? Because the legacy security stack, I'm it's crazy to say I'm saying legacy security stack, right? But there was a delay, right? There was a learning period where it would catch up. It may be a day, maybe an hour, maybe a month, right? Whatever that looks like. But it takes a little bit of time for these products to learn from it, adjust, and then provide that security. they do with all the other attacks on the market. But now we're in we're in a state where not just is it faster, but the bar of entry is so much lower. I
Shimon Tolts (Copperhelm): Lower.
Joe: when I was getting into security, right, I was looking at different career paths like offensive security and malware engineer. And when I was doing my masters, I took a class on reverse engineering malware. Right. So I got access to a malware database and I was downloading different malware on sandboxes, infected myself too many times to actually admit, you know, and went and and started reverse engineering them, right? Like kind of learning just doing it myself. And
Shimon Tolts (Copperhelm): And today we have REA which has two hundred thousand stars
Joe: Right.
Shimon Tolts (Copperhelm): and does it for you.
Joe: Right. Right. and that's the thing was like I was learning how to do it like manually. I was learning how to do it myself, y And now that barrier to entry is so extremely low. And again, right, in that class, I only spent 16 weeks doing it. I would not say I have a skill set in it by any means, right? I know how to plug it into a tool and have a tool like step through it. Right. But where that barrier to entry used to be, it no longer exists, right? Because AI just does it for you.
Shimon Tolts (Copperhelm): And I wonder, I wonder if this is the same as Python and JavaScript developers don't know how to do alloc, malloc, free in C. And like will it be the same thing? Or like are we the last generations? Are we the last coders? Are we the last like people who actually know what happens? And and are we going to be the future COBOL developers? The only ones that actually know what to do with the mainframe? I don't know, maybe?
Joe: Well this is this is gonna be the problem and it's probably almost like an existential security risk, I I would say, right? And I I don't want I don't wanna like play off of Sam Altman's and you know the Claude Founders. I I wouldn't call it fear monitoring, right? But they're saying like, hey, like there's a risk with AI actually being a little bit more powerful and strong. than what we expect and what we're designing it for and whatnot, right? they're threading the needle, so to speak, on capability and it being able to like build itself. but think about it in terms of this, right? With AI now coding, you know, I I don't know, let's just call it 80%, right? 80% of code is created by AI in some way, shape, or form. Right now, it takes someone with a reasonable skill set to look at it and say, you know what, you need to adjust this or you need to remove this or modify this in that way. Right. you still need that skill set to some extent. And I'll be honest with you, I really only apply that after I see the code run and I'm saying, hey, that didn't run right. you need to adjust this. I'm not looking at the code itself, which is a different skill set. And I know devs that that do. But we're going into a place where you don't have to look at it anymore. So if you don't look at it and you're no longer having that skill set of reading the code and understanding what's going on, who's to say that the AI isn't putting, malicious things into the code that's now being run aclo across the globe in what was supposed to be a benign way? You are trying to achieve some task, whatever it might be, and now Claude decided, or OpenAI or Grok or whatever other model you want to choose, decided, you know what, I'm gonna self-replicate myself. And maybe they just put in a little stream and it starts going from there, right? Like that's something that it's something that like you wouldn't think to catch and un and unless you're reading every line and understanding every line, you're not gonna catch it. Right. So like we're kind of going into a a gray area where we're losing the skill set, like you said, of coding and understanding it and writing it ourselves. And now we're relying on this technology that we're being warned about could become a little bit more powerful than what anyone is expecting or intending.
Shimon Tolts (Copperhelm): Absolutely. So, going back to the cloud CVEs, so now they have the capability, we have the volume, like never ending amount of CVEs, we have the autonomy, like you said, they you can spawn spin up agents and they do it for you. And we have the cost, which is going down constantly, and you can like if you used to have an attacker go look for the top ten low hanging fruit and then move to the other target, you can just go like For 100% of all of the exposures that you have. And a fun fact, or maybe I would say interesting statistic, is that FedRamp has changed their severity-based timelines and they've actually shrinked the SLA for taking care of like what they they changed the name also of a like a critical severity, they changed it to pain five, and they changed it from having to remediate it in 30 days to 12 hours.
Joe: Huh my gosh.
Shimon Tolts (Copperhelm): From 30 days to 12 hours. Now there's like you can't like you gotta ha fight fire with fire. Like you can't do this manually. So we can see that that even the industry and and FedRAM is catching up and there is no other way than like going autonomous remediation. Now it's a spectrum. And we we can talk about like how do you do this? I believe in crawl walk run. And I think that currently it's more of a people process problem than a technology problem. because we work with many enterprises and Fortune five hundred companies. It takes time and it's also a state of mind to to to allow something to automatically apply WAF rules. Or do network segmentations and so on.
Joe: Yeah. That's pretty crazy that it's moving its recommendation or moved its recommendation to twelve hours, right? And I can I can count on one hand the amount of companies that I've been at that have had twenty four seven capabilities that would be able to fit that twelve hour time frame, right? you think about it, most people are starting their day at like what, eight o'clock, they're ending it by four or five, right? Companies typically outsource overnight stuff, you know, to another company. And the overnight stuff is supposed to be like, hey, I needed like this is bad enough for you to call me, right? Like that that sort of thing. Not hey, I need you to actually work on my systems and whatnot. Like th those are two one, those are two very different like skews skews for for these companies, right? To be able to actually have them operating in your environment for that amount of time. But two, I I I mean it's a difficult thing to pull off. Like I said, like there's not that many companies that even have that that capability, that twenty four by seven by three sixty five capability. I mean, know everyone, you know, kind of says that they do
Shimon Tolts (Copperhelm): Think if you're a FedRamp, if you paid $1 million to be FedRamp, then
Joe: Mm.
Shimon Tolts (Copperhelm): I I assume you're a big company and like also to Yeah. Yeah, yeah. You you need
Joe: Is that a requirement of FedRAM? 'Cause I haven't looked I haven't looked at it. So is that a requirement?
Shimon Tolts (Copperhelm): to pay like around $1 million and you need to have a sponsor, one of the government agencies, and and then I think it's like the one million dollars gets back in the contract or something like that. But but still you you gotta
Joe: But
Shimon Tolts (Copperhelm): Yeah it's not a three person startup.
Joe: but as a part of FedRamp is the twenty four by seven coverage like required? Like is that a requirement? So like you already see it up front.
Shimon Tolts (Copperhelm): Well I think it's
Joe: implied.
Shimon Tolts (Copperhelm): like as as you say, in order to be able to do it, you need to and and and also you know it's it's also has to be US employees sorry, US residents
Joe: Mm. Yeah.
Shimon Tolts (Copperhelm): in the US. So like you can't like India it or Australia or like follow the sun. You can't have the model, so so I think
Joe: I gotcha. So the bar is the bar is higher. The standards are are higher in terms of basically everything on the back end.
Shimon Tolts (Copperhelm): Yeah, but if you run your software with access to the FBI systems or the CIA system.
Joe: Yeah.
Shimon Tolts (Copperhelm): Yeah, it could also be just Department of Health, but still also important. so I think really the tectonic shift here is companies used to do like severity based prioritization, and I think that I would say prioritization is dead. You can do this on the thumbnail, because you can't do this in a statistical model anymore. So, what we see with companies is that we come in, we connect to their scanner, we connect to their cloud assets, and then our AI agents go over all of the CVEs one by one, and they don't care about EPSS or severity or anything. They go and they do two main verifications. Number one is understand whether the asset is reachable from the internet. And this includes internal, external, probing. It includes looking at compensating controls on the way and it's really building this map. And the second thing is doing a runtime analysis, which we do by leveraging native APIs like Amazon SSM or Azure Managed Run Command. So it's really easy to install. And then we're actually able to connect to the workload and see, okay. The scanner says that I have this Apache version 5 that has this critical vulnerability. But the fact that it is sitting on the disk doesn't mean that it's actually running. It doesn't mean that it's actually bound a socket to a port. And it doesn't mean that it is configured in a way that is exploitable. For example, it has a certain configuration toggle or a certain toggle on the ENV flags. Or just in general. maybe it's a package like a React to shell, maybe it exists in package.json, but no one ever imported it into code. So the first step is like a reduction of noise by doing something that today was done either by humans in two ways. Either they would actually go and do those checks manually, or they would just say, you know what, I'll just upgrade instead of like Checking whether it meets, I'll just upgrade it and that's it. But now the volume is so huge and the window is so short, you gotta do this automatically. And the second part, which you you choose which one we zoom into. The second part is how do I apply how do we get to remediation? And I think the way to get there is by using mitigation, which is compensating controls, like. network segmentation using security groups, applying WAF rules, applying XDR policies, and really doing anything you can that is still within the security team's realm and can happen immediately to stop the bleed.
Joe: Yeah, so really applying the best practices broadly, right, as as secure as it possibly can be, essentially, right from the very beginning, to buy you time to actually remediate these items within your environment. Is that kind of what you're thinking?
Shimon Tolts (Copperhelm): Yeah, but it's not best practices. We tailor make a policy. For example, let's take WAF. We create a custom WAF rule that is relevant for this for the for the specific CVE on the specific workload within the asset within the cloud environment. So think of it as a tailored rule. Or think of us enabling EDR policy that does EBPF. in network capturing and and looks for the payload of the exploit. Now this cannot be done because you have ten million vulnerabilities, you can't enable 10 million well not 10 million but even 10 10,000 CVE protections because your your network is gonna die and your CPU is gonna die. But if you know how to apply the relevant policy and say those two policies on this workload and those five policies on that workload and those two WAF rules on this server, then all of a sudden you can tailor the right protection on the right asset.
Joe: Hmm. Okay. That is interesting. you know, at at at like a former role, I actually was tasked with deploying the AWS WAF to over a hundred a AWS accounts. And it was difficult. It was difficult to say the least, you know, creating like kind of a uniform template of okay, this is how it should be getting to. deployed in a default manner throughout the environment and then working with all the different devs to tailor tailor make or create these WAF policies specifically for their application. it took me 18 months. Easily 18 months.
Shimon Tolts (Copperhelm): Two to deploy the WAFs or also to put the policies?
Joe: To put in the policies. the deploying the WAF was the quick part, that was probably a month worth of work and that was us being pretty careful and whatnot. But actually tuning it and, you know, conforming it to the applications and whatnot was just brutal. I mean, it was weekly calls with these devs.
Shimon Tolts (Copperhelm): And and think of that I I guess I guess what you did is like apply OS top ten rules and SQL injections and stuff like that. Now
Joe: Yeah.
Shimon Tolts (Copperhelm): think of a world where you have a thousand new CVs every day and you need to to apply the rules for the relevant CVs. It's you can't
Joe: Right.
Shimon Tolts (Copperhelm): do this manually.
Joe: And that was that was even at the very beginning of Chad GPT, there were were no AI agents or anything like that that we were concerned about in the world at that point. So like the the old architecture and it's four years old, right? I'm talking about something that is literally maybe three years old and
Shimon Tolts (Copperhelm): Yeah.
Joe: I'm calling it old. at this
Shimon Tolts (Copperhelm): Yep.
Joe: point because it wouldn't stand up to AI agents and you know all the different issues that are coming out right now. Like it just wouldn't h hold up. It's not flexible enough.
Shimon Tolts (Copperhelm): It's crazy, but today I think there's the pre AI, post AI agents world and and that's it. Like today, I don't know when we're gonna stream this episode, but as of today Entropics released like they stopped using Git because they're like they broke git.
Joe: Yeah.
Shimon Tolts (Copperhelm): And they're like this is like made for a different era. Like we build our own git because we do like thirty thousand changes a day, it's like on a monorepo that's like that's not gonna work.
Joe: D weren't in the beginning, I mean, you know, granted, right, like I'm I'm working on my dissertation, so there's a lot of information and news that like just completely surpasses me and you know, I kinda like come up for air once a month and am like overwhelmed with the updates of the month and whatnot, right? But wasn't there a time, I think it was like last year, where like Git was down because there was too many like commits to it and whatnot, and it was because of the AI agents, not just these companies were now making insane, pushes every day, but now everyone else was making all these commits that probably never would have beforehand. is that a thing or am I imagining it?
Shimon Tolts (Copperhelm): well if we look at the GitHub status a history I don't wanna like bash on anything in the podcast but like that's that's a known issue for some time now. And now you could debate like why and how and what are the reasons, but that's like a known problem. that GitHub has become a if you think about it, it's sort of a single source of truth. It like it holds all the source codes and and a lot of packages. Now of course for production systems and companies have like JFrog's Artifactory and stuff like that, but Everything's getting to its limits. I
Joe: Yeah.
Shimon Tolts (Copperhelm): think there is going to be a total replatforming. I don't think my daughter's gonna use GitHub, unfortunately. She's three years old, but you know. I I think this is like d the great replatforming.
Joe: Yeah. Hmm. It's interesting. I I've also been looking at other options outside of GitHub for for me to move to in twenty twenty seven. and I I don't know. I I don't know if there's any good options, you know, like it it's on my list. I didn't think about it very much more than that. But, you know, like you said, like GitHub is kind of I don't know, it's past its prime, so to speak.
Shimon Tolts (Copperhelm): And it's it's it's not only GitHub and I don't wanna I I just think that the the the rate and the the way we code is very, very, very different now. So I don't
Joe: Well, the architecture. The architecture of it is the antiquated part that doesn't keep up with what we're talking
Shimon Tolts (Copperhelm): Yeah.
Joe: about.
Shimon Tolts (Copperhelm): The the Git protocol. Yeah, the Git protocol. I I think that another major major problem that we see is which is also interesting fact, I I made a research and I saw that when we look at the global available CVEs, we see that only of s officially there's only twenty-five percent of them are stated as having a public exploit. And roughly like roughly twenty to twenty five percent. And one of the main issues that I see is that because the scanners and a lot of the security companies everything's lagging. Like the the world is changing fast, but all the so one of them the main things that They prioritize, and again, my statement is prioritization is dead, you gotta validate everything. One thing they is very important to them, they say is this weaponized, which is a cool way of saying, Does this have a public exploit? And I think that it's a major problem because 25% is not true. My friends, go get Kimi or JLM and have a fun at it. And and I think we should really look at it as one hundred percent. Now I I don't envy the companies because I understand. all of a sudden, you you you're fighting with the business all the time anyway, so they will fix your stuff. And now it's not only that they you you're you're fighting them and because they wanna, release features, make money. Now all of a sudden it's like, no, now you gotta fix seventy percent more. And you're like, are you kidding me? this security tax. I I just wanna r run my code, build build my business. so we we gotta we gotta wait and see. Like I was at an event with Jason, the deputy CISO of Anthropic, and he said like the Vonpocalypse is here. And he says he thinks that the next two years is like the apocalypse. And you gotta fight fire with fire. And The interesting thing is I asked him, so Jason, why do you do it anthropic? And he said, Well, we do have autonomous remediation. We spin up a separate environment, we test all the patches, but then a human merges them. And I'm like, Really? Entropic?
Joe: Hmm. Hmm.
Shimon Tolts (Copperhelm): If you're still merging things manually, okay, like you know. So
Joe: That's interesting. Did did he have a good reason for it?
Shimon Tolts (Copperhelm): Yeah, they still what a human in the loop and trust verification. But I don't know, when when I met with him it was only like Mythos and Fable. Now Opus five point five is amazing. By the way, my company were part of the Glasswing project, so we got Mythos, which is cool. but really I think that Opus five point five is really changing the game now. And we'll see. We have meta. A Gemini four a was released, we'll see. It just continues growing.
Joe: Where do you think where do you think the just like the AI sector itself goes in twenty twenty seven? Right. And I'm thinking in terms of, you know, a couple of years ago, LLMs were basically all that we had, right? which was essentially like the next generation of search engines. And then, you know, twenty twenty five, we get AI agents. Now we have a whole bunch of harnesses that we can, work with those agents within. What do you think comes next? What's that next iteration of this space that people should be, keeping an eye out for? Have you heard anything?
Shimon Tolts (Copperhelm): Well the the the name of the game today is the loops, right? So they say like you have a loop and it just does a version, monitors, improves the version, monitors improves the version. I have my own small version of this. I have a use case which is I have gazillion emails and So I have routines that that run in in in every four hours, they categorize it, they catalog it into they put tags they credit to archive, they create drafts for for new ones, then they go to my AI recorder, they get the context of the inform of the of the call, they write it in, then they go to calendly, they get my my open slots, they offer the next meeting and so on. And then I I have the the creator of the routines. And and I find myself every couple of days, I tell it go over the routines, check how they worked, and improve them. And and I'm like today I'm like sort of like, shouldn't I have like a routine that enables the routines to improve the routines? And
Joe: Mm-hmm.
Shimon Tolts (Copperhelm): and and that's like it, you know, that that's like the
Joe: Yeah.
Shimon Tolts (Copperhelm): Because it learns my usage, it looks what it did, it looks what was my reaction to what it did and it
Joe: What what I always hear about loops and I'm hear I hear about loops from Claude and OpenAI and Grok employees, right, that have unlimited tokens to use. What does a token hit look like with a loop? obviously it would depend upon the task and whatnot, but have you set loops that burn through all your token usage in a short amount of time, or is that not a thing? Does it get more efficient over time or something?
Shimon Tolts (Copperhelm): I think they do get more efficient. Again, Opus 5.5, I love it. I I think that like I used to put the effort too high. I would like max and extra and ultra core and stuff like that. And today I see that like medium and high is enough. And but I think that we we have the gatekeepers of of of the bottles, which is like a sort of like the five-hour window. So even if you burn something, it's not like you're gonna burn everything. And so there's like people love to hate the five hours, but in a way it's like a stop-loss mechanism. but I'll switch hats from Shimon, the CEO and the cloud ambassador, to to like Shimon's company. So in our company we use Amazon bedrock, for example, to run the agents to connect their customers and so on. There there is no five hour window that that's like it's r just running
Joe: Mm.
Shimon Tolts (Copperhelm): using API. So you gotta use the observability stack and the long chain and long graph and then all the stack and and and and like leverage bedrock or whatever, like agent core, whatever you're using to to help you do that because once it's it's not like a user subscription, but it's just like the API side, and then you know. The sky is the limit or or your bank account is the bottom, I guess.
Joe: Hmm. Yeah. That that is like a very real issue and risk with with AI. You know, like I was looking at it yesterday and just thinking of like, okay, well, I need to get all this stuff done, you know, this week. Like you, I now have an agent that just like manages my inbox because my inbox is just completely impossible to stay on top of. Like it, you know, I I I feel embarrassed, but like for a long time, my inbox would just get inundated with podcast guest requests and speaking opportunities and everything else, right? And it would get buried. And I would get to it four or five, six weeks later, and I'd have to apologize, like, hey, I'm sorry I got to this so late, you know, like that sort of thing. And it's not a good look, but when you're just getting started, I mean I've been doing this for five years. I still feel like I'm just getting started. You don't have You don't have the bandwidth for a real, assistant. And y you know, for a little bit of time I was like considering having, you know, my wife be my assistant. But, you know, then then like Grokbot came a became a thing and I like, All right, well let's just try this out before I start paying her, you know, to be my actual assistant. And unfortunately it took her her theoretical job, I guess.
Shimon Tolts (Copperhelm): I I can't see myself really doing stuff now without it.
Joe: literally just last night I like my new website went live, right? told Grockbot,
Shimon Tolts (Copperhelm): Nice.
Joe: here's the code base, review it, make all these fixes, pushed a whole bunch of changes to GitHub.
Shimon Tolts (Copperhelm): Why why grogbot? Why grogbot and not claude or open AI? what's why grogbot?
Joe: Because so I was already a a big Grok user. and then Grokbot came along and just made agents like so much more easy to use, so much more consumable. You know, the connectors are very easy to use to connect it to everything. it's on my phone, it's on all of my devices. And yeah, like I I was using Hermes before, connected it to Claude, and that was fine and all, but I I just feel like Grokbot is I don't know, one maturity level above Hermes in terms of ease of use. And so I I don't know. I view myself as a technical person and I still leaned more towards, Grokbot at the end of the day.
Shimon Tolts (Copperhelm): So ease of use, ease of integration. Do you use instinct?
Joe: Yeah. Yeah. Integration, no.
Shimon Tolts (Copperhelm): Yeah, so I think that's the it less control but that's the easiest one. d that's what you do for your mom. you connect it or you
Joe: Mm. okay.
Shimon Tolts (Copperhelm): tell her just talk to him. Just he he can do stuff for you.
Joe: Hmm. That's interesting. Yeah, I I never I haven't thought about that. Yeah. I would definitely need something like that for my mom for sure. but hopefully she never gets into AI.
Shimon Tolts (Copperhelm): I gotta tell you personal story. My dad, he's 76 years old. He's a PhD at the university. He's a statistics and demographics academic scholar, right? And he've been wanting to write a book for 20 years. And he has like a lot of publications that did. And you know, for 20 years, like, yeah, I wanna write the book, I wanna write the book, but I don't know the editor and this and that. And now he I He's part of my family Google account and so he has Gemini Pro and stuff. And now he's writing the book together with Gemini. And he's like he already ate re wrote nine chapters and my dad's 76, yes?
Joe: Yeah.
Shimon Tolts (Copperhelm): It's crazy. I was like and I see like it actually happening, so it's crazy.
Joe: Yeah. No, I I was just talking to some friends of mine. We're you know, we started this group chat as like an investment group chat during the pandemic. and now just recently it's basically turned into like business management with AI group chat basically. And you know, a a side job that I still do to this day is create educational courses and content online. I don't really talk about it that much, and that's gonna change, but it's something that I've done for the last five or six years at this point. And it used to take me probably about four to six months, start to finish, you know, when the content goes live. Because you're creating slides, you're recording it, you're editing, you know, all this stuff. Like it's a lot of work. It's a it's a whole lot of work. And now, like with with Grock and Claude working together, right? it gets done start to finish in a month rather than, you know, six, right? And the reason why it takes a month is because I'm slow with doing it, right? Like if we're talking about like actual time, it's like, you know, twelve hours or something like that, right? but it's insane because it it just I don't know, it it it makes everything so much easier. You don't have to rely on your own willpower anymore to to get things done that you used to have to.
Shimon Tolts (Copperhelm): I agree. So before like we we we run out of time, so
Joe: Yeah.
Shimon Tolts (Copperhelm): so what's your take on on the cloud security and vulnerability space? Like as as an expert that works like in the security field, like do do you think the market is adopting it fast enough? Do you think there's not gonna be any other way? Do you think we're gonna have the major, major incident, like the log for J of the thing and then it's gonna change? Like what do you think?
Joe: Yeah, you know, I'm I'm sure a lot of security professionals are gonna hate me for saying this, right? But all the big hacks, all the big exploits lately at least, they tend to come out around the holiday season, and we're going right into the holiday season right now. and so, you know, I I think we're in a place where security teams are no longer able to keep up, right? Like We caught up to cloud security and what to do there and everything else like that, but AI just completely throws all of that out the out the window, right? And so we're more reliant on solutions to keep up with these threat actors and exploits more more now than ever. And to be honest with you, there isn't very many solutions that can keep up. The ones that claim that they do, they They typically, have some funny money or funny numbers, in their in their statistics and whatnot. And when you start digging into it, it doesn't quite hold up like you would have thought.
Shimon Tolts (Copperhelm): that are doing a lot of innovation now, so
Joe: yeah, for sure. there's definitely a lot of startups, and there's some that are sponsors of the of this podcast that, you know, really they're they're pushing the way forward of how to actually secure your environment by default without thinking about it, really. so it's really and I don't want to say it's shifting left. It's like almost like shifting up out of our hands into the solution that we're choosing. You know, whatever solution we choose for our environment. has to not just be able to grow with our environment, but it has to be able to keep up with the threat actors that have access to these models that are generating all these exploits. I think that that's that's probably what we're looking at as the next iteration of the security tech stack. Shimon, like I I know we're at the top of our time here. I really appreciate you coming on and I really enjoyed our conversation. It was a great it was a great conversation.
Shimon Tolts (Copperhelm): It was great being here. Thank you so much. And it will be interesting for me to look back at this in one year from now. And we'll see how it aged. Because when I think of one year from today, be like going back, I would have never imagined this is what would happen. So I'm very interested in seeing how the future unfolds. Thank you so much for having me. It was wonderful being here.
Joe: Yeah, absolutely. Likewise. Well, thanks everyone. Make sure that you check out Shimon and his company at the links below in the description of this episode. I hope you really enjoyed this unfiltered conversation about the state of the market with AI security. Thanks everyone.