The Surprising Truth About How AI’s Role in Coding Will Reshape Developers’ Futures

Welcome And 200K Subscribers

Nir, you know, it's it's great to have you on the podcast. I uh literally like 30 minutes ago, right? I I looked and I was like, okay, I'm recording today. Who is it with? And I looked and I saw your name. I was like, wait a minute. Haven't we been trying to get this thing going for like a while? And I looked back, it was like the beginning of December, you know, like I think you got in touch with me actually like right when I cut off interviewing anyone else for the rest of the year, because I was just so I always have to take a break, you know. And this year it was like December and January. I just completely took a break, didn't do anything with the podcast. So I apologize for the delay, but I'm really excited for our conversation at least.

Likewise, you see, we ended up uh making it, so the that's great.

So yeah, it's just a crazy time, right? Like I I made a LinkedIn post earlier today where you know the podcast hit like 200,000 subscribers, right? And I started this thing five years ago, and you know, my goal when I started it five years ago was literally just that, you know, I would get 10 listeners, right? If I get 10 listeners, that's a success, right? And then I started to, you know, kind of like play the mind games with myself and say, like, okay, well, let's look up some stats around this, right? And something like 98% of podcasts don't make it to episode 10, right? So then I was just like, all right, well, let's get to episode 10, you know, and then it was like, okay, like only 0.1 hit like 20, you know, like whatever it was, you know. And whatever that bar was, I just always thought to myself, well, that's easy. Like, I could get there, you know. But the journey's the journey's been crazy. I mean, I just had to like send out like 50 emails saying, you know, I'm booked for like the next six months. It's stupid.

Yeah, listen, it's it's great that you always can set up a few goals for the next thing and the next thing and the next thing, and when you hit the goal, you just kind of no uh, you know, it was it was a nice goal for this time, but I actually have you know bigger aspirations and set the next and the next goal, and that's how you grow eventually. If you don't set goals, yeah, you don't know where where you will end. Yeah. Definitely not in the place where the you know, whether you were setting the goals. Right.

Yeah, it it's I heard I think it was Jim Rohn talk about this before, where you either set the goals for your life or you you either chart your own path or someone's gonna do it for you. And if someone else is doing it for you, you're not gonna get the most out of out of your life. You're not gonna make the most money, you're not gonna do the most things, you're not gonna explore, you know, your your passions and whatnot. And uh that that really hit home to me. I heard that when I was in college. Oh, geez, 14 years ago at this point, right? And kind of like really shifted my whole mindset, you know, and then fast forward, I was working for a large company. Someone w someone on my team was there for like almost 30 years, you know. The guy was like getting ready to retire. You know, whenever I would talk to him, like he knew the environment inside and out and whatnot, but you could just tell he wasn't like excited about what he did. You know, he was just there because he was there for 30 years, and that was it, you know. And I then I asked him about what he planned on doing when he retired, which was like a couple years away. And his face was lit up and said like he's gonna travel the world, he's gonna spend, you know, three months in Japan and you know, all this stuff, right? Like amazing stuff. Stuff that I want to do in my life. And, you know, like I didn't say this to him, you know, because I didn't want to like crash his dreams or anything like that. But like, what if something happens before now and then and you didn't go on your trip to Japan, you know? Like, how important is this at the end of the day? You know, like you've already met your your financial goals, like I know that you did, but you're not fulfilling the other side of you know your own wants and needs and dreams, you know. Like, I don't know how we went down this path, but like I just think of I think of that, you know, because so many people say all the time they they convince themselves, oh I can't do that. I'm not

Goal Setting And Not Waiting

I'm not smart enough. I don't have a name, I don't have a personal brand, I don't have X, Y, and Z, right? Like whatever it is. At the end of the day, I didn't start this thing with a personal brand, right? This created my personal brand. So like I started from nothing, you know. My first episode got like a thousand views, right? And then I went on a drought for three years, got like a hundred views, like at most, you know. But you know, I say all of that to encourage people to get started, I guess.

Mindset is the thing that eventually differentiates, you know, the best from the rest, essentially, right? It's it's if you work, you know, for 30 years in a specific company, it's fine if you meet other goals that you have in life. And if you're fine just doing your job day to day and still being able to achieve other goals, maybe like traveling uh here and there, having maybe more time with a family, it depends what is the priority, right? And if your priority is to travel and your job allows you to travel, I don't know, a month in a year or two months in a year and work remotely, awesome. Maybe that's the best job for you. And stick around for 30 years, but you can achieve other goals. And so I think it's really it depends on on the kind of case by case and the goals that each person sets for for himself.

Yeah. Yeah, absolutely. So, you know, how did how did you get started in the space? Like what does that journey, you know, look like? Like what was there something that interested you in the in the area or field? Or, you know, what what does that path look like for you?

So I actually so I was interested in technology since I was very young, up to the point in which it was, I think, the age of eight when I wrote, I don't know if we can call it code, but logo. Logo was a language that you can instruct, you know, a cursor what to do, right? And I got really excited about that. It was that or playing games, and I liked that part. Okay. And later on, my parents sent me to the Tel Aviv University. I grew up in Israel most of my life, and it was a visual basic course, and they taught me how to develop games, which obviously I couldn't care less about games. But the game that I developed uh messed up a little bit with the file system in the lab. And long story short, they kicked me out because it was a prank, but it was a prank that messed up the entire lab. No one could start the computers. So the, you know, they kicked me out. That's kind of how I realized that this is what I want to do. And then it was just pranking a lot of friends with uh different security things, like you know, the top sevens of the world and so on. And eventually it became more professional when I started my army service, and after the army service, I immediately started with a cybersecurity job, consulting, and consulting is great. It opens you up to see how companies operate in different types of businesses, different types of risks and so on. And then later on, you know, after being in a few of those consulting companies in different roles, mainly in like IT security and application security. I was a penetration tester at some point. And by the way, on the pen testing side, I actually have two CVEs on my name. Back then it was not it wasn't that common to open up a CVE. And uh later, later on, just made some progress and moved into more of the management type of roles where I headed OPSEC for uh NCR, the National Cash Register, then you know CISO at a fintech and and so on and so on until I became founder of a cybersecurity company, the thing that I like to do.

That's interesting. Okay, so you said that you went to university before you went into the military. Is that is that normal in Israel?

It wasn't it wasn't a degree. It was more of like a boot camp. Yeah, I didn't get a degree before I mean I I did a degree I did get the degree at a reasonable age. I did it right after the army service, but the the boot camps, I I could do as many boot camps as I wanted. It didn't contribute anything to my bachelor's.

Yeah. Okay. I guess that makes sense. So you were essentially just like taking some courses there or whatnot to like kind of upskill and whatnot. Yeah, I I'm always curious because like I actually have on a lot of people from Israel, and you know, all of them, like every single one, was like in the 8200 group, and none of them can ever talk to talk to me about anything that they did. You know, they just be like, Yeah, I I typed on a keyboard, you know, and that was it, right? Which is like it's like the biggest like downfall for me, because it's like, man, I have this really interesting person on, and he can't even tell me the interesting stuff, you know? You know what I mean? But yeah, I've just never heard of people like going to university before their military service. That's why that's why I asked. It's a it's an interesting like differentiation, I guess.

There's a program like this in certain areas you can get an approval from the Army in Israel to go and get your degree before the the Army service. So it postpones your army service by, you know, let's say three years. But then you when you start your service, you start it as like in a fast-paced process to get into being an officer. Like within months, you become an officer. Wow. And so so it and it fits very specific types of profiles. So for example, in like getting a computer science degree is something that you can do before you go to the army, and then you will serve in one of the technology units, most likely. Because they invested by waiting for you. Now they can use it. Right.

Yeah, that makes sense. It's uh it's interesting to me talking to people that like were in the military, either the US or Israeli military, talking about like, you know, the officer rank and whatnot. Because like as a civilian, I I know nothing about any of the ranks. I know you go in and you come out some other rank, hopefully, and like that's it, right? But like I I remember when I was trying to get into the military, and everyone, everyone that I talked to was like, oh yeah, like you you go in, go to officer candidate school, you know, you'll be an officer. And I'm like, that doesn't like I don't even know what that means. Like, what does that mean? Like I I go to I become an officer and like I'm in charge of someone now. And like I was I was like talking to the Navy and they're like, Well, yeah, you're in charge of like 2,000 people because you'd be like in charge of a ship. And like, oh, well that that sounds interesting. That sounds like a promotion, you know?

Yeah, I mean, listen, it's I I I can definitely tell that the the army shapes you. Uh you know, in Israel it's it's a mandatory service, so you you need to go, you need to contribute to the country because it's fairly small, and obviously, you know, its location in the Middle East kind of requires the protection. This is why 8200 exists, this is why you know the Air Force has so much technology built in in Israel. This is why every piece of this survival country depends on on this workforce. Yeah.

No, that it actually makes a lot of sense, you know, and granted, like I've never been to Israel or anything. I would like to go sometime, but you know, like recently, right, with with America attacking Iran and everything, like I wanted to just like look at the size of Israel just for you know comparison to Iran, right? And I was I was looking because like it always just seemed like, and of course it's it's a big deal, it's not try I'm not trying to like downplay it or anything like that. It seems like whenever you mention the two nations, immediately like the next word or sentence is nuke. And so I was thinking to myself, like, because I'm used to America, and so if a nuke drops in Chicago, like, yeah, I'm dead. You know, the other the other states are like still intact, they're there, you know, as long as a hundred more don't come. And so I looked and then I was like, the size of Israel is

From Pranks To Pen Testing

like significantly smaller than you know, Iran and like everything else around it, basically. And I was like, well, that's really interesting. And then you know, you go to like a website and it like simulates you know the s the the size of impact for like a nuclear weapon, right? And it's like it's like one of the smallest ones takes up like would impact like 90% of the country, you know? And immediately I was like, oh, I get it a little bit more now, you know, because it's like, yeah, like if that were to ever happen to that nation, like and they're done, you know? Like that's a pretty crazy scenario to be in that I think a lot of Americans never even realize. Because, like, I mean, like, literally, I could fly from one coast to the other coast, and it is an eight and a half hour long flight, non-stop, right? Or I could fly from Chicago to Europe eight and a half hours. Like, think about that. That's insane. I just traversed half of my country, an entire ocean, and now I'm in Germany eight and a half hours in the air, and that same exact flight time is me going across my country. Like, that's insane.

You know, so we have four time zones, right? Well, it depends how you look at that, but let's say four. And like if you look at Russia, for example, you will see they have, I think, ten or eleven time zones. Yeah. That's crazy. Yeah.

Like there's a lot of room. I wonder how long it takes to like actually fly across Russia. Probably a lot longer than eight hours.

Yeah, I can check where I have claudia.

That's actually really interesting. I'll have to look that up. But but yeah, you can you tell me like maybe what you specialized in when you were in the military? AT and security. Okay. Yeah, I I had someone on that like told me uh reverse engineering one time, and I was like, oh, you probably told me too much. Like, like, because that was like the most I've ever gotten out of someone before. It was kind of funny.

Yeah, but I I was not in 8200, I was in the Israeli Air Force.

Oh, okay.

A different service.

Okay. Gotcha. Yeah, I've I'm I'm sure like everyone else, right? I I've been watching a lot more videos on like pilots and what they go through and all that sort of stuff. Not saying you were a pilot, but you know, just looking into that a little bit more and then seeing like how the F-35 performs against the you know opposing jets is just like, how is like how are we even utilizing physics right there? Like that is insane, you know, to see it perform like that, it's just absolutely amazing. But uh, but yeah, I mean, we went off on like literally a 20-minute tangent right there. You know, I I wanted to ask you, because you know, before the call, right, you were talking about like AI coding and you know, the the kind of like the automation side of AI, which I guess, you know, people would argue is the entire you know precipice of AI. What's your thoughts on you know these AI coding solutions like Claude, right, taking over like software engineers and devs roles, like completely eliminating them. And the reason why I ask is because I'm very much a pessimist on this, right? Because I immediately go and say, like, well, yeah, Anthropic CEO, every six months he has to re-up his claim of six months and you know, every dev is gone, right? And then sure enough, AWS goes and has a massive outage because they had some AI bot create some code in their environment and deployed it into prod and it brought you know down their DNS solution, right? So it it just feels like we're in a state of like uncertainty, right? Because at some point in time, the devs are gonna be wrong and Claude's gonna be right, and now we're in a world of hurt. What's your thoughts on that?

Think about it this way. You know, before World War II, an alarm was a job. Now it's in the phone, a computer was a mathematician's job. Well, now we know it's a hard one, right? I think that the developer's job will change with AI, but I think that the skill set that developers have will be highly needed and even boosted with AI. Because at the end of the day, maybe developers are married a lot to their code if they're not using AI. And those that are using AI are less married to like how pretty it looks like. Does it work? Does it scale? As long as it answers the questions of the intent of what they want to achieve, it's good. And I think that what we'll see is that developers' role changes to more of the guidance of what needs to be built and how things need to be built. Just you know, an interesting anecdote. I I, you know, I don't write code for Arnica, but I know how to prompt Cursor and Claude, and we use Cursor Claude and Copilot, and they're gonna have a lot in Arnica. And at some point, I just need to know how to prompt, or I just need to know how to build the software that I want to build, or what do I want to change? For example, we can we can have a customer asking for a feature and someone posts it on Slack. It's real use case in Arnica. We get some of the social engineers posting something on Slack, saying, hey, we have this feature request from the customer. And the next response in the thread is at cursor, build this with those and those and those guardrails. And it just goes and builds it. So so as long as you know how to prompt it and you know how to set up the environment and set up the guardrails in the right place, there is a an art in it that you can master to be the next level, the super it can be the super dev, the super QA. I think that's pretty much where it goes. Yeah. And at the same time, there are some challenges. So if you think about the world and where it goes, you can see that roughly today it's roughly 60 to 70 percent of the new code is generated with AI. And the expectation is that it will be closer to 90% by uh 2030. So there's still some room, but a lot of code is already generated by AI. There's also there are also a few pieces of research that show that uh AI actually generates roughly 40 to 45 percent more bugs. Bugs can be security or non-security bugs. It doesn't matter. It can be a vulnerability, it can be just a logical bug. And the thing is that you have those bugs and it will happen because AI is used to repeat existing behaviors in code repos. So if you have a specific like vulnerability, it will likely repeat the vulnerability. If it has a specific logical bug,

Israel Service And Why It Matters

it will repeat the same logical bug. And then because you have more issues, you require more reviews, more thorough reviews of the code before you ship it into production. And that's the bottom line today. So so developers are less focused on writing code, they're focused more on reviewing the code. And the review of the code is also problematic. Because not only that you you you have high cognitive load on code reviews, and you can spend very limited time on this because of that reason. Now you get to the point in which more than 60% of the issues flagged by tools, AI code review can be security and non-security, more than 60% of those tools, comments actually end up being dismissed by developers, either bypassed or just don't care about them. And and that's huge because we think the developers get more productive. Not really, because the tools actually make us less productive. And and that's that's the problem that we're facing today. And everything that I just mentioned to you happens across multiple AI coding tools. Meaning, you know, developer in typical companies you will see two plus uh AI coding tools because every engineering team selects their own. So how do you govern this? You can say, in one place I will set up certain rules and settings, and other tools I won't, because I don't know how to operate this, or I don't even know that the tool even exists. Yeah. The engineering team procured something for themselves. And can be even personal accounts the developers are using. You know, I bought my 20 bucks subscription with Claude, and that's it.

Yeah, no, that's a really good point, you know, and I always like tell people I'm not a developer by by any stretch, you know, of the means, right? Like I I've tried to I've taken like you know, Python courses, I don't know, five times, and I just haven't finished them. And the problem with coding is that like if you don't have a project, you know, to keep you going, you kind of lose those skills because there's a it's very much a diminishable skill set. And if you if you're not using it, you know, if you're not using it regularly, you also straight up lose it. I mean, just there's no other way to to go around it, I think. And so for me, like I'll I'll use I I'm a heavy user of Grok heavily. And so like I use grok to you know code out different things and whatnot. And you're you're right though. You you have to get really good at prompting it to tell it exactly what you want, and it'll give me something. Like now I've built in the context and the different projects that I'm working on, but you know, I would have to say, like, all right, review it, analyze it, make it better, make recommendations. You know, and it would make all the recommendations, review it, change different things all the time, you know, and when it stops like making those recommendations, now I know, like, okay, we're in a pretty good place. This is pretty solid, which is pretty impressive, to be completely honest with you. You know, I'm getting my dissertation in something, and a part of it is actually running like a PyTorch model on my system. And so I like I absolutely hate Windows for basically everything other than gaming. The only reason why I have Windows installed on my desktop is because I also game on my PC, and like that is literally it. If I didn't have to game or if Linux would pick up their gaming, like I would switch today, you know, because I hate Windows so much. And so, like, with that, you know, like I need to basically utilize my GPU. And so I'm not gonna deal with a Windows CMD, you know, bash prompt, trying to freaking get this, you know, this cluster up and running to run the PyTorch model and all this stuff. So I just spun up WSL2 with Ubuntu, launched it, and I actually just followed all of Grok's steps, right? Like it gave me like 50 steps, literally broke it down like I'm a 10-year-old, right? Told me everything. I followed it. I mean, I didn't the only time I ran into issues was when I didn't read what grok was telling me to do. Like when I deviated from his path, that's when I started failing. If you can't copy and paste it, it's not a real clinic you need to do.

It's just a bunch of war, it's in between.

Right. I mean, at least like for me, you know, I feel like I understand what's going on. Like I'm I'm well versed with Linux, and you know, I was basically like a Linux admin earlier on in my career, right? So like I understand Linux really well. It's literally like the coding part that's like, yeah, I just haven't learned that skill, you know, at this point to that extent, right? And so that's where like Grok actually makes up that difference. It's like, okay, here you go. And if I run into an error, I give the error back to Grok, you know, I've had to do this before, and it'll say, Oh, you know what, this line we messed up, this is the modification, right? And it goes through and like I'm learning as I go too. Like, I'm not just kind of just you know, copying and pasting and not understanding what's going on. That'd be a great way for you know anyone to exploit my system then. So like I I am definitely learning, but it a hundred percent bridges a skill gap that like exists. I feel like it probably exists in a lot more IT personnel than than you would expect, even.

Gap.

Okay.

The gap necessarily needs to be filled with security. Think about it this way. Let's say that you use in a company, I don't know, like uh Claude and I mean cursor and a claw. Okay. And each of those ecosystems have a like a markdown that you can provide to this, like a s a skill or a markdown that could be claw.md and cursor rules

AI Coding And Developer Futures

and uh agents.md, a bunch of files that you can set up. And thing is you can provide it with the base instructions that uh you want to be repeated every single time before code is uh generated. So for example, you can provide it with a prompt that says write code security by default. Now, not in these exact words. We actually have a prompt in our product that we can inject to all of the repo in the company and it's a special workflow. And and that prompt is roughly 8k tokens. But it's 8k tokens that you can justify why you followed OWASP ASVS best practices and uh how does it actually respect the code changes, such as, yeah, I want you to make uh every piece of code secure by default, but if I already have authentication, for example, in my app and it's not as secure as it should be, don't change it because it will break my app. Okay, things like that. So obviously it's very thoughtful on how it's prompted. And and then if you can if you can inject this or any other requirements, like write tests with every new feature. Whatever the thing that you want, you can actually bridge that gap because developers, for example, don't need to be security experts. But what they need to do is to know that security is checked because they already have the rules. Now, it's not enough to have the rules because certain patterns will be repeated, right? Like, I don't know, I have this problem with my authorization mechanism that every single time a scanner will find this because of the way that it's implemented and can't be changed. But you want to have this in in multiple steps in the development lifecycle. That I think just by doing that at a prompt level, closes most of this gap. I think that can Grok, I don't know, maybe I need to check. Not sure if Grok supports agents.md, but it must have something that that becomes.

I'm not sure if it does. You know, I I know that they're coming out with like their standalone Grok code product, so I wouldn't be surprised if it's getting built into there. But recently, you know, when you even just ask it any any kind of prompt, if it has to think deep at all, you can see like it, it's like, all right, agent one is doing this, agent two is doing this, agent three, like it just starts spinning down, you know, all of the agents that are doing different things and coalesces that response into one response to you.

Yes. By the way, I you know, it's nice that you mentioned that agents are operating differently. I I can tell you, I mean, I route, I I write multiple, you know, apps for different needs. It can be ops, can be scripts, can be something that makes my life easier. And and I my my latest trend is actually asking whatever. Sometimes I have, I want to do it with cursor, sometimes I do it with plot, sometimes I do it with codex, whatever I have the mood for. Whatever the model looks to be the best at the this time. And the first thing that I say almost every single time is that, hey, you're a product manager, these are my requirements. Break it down to a PRD, which is the product uh requirements document. And for anything that is vague, don't assume anything, ask me. So it interviews me, it takes me roughly 30 minutes until like there is a full SPAC doc. And then my next command would be okay, so now your program manager estimates which resources do you need to develop this PRD. So it gives you, I need the this QA, and I need the developer and the security expert, blah, blah, blah, blah, blah. Okay, awesome. Build them, those agents. So it builds those agents. And then you say, okay, build an orchestrator that will be able to run those agents. So it builds the orchestrator. And then the next thing that you say is run. And it goes and built it. And it takes time, you can leave your computer working, you know, come back in a few hours. You'll see something working. And at the same time, you can say revalidate, retest, ensure that it meets those strict requirements. All of it. And it just works. You just need to be patient with the multi-steps that you're willing to take to build a workforce that will write software for you.

Yeah. Yeah, no, that that makes a lot of sense. And I think I feel like that's the that might be like the root of a lot of the issues that we see, you know, with like agentic AI coding is that people aren't prompting it or challenging it enough, you know, to say like, okay, we'll make this better, or find the flaws in it, or you know, make sure that the rules were run and everything. Because like, you know, I I uh, you know, like I said, I'm getting my PhD, right? So when I first started to like actually do my research, I pulled up Google and Google was completely useless. I mean, it was giving me stuff that like wasn't relevant, I couldn't find any articles or anything. And I'm sitting here like, okay, either I'm in a brand new field that doesn't exist right now, or Google is stupid. And so I went to ChatGPT, and ChatGPT would just like lie to me constantly, you know, and I was literally about to like drop out of the program because I'm like, if I can't research this thing, like what am I even doing? You know? And so then I went to Grok, and Grok was actually really good at finding articles, and it would hallucinate here or there, definitely less than ChatGPT at the time. I mean, I'm sure someone will like, you know, destroy me, ChatGPT 5.2 is like the greatest thing ever or whatever. But at the time, ChatGPT would hallucinate like 90% of the conversation. And Grok was a lot more efficient. And once I built in the prompts though, and that context memory of saying like every single resource that you give me has to be validated, you have to read the entire article. If it doesn't correlate to these topics, you cannot refer it to me. It also has to exist, it has to be publicly available. If I cannot download the PDF version of the file, it does not count. Like, and it as soon as I built that into it, you know, through trial and error, like it, I could ask it anything now, and it'll it'll give me like the proper response. I know that like what I'm referencing now, like is the actual thing, which I'm sure if my PhD chair is like hearing, he probably just fell out of his chair, you know. But it's like, hey, we're in 2026. This is a cybersecurity program. What do you think I'm gonna be using under the hood to like validate things? Right? Like, I mean, what are we talking about right now, you know? So hopefully I don't get kicked out. That would suck. But that's not part of the audience. Yeah, that one guy. Yeah. No, I I'm sure that they probably wouldn't even mind, you know. We all went through education, right? So as long as the as long as the scanner doesn't catch it, it's fine. Which, you know, it's not like I have groc write my whole paper or anything like that. You know, it's literally giving me guidance. Yeah, this is a crazy conversation that's probably gonna get me in a lot of trouble.

But oh well. Listen, I mean, that that's that's the world we we live in today, right? You need to think differently about the the workflows. There's a reason why you see the SaaS uh market going down and a little bit up and a little bit down. Because eventually, in a world in which a prompt is

Code Review Fatigue And Governance

not a note anymore. You can do so much. You can let's say you have an incumbent, you will come up with a new process, a new company, it doesn't matter what it is. And by the time that you started implemented yours, someone already thinks about how to replace you. So so there is a stream of five to seven years until you know a product can become an incumbent and maybe not keep keep up with the pace. Same with the process, right? Like learning, right? So you will always have something that will innovate and faster, especially with with AI. And that also brings the topic of what will actually be viable later. Like the fact that I build apps, is that sustainable? I don't know. Maybe it's cheaper for me to use Calendly than develop a Calendly and you know 300 bucks of a prompt. Right. They will maintain it, right? Or if I wanted to build a linear, yeah, it's a prompt away. Linear maintain the product and they have you know uh features uh that they develop and improve and so on and so on. So so I I think that we will all kind of get back to well, there are there's a SaaS product that it just doesn't make sense for me to develop. And maybe the competition will be companies against those as opposed to individuals trying to chase.

Yeah, maybe you know, really it sounds like what you're describing is that like the competition is like leveling out, right? Because like if you look at you know these SaaS apps, you know, not to name any of them, but if you look at the different SaaS apps and how much they charge and whatnot, right? It used to be reasonable for them to charge that much, even though it was a begrudging amount. You know, you kind of like begrudgingly pay, you know, your sales solution every month or every year, because it's like, okay, I absolutely need it, they do it, you know. So I I have to do it, right? But now maybe the competition is is leveling out where it's like, hey, the only option is no longer to go with the incumbent. I could potentially build this plug-in myself and integrate it into the tech stack and maybe get the same result in a different way, right? And saves me a whole bunch of money. So theoretically, that would force the SaaS product pricing to go down to kind of keep up with that level of innovation, if that makes sense. It depends.

It depends on there's so many dependencies here, but I'll I'll give you an example. Let's say that you use a cloud code review, they just publish their pricing. It's 15 to 20 bucks per pull request. That's pretty expensive. The numbers add up very quickly.

Yeah, how many times can you pull? How many times can you pull from GitHub? Like that's that's crazy.

It's crazy, right? But but this is why, you know, so so despite the the the value that ETN provides, which I believe it's huge value to scan every pull request, the cost is not necessarily something that makes sense, but this is the cost to run for them. Right? So, but on the flip side, if you if you have a company that does optimizations on what shouldn't be prompted, when, how should it be run, maybe you don't need to scan all of these pieces because you already scanned them and you'll you'll scan the subset of the code because you store the context for the rest. All of these optimizations, which by the way we do in Arnica, we see that you know, roughly three bucks, like in like good sized company, you know, three bucks is is what we see today for a PR that we scan. So so eventually there is room. I mean, you can say, well, you know, cloud code will replace everyone, but what about managing a backlog? Well, now you need to ensure that things that you merged into production, like how do you take care of them later? It doesn't happen. How do you deploy it across the company? How do you make sure that you have reasonable consistency of results? All of the enterprise level questions that come with scale and process are outside of, let's say, a foundation model company. But on the flip side, everyone uses like not everyone use foundation model company products because developing an AI model is very expensive, right? So so we'll see more usage of the foundation companies. They will come up with products that seem to replace the SaaS industry, I guess. Bottom line is just every company will need to analyze where does it actually make sense to build versus buy. And there's just gonna be way more appetite towards build when there are very specific problems with a very limited set of use cases.

Yeah. Yeah, absolutely. I feel like we're going into like an evolution stage rather than like, you know, just straight up like job elimination and whatnot. I I I think that we're in an interesting place because it almost feels like the skills that used to be extremely difficult to get, right? Thinking about a lawyer, becoming a doctor,

Secure Prompts And Guardrails

stuff like that. I feel like those are more prone to be taken over by AI quicker than you know, a dev or an engineer, right? Because like we're kind of working more in tune with that technology. Working, we're working with it more. A lawyer, in some respects, is working against it, or a doctor is almost working against it, right? Where you have to be more accurate than the machine. It's like, well, eventually that machine is gonna beat you, like no matter what, you know. And I so I I I actually have like my own personal experience with it to an extent, and I'm sure people are gonna just like you know, rake me over the coals for saying this. But last year, I actually got some pretty extensive blood work done. And it was like through a clinic or whatever, because my normal physician, my general physician or whatever, just wasn't giving me, you know, the the results or the feedback that I wanted or needed. He was giving it, he was basically treating symptoms and not the root cause. And he was only going to make things worse, whether he wanted to believe it or not. Granted, if I would have gone through this process that I'm about to explain and they give me the same response, it's like, okay, there's no questioning it at that point. But I was doubting my physician. So I went to a clinic, got a whole bunch of blood work done, took the results, put it into Grok, and I said, What are your recommendations? Right. And Grok literally gave me, you know, top like five recommendations. It's like, here, here's the five things, right? And a part of the service that I use, like you meet with a doctor, they review the blood work with you, you know, all that sort of stuff, right? I meet with the doctor, the doctor gives me the exact same five recommendations that Grok gave me, right? Which differs from my general physician. Completely differs from it, right? And I was struggling with like, you know, high cholesterol, high triglycerides, high blood pressure, you know, normal stuff as you age, I guess like normal quote unquote, right? Shouldn't be normal, but it is, apparently. And uh, you know, as soon as I just followed the recommendations, everything went back to normal for the first time in 12 years. Everything was back to normal, like 100%, right? Which, you know, like it put me into an interesting mindset because now it's like, well, wait a minute, like these doctors, like they might actually be offset by AI. Because I mean, literally, Drock told me the exact things, like it gave me the right dosage and everything. It was like, I need to know your weight and your height for me to be able to make this, you know, the dosage recommendation. Gave it to him, gave me the dosage, right? It was the exact same dosage that the other doctor was going to give me. I'm just I'm just sitting here like, man, these guys are kind of screwed at this point. And the same thing with the lawyer, right? For the podcast, like to like create contracts or whatever it might be. I'm just going to grok and being like, this is what I'm using, make it better. Right? And it seems to do a pretty good job.

I think we need to think about it as as a an accountability term.

Build Vs Buy And AI Accountability

If you need someone accountable, yeah. That's a good point.

That's it. Just do your thing with the eye. I I did my taxes. I wrote well, not I, but the the agent wrote another validation.

Yeah.

For itself, and it ran its own validation. So I have six eyes, quote unquote, looped in this. Okay, you know, I'm good. I'm feeling that I can be accountable for that.

Yeah. Imagine if you had to pay six accountants.

Exactly.

You know? Exactly. Six accountants to do your taxes. Like that would be insane. But you're given that same you're getting that same level of quality. I feel like taxes too is another one. Because at the end of the day, it's literally reading the law, looking at numbers, and applying it, you know, to whatever to whatever limits there are, and giving you a number back, right? Like it it'll even give you the form to fill out and walk you through how to fill it out. Like I was thinking about this literally yesterday where I'm like, man, I don't want to go to my accountant on Thursday. Like, can I just have AI do this for me? You know, like that was my actual thought.