Transcript: Balancing Innovation and Safety: CTOs Check-In About AI Evolution
0:00 Scott With all these things people are creating that aren't being created by professional software engineers or being created by people in the marketing teams and finance teams,
0:07 Scott what are the concerns of a professional software engineer that maybe they aren't aware of?
0:30 Scott Welcome to another episode of How Many CTOs Does It Take?
0:35 Scott My name is Scott Porad and I'm here with my trusty co-host.
0:38 Brad Brad Hefta-Gaub.
0:39 Scott Brad,
0:40 Scott how are you doing?
0:41 Brad I am doing great.
0:43 Brad I've been busy.
0:44 Brad And it's been busy and fun.
0:47 Scott So today's episode is
0:49 Brad Yeah.
0:49 Scott a what's on your mind episode because
0:51 Brad All right.
0:52 Scott it's been a while since we just talked about what's on our minds. What has been on your mind lately?
0:57 Brad I've let's see one of the things that I've been really busy with lately and been on my mind a lot is security. And I've got, you know, I've got multiple clients that I work with and they're all.
1:11 Brad all going through various different stages of either preparing to have to really worry about security or having to worry about security or and I personally I just I don't know if you saw this but there was just a amazing like exploit that got released as part of a standard LLM LLM mini LLM pack
1:36 Brad package that there's a python package that a ton of people use and somebody did add a full secrets uh exploit uh if you have installed this thing now fortunately i don't install this thing i don't have this thing installed none of the teams i work with have it installed but it was like literally getting environment variables secrets github secrets it was going after crypto wallets and it's just like
2:01 Brad Boom,
2:01 Brad that kind of stuff is happening all over the place.
2:04 Scott And.
2:06 Scott I've heard about this. It did not affect me or the company I'm working for, but was this a malicious attack or was this the LLM going off on its own and being rogue?
2:17 Brad No,
2:18 Brad this was a intentional malicious supply chain attack by somebody who knew what they were doing.
2:24 Brad And it was very hard. Somebody found it. It was very well obscured.
2:29 Scott We had an incident last week as well.
2:32 Scott as well and I wonder I actually don't know what it was because our we CrowdStrike cap caught it for us in just a matter of seconds and it was also a supply chain issue I wonder if it was the same thing okay
2:44 Brad Yeah.
2:45 Brad Well,
2:45 Brad anyway,
2:45 Brad so I would say security and AI and packages and supply chain,
2:50 Brad that's on my mind. I'll tell you that.
2:52 Scott it's interesting because that is on my mind also but in a different way but I want to ask you a question first before we explore that
2:58 Scott Do you remember about 2001 or so Microsoft like called a code red and like for six months they just went on like security and we got to fix all of our bugs security
3:09 Brad Yeah,
3:09 Scott holes.
3:10 Brad right.
3:10 Scott Do you remember that?
3:11 Brad Sure.
3:11 Scott I have a very uncomfortable feeling that.
3:15 Scott We are about to head into another era like that of just security vulnerabilities all over the place.
3:22 Brad This speaks to the other sort of things, and I posted this on LinkedIn.
3:25 Brad I basically posted something on LinkedIn the other day where I was like, hey,
3:29 Brad two credential theft stories that are kind of related and one's easy and one's hard. And the easy one is this exploit that kind of popped this week.
3:41 Brad But the other one is I am convinced that all of the LLM data like Anthropik and OpenAI is just littered with secret keys and the
3:54 Scott Yeah.
3:55 Brad following evidence I have for it.
3:58 Brad I was setting up a package.
4:01 Brad It was a basic scaffolding for a React Native web app and it was it's open source and it was really well done with respect to it.
4:09 Brad All of it had no secrets in the code and it had a very clear read me on how to set up all your secrets, you know, to either install them as part of your GitHub secrets or use a secrets manager,
4:19 Brad set them up in your environment variables. But like the code was very much like.
4:23 Brad you know, secret agnostic.
4:25 Brad I asked Quadco to,
4:26 Brad hey,
4:27 Brad go set this up for me and configure it and get the basic framework up and running.
4:31 Brad And it was running away and I looked over and I noticed that it was generating database passwords and generating API keys for various services like, you know, Supabase and Expo and all these basic tools that this framework was based on.
4:48 Brad and it was literally passing these secrets through the llm in clear text and i had to go stop stop stop stop stop no no don't do that don't do that but by default it was basically just passing secrets in clear text and that means that it's sitting in a log file somewhere in anthropic you know and if anthropic ever gets hacked all that stuff is going to get exposed like i don't think i don't i doubt that anthropic would ever violate my secrets
5:15 Brad secrets but if they ever get hacked all those secrets are going to get exposed
5:19 Scott How did the LM get the secrets?
5:20 Scott It read them from something you had told it to
5:22 Brad oh
5:22 Scott set up?
5:22 Brad the setup and the setup instructions were like configure super base so you know super base is a you know online uh database platform so you know super base command line basically says create create database and it like you have to create a you have to set your own password when you create the database and so it went off and
5:41 Brad ran a you know ran a cryptography library to come up with a random password took its value and use that to go create the the database right
5:50 Scott Interesting.
5:50 Brad if you were the sysadmin right you would have done that manually you would have been like oh i need to set up this database i'm going to go create a secret i'm going to go store my secret in my secrets manager i'm going to go set up the database and and at no point would you have been passing you're not emailing those to your devops team saying hey here's the password i
6:08 Brad route I want you to use on the database you're setting up please yeah
6:10 Scott Right. You use some sort of disappearing kind of thing to
6:14 Brad exactly
6:15 Scott pass those around.
6:16 Scott Interesting.
6:17 Scott Okay,
6:17 Scott so here's the thing I had recently happen that's been on my mind.
6:21 Scott We had a false alarm where it appeared that using Lovable,
6:26 Scott somebody at our company had exposed all of our company's data by plugging it into our data warehouse.
6:31 Scott Now between lovable and getting to the data warehouse,
6:34 Scott there was a page where you had to authenticate with Okta and so it wasn't a risk.
6:37 Scott But at first it was like,
6:39 Scott oh,
6:40 Scott it's really,
6:40 Scott really, really, really, really, really bad.
6:42 Brad Yeah. In other words, like the fear in the moment was somebody minted some kind of a token that has enough credentials to read our database and that token was put up into Lovable. So Lovable's got.
6:58 Brad It's got access to that token,
6:59 Scott Or,
6:59 Brad something like
7:00 Scott I
7:00 Brad that.
7:00 Scott don't know, I don't know, we use Snowflake, so maybe Lullable has some Snowflake connector so you can like just,
7:05 Scott I don't know,
7:06 Scott I don't know what the answer is.
7:07 Scott But okay,
7:08 Scott the point is this,
7:08 Scott we've got company values.
7:11 Scott And we've got two values that are in conflict at the moment.
7:15 Scott And I'm going to abbreviate these one, one value is obsess over customers,
7:20 Scott and the other is move fast.
7:23 Scott Okay.
7:23 Brad Yeah.
7:23 Scott And so the tip of the spear on the AI stuff is saying, hey,
7:27 Scott our values say move fast.
7:29 Brad Yeah.
7:29 Scott And the InfoSec guys say,
7:31 Scott hey, our values say don't lose our customer data.
7:33 Brad Yeah.
7:34 Scott And as the CTO, I feel stuck in the middle.
7:37 Scott I actually don't know. I'm trying to figure out how to navigate this. On the one hand, yes,
7:42 Scott I absolutely want people to move fast, experiment with these things, figure out how to do stuff with.
7:48 Scott stuff with it. The landscape and the marketplace is changing and we got, you know, I want people using these tools and figuring it out.
7:54 Scott That's the tip of the spear move fast crew.
7:56 Scott On the other hand, I think about it this way.
7:59 Scott With all these things people are creating that aren't being created by professional software engineers,
8:04 Scott they're being created by people in the marketing teams and finance teams.
8:08 Scott What are the concerns of a professional software engineer that maybe they aren't aware of? Stability and scalability.
8:14 Scott and performance and cost and those
8:17 Brad Sure.
8:17 Scott kinds of things and security and all of those things are two-way doors with the exception of security meaning if you're not scaling or if it's not stable or if it's costing so too much you can either shut it down or you can throw people at it to try and figure it out and make it better
8:32 Scott But that security is a one-way door. Once your customer data goes out the door,
8:37 Scott it's never coming back.
8:37 Brad Yep.
8:38 Scott And I'm just trying to figure,
8:41 Scott I'm just in the midst of it trying to think about and navigate,
8:44 Scott how are we going to handle this in our company?
8:47 Brad Yeah. And I guess that's part of the point behind my story, right?
8:51 Brad Like, I
8:52 Scott Yeah.
8:52 Brad think I know a thing or two about security.
8:56 Scott Yeah.
8:56 Brad And it was a demo app of a framework that and so
8:59 Brad And so it doesn't matter, like there's no customer data in it.
9:01 Scott Mm-hmm.
9:02 Brad So like the AI agent was,
9:04 Brad it was being professional,
9:06 Brad it was being proficient, it was doing things mostly correctly,
9:09 Brad but it was this one thing that it didn't consider about the fact that it was passing information,
9:14 Brad you know, it was treating the security in a way that it was treating it like it was in a sandbox when in fact it was sending itself back and forth.
9:25 Brad to this cloud-based LLM, right?
9:27 Brad And it was so fast and it was so subtle and it only happened because I glanced out of the corner of my eye.
9:35 Brad and watched it do this command and i was like oh shit i can't no wait did you just do that like i like i had to like rewind the tape and go wait did you just do that Mm and then
9:46 Scott -hmm.
9:46 Brad i had to basically stop it because it had done some very it had done the similar move already a couple times so i was just like fuck this whole like all my databases and all the environments that it just set up for me i gotta tear all those down because
9:59 Brad you know all the keys were basically like burned at that point it's like then i got to go and that's the funny part about that is to then tell the agent hey i still want you to set this up but don't do it in a way that you're passing the secrets then it's like the agent's got to work extra overtime to basically write scripts that it's going to run locally and not you know it's just and so you know and i talked to some of my some of the engineers on one of the teams i'm working with i talked to them about that i told them the story
10:25 Brad the story and they were like oh yeah dude like there's no way that tens of thousands of people aren't making that mistake every single day today and and they don't notice it
10:35 Scott What about the e-rotation agent you have that you'd say,
10:39 Scott hey, every night go run around our environment and change all of our keys so that we're secure.
10:45 Brad right right well
10:46 Scott Yeah,
10:46 Scott man, weird.
10:47 Scott It is weird times out there,
10:50 Scott man.
10:50 Brad if somebody if somebody okay if
10:52 Brad If somebody, if an external bad actor hacks Anthropic and gets into their logs and their telemetry, right, is able to basically see those chats or
11:02 Scott Or OpenAI or Gemini or any
11:05 Brad yeah,
11:05 Scott of them.
11:05 Brad yeah, I'm picking,
11:06 Brad I just picked one of them, right?
11:08 Brad They're going to sit on it.
11:09 Scott I guess this is where job security is,
11:11 Scott is be the human key rotator.
11:17 Brad Maybe that sounds like a grim job,
11:20 Brad but yeah.
11:21 Scott You know, I mean, I met a guy this week.
11:24 Scott This is his business, his job. He has a specialty and he fixes the chips.
11:32 Scott in marble countertops in like kitchens you know they gets easily chipped on the edge and he's got a bunch of he's a crafts person has a bunch of techniques that like he knows how to like do that which
11:42 Brad Oh, yeah, absolutely.
11:42 Scott if you have a marble i don't have a marble countertop but that is i guess they chip easily because it's just stone so you know he's found this little niche for himself and charges a lot because
11:52 Scott not
11:52 Brad Yep.
11:53 Scott a lot of people do this you know sort of do it so my point simply is i guess maybe a human key rotator is
11:59 Brad Yeah.
11:59 Scott the job of the future kids
12:00 Brad Yeah.
12:00 Scott kids if you're listening yes
12:02 Brad You were zooming the camera out and saying like you're sort of seeing this as well.
12:08 Scott it's it's a similar problem it's it's the similar category which is this is all moving so fast and there's this one and i guess it's okay to be that a lot of it moves fast
12:18 Scott fast but you've pointed out an example i've got an example of like oh actually in this area it's not okay to move fast this area and it's over 30 years as an industry because this wasn't as much of a concern it was just sort of client server local network kind of situation so like over 30 years as an industry we've developed all these techniques and practices to prevent these problems
12:38 Brad Yeah.
12:39 Scott because they were real problems and
12:41 Brad Yeah.
12:41 Scott i just think at jurassic park we're so what was the expression you know we're so focused on whether we could we didn't think about whether we should that's
12:50 Brad Yeah, right.
12:50 Scott on my mind big time and it's probably you know i mean i think we're going to see some big breaches because of it
12:57 Brad Yeah,
12:58 Brad I think I think you're right. I think that's a given.
13:00 Brad Well,
13:00 Brad by the way, recent episode that you weren't able to attend because you were.
13:04 Brad on a vacation or a business trip, but I was speaking with a security expert and,
13:10 Brad you know.
13:11 Brad Though one of the points that we ended up talking a fair amount about was the fact that it's so easy for humans to essentially become fatalistic about whether there's anything they can do about security and therefore they then like just don't bother because they're just, what can I do about it?
13:32 Brad You know, there are a combination of he made the point that there's a combination of evolutionary reasons why.
13:38 Brad is why by default we kind of we're an interdependent species we kind of have to trust each other and there is a percentage of us that are absolutely untrustworthy but the majority of us are trustworthy and we're so interdependent that we kind of default to trust but if there's only three percent of us that are psychopaths or sociopaths that's a lot of sociopaths out there yeah
13:59 Scott you don't even need three of three-tenths of a percent three-tenths of a tenth of a percent what is that i don't even know that's a
14:04 Scott thousand percent well
14:05 Brad Yeah.
14:06 Scott so like we've developed things like SOC2 so that I'm sure came in response to some breaches and things like that right do you think we're going to we're going to get the next generation of that something like that'll come along
14:19 Brad Well,
14:19 Brad it's interesting you say that as I'm working with a team to working with one of my teams to get their SOC2 compliance done.
14:25 Brad And I'm reminded by our work that we did at.
14:29 Brad health tech company when we worked together last that I don't want to say that SOC 2 is like security theater, but it can be done in such a way that it's really just kind of checking boxes.
14:41 Scott Totally.
14:41 Brad How many people out there have SOC 2 logos on there?
14:45 Brad You know, on their websites and actually are just checking boxes.
14:49 Scott I don't know, but I do know that when I have been in the situation of working towards SOC2,
14:55 Scott I have preached to my teams, the most important thing is to actually be secure.
15:00 Scott I remember having an incident,
15:01 Scott I mean, it wasn't an incident,
15:03 Scott a conversation bordering on butting of heads with the CISO at a company I was at who was asking us to do something to meet a compliance.
15:11 Scott regulation. I was really frustrated by it. And I finally said,
15:14 Scott look,
15:14 Scott I will do this as long as you will just agree that this is less secure than what we have now.
15:20 Brad Oh, my God.
15:22 Scott and and she was like yes i understand your point it is less secure and i was like okay thank you yeah
15:26 Brad That's wild.
15:27 Brad To meet some security requirement,
15:31 Brad you actually had to make it less secure.
15:33 Scott i mean we we found a way to not do that but like it was this it was this the the framework was saying do something a certain way and we had done it a different way which was better so i've always just kind of said like tried to reach the team like
15:46 Scott At the end of the day, I will go to bat with the auditor for the more secure option.
15:50 Brad Yeah,
15:50 Scott We
15:51 Brad totally.
15:51 Scott should not be,
15:51 Scott I will fight to the death on that one because I actually want it to be secure.
15:57 Scott The framework is just a guideline for how to get more secure.
16:00 Scott And by the way,
16:01 Scott it,
16:02 Scott Ascrab, it has saved our bacon recently. Well, I mentioned in this podcast,
16:06 Scott it saved our bacon this week.
16:07 Scott Framework said have malware.
16:10 Scott malware showed up malware and you think to yourself oh malware like blah blah blah malware system catches it within seconds and you're safe and
16:19 Brad Yep.
16:19 Scott that was because the framework served as a guideline for us and knock wood paid big big dividends
16:28 Brad Yeah,
16:28 Brad totally.
16:30 Scott That's on my mind. What else is on your mind?
16:32 Brad Well, here's another thing that's on my mind.
16:33 Brad I love a good automated CICD, automated,
16:41 Brad high quality, like local dev staging production.
16:45 Brad I love it when you've got all that stuff dialed because I think it just makes your engineering team go so much faster.
16:54 Scott So beautiful.
16:55 Scott It's just beautiful.
16:56 Brad And I'm reminded of this because one of the teams I'm working with, which has a really good setup,
17:03 Brad they started building a new piece of their platform.
17:06 Brad It was difficult to give themselves like dev staging and production.
17:12 Scott Mm
17:12 Brad And
17:13 Scott -hmm.
17:13 Brad so they don't quite it's like, you know, it's different enough. It's like shared services and it's like it so they don't have that. They don't have the clean automated.
17:23 Brad I mean, they have, you know, they still have a deployment system,
17:26 Brad but they don't have the ability for the developers to run local builds and they don't really have the ability. I've always told you about how this this one team I'm thinking about, they've got as soon as a PR gets made, boom,
17:38 Brad an ephemeral environment.
17:38 Brad environment kicks up and anybody that wants to go play with the full stack ephemeral environment can go play with it and test it and critique it but they've been starting to add some new services that are these shared services and there's not a clean way to kind of create that and it's slowed them down slowed them down big time this past two weeks and i'm struck by the it's
18:01 Brad Not that they're going that much slower than anyone else would.
18:04 Brad In fact, I'm sure other teams go much slower.
18:07 Brad But they were so fast with their really well-run dev ephemeral environment machine.
18:16 Brad They were so fast with that that they look like they're crawling.
18:21 Brad without
18:21 Scott yeah
18:21 Brad any is
18:22 Scott one little bit of sand in the gears why
18:24 Brad yeah pretty serious piece of sand in the gears yeah well
18:27 Scott why is that whole smooth cic ephemeral environment blah blah blah that whole stuff why is that so hard you think that would just be like out of the box we're 30 years into this that should be out of the box
18:40 Brad i mean obviously tools like
18:43 Brad Docker Compose or Kubernetes,
18:45 Brad these types of tools make it a lot easier,
18:48 Brad you know?
18:48 Scott Mm-hmm.
18:49 Brad You know, those are relatively new tools, like 10 years old,
18:54 Brad you know? They're not like, Docker yeah,
18:55 Scott Compose, let's see,
18:56 Scott I did my first Docker in 2013.
18:59 Scott That's 13 years ago.
19:01 Brad my
19:01 Scott Did my first Docker, that's a funny way of saying it, by the way.
19:05 Brad first Docker.
19:07 Scott Sidebar.
19:09 Scott I had the hardest time understanding containers when they first came out.
19:12 Scott Like I knew machines,
19:14 Scott I knew VMs and then containers.
19:15 Scott I was like, what?
19:16 Scott I don't get it.
19:17 Scott I mean, I figured it out. I mean, I was so confused at first.
19:20 Brad yeah it's a good question why isn't that just out of the box i think maybe that's i i don't know maybe it's because there's just so many i don't know you i don't know that's a good question i i think that like a lot of development teams think they don't
19:34 Brad don't need that sort like there's i think there's this at least in the startup space that i'm in right the startup space that i'm in like you're so many people are just like i just got to get it up crank crank this out really quick like the demo like they're trying to crank out the demo right and i think with like ai coding agents it's going to be even more of this where it's like you can ask the agent to build the thing and it builds like the full thing running and you're like woo i got a whole app and everything you know but then to actually
19:59 Scott Like, make it have a clean deployment to production and then to say to yourself,
20:04 Scott well,
20:04 Scott I want to have this staging environment where I can let the stuff settle before I release it on the public,
20:11 Scott you know, that's like you have to have that thought process that say that you want that and then to say, oh,
20:15 Scott I want to be able to actually like have full stacks that are.
20:21 Scott ephemeral for every feature change like i think maybe that's a pretty sophisticated thing to want yeah
20:27 Brad maybe i thought github had it or maybe microsoft i should say had it nailed for a while they had this trifecta this this three-legged stool they had visual studio code they have github and they have this thing called github codespaces and i thought they were just going to make it all just like effortless to work together and i know a few teams using it but like it doesn't seem like i thought it was
20:47 Brad I thought it was just going to take off like nobody's
20:49 Scott Yeah.
20:49 Brad business, but it doesn't seem like it did.
20:51 Scott Maybe that's just because the architectures are always so different,
20:55 Scott you know, like how many different containers do you have in your.
20:58 Scott in your cluster you know yeah
21:00 Brad But I think that's what I think I don't know I never really worked with it that closely or here's another thing on my mind
21:07 Scott yeah
21:07 Brad The kids so you and I both have kids that are in like their 20s, right?
21:12 Scott i just listened to your episode with the that you had with the david espinoza on with the ai co-host of his podcast and you guys you guys talked about the kids
21:23 Brad Boy, did we because I don't remember what I said
21:25 Brad Yeah, you were just like, hey, what are we going to do? Like if AI takes over all the jobs,
21:30 Scott the entry level jobs, then what are the kids going to do?
21:33 Brad that's well i've been thinking about it more from what are the kids going to do perspective not the job i wouldn't think you know like
21:38 Scott Yeah,
21:38 Brad i've got
21:38 Scott exactly.
21:38 Brad two
21:39 Scott What are the kids going to do?
21:40 Brad i've got two young adults in their 20s one has recently graduated college one is uh still in college and you know i saw this report recently it was pretty cool it was um a vision a data visualizer of jobs in the united states it had like a
21:58 Brad from Bureau of Labor Statistics jobs that had like, I don't know, 200 million different jobs, like actual individual jobs.
22:04 Brad It showed the number of jobs in each category and then they were grouped by all these different categories. And then it was color coded and you could have a little toggle where you could change what the colors meant. So you could have one which was like the growth of the sector. Right. And then another toggle was.
22:23 Brad the average income in the sector and so like if you did average income like you know low paying jobs were red and high paying jobs are green so like the green ones were things like finance and tech you know and those squares had sub squares you know which got more detailed and then there was another and I don't know how the author of this little visualizer got to this data exactly but um
22:47 Brad digital ai exposure you
22:49 Scott Yep.
22:49 Brad dabble that one and it was like an inverse all of a sudden tech and finance went red and things like labor went green right from red to green and so i just it's gotta be crazy hard to be 23 years old and graduating or 22 24 graduating from college right now
23:09 Scott Yeah.
23:10 Brad unless you have got a really clear thing that you really really want to do just trying to figure out what to do with all this uncertainty forget climate change forget war in the middle east forget all that stuff just the ai thing and what's that what that's doing to your head about the rest of your life i
23:26 Scott I have a there's an engineer working for me that's he's like in his mid 30s and he's just like, I don't know what I'm going to do for the rest of my life.
23:34 Brad have a friend who's my
23:36 Brad my rh he hung it up he's like yeah you know i gotta i've been doing this for 30 years it's been a great run and i just don't think i have the energy to go learn it all again he's not as excited about it like you and i've been excited about he's like i'm gonna go be a preschool teacher
23:54 Scott Yeah.
23:55 Brad i like kids and go hang out with kids all day that'll be fun so
23:58 Scott There you go.
23:58 Brad i've been thinking about that that yeah i've been thinking about not those little kids i've been thinking about these college age kids who are graduating and who their whole world is so uncertain right now
24:07 Scott Yeah.
24:08 Scott As you know, I have kids same age as yours, a little bit older,
24:12 Scott but basically in the same cohort.
24:13 Scott And yeah,
24:14 Scott it's a real concern.
24:17 Scott You know, I guess if I try to put the CTO hat on for a second, take off my.
24:21 Scott my Baron hat and CTO hat on.
24:23 Scott I think that I also, in a world where capitalism drives us to not hire junior people anymore because
24:31 Brad Mm-hmm.
24:32 Scott the robots can do the job,
24:34 Scott what are we, what's the end game there?
24:37 Scott Because you, how do people become senior?
24:40 Scott They become senior and experienced by first being inexperienced and junior.
24:44 Scott So if we're not hiring any new, new talent.
24:47 Scott then eventually all the old talent ages out is the end game that we think the AI will eventually reach that level and they'll be taking those jobs too and so we don't need any more senior engineers we don't need any more architects we don't need any more principal engineers because we're not going to we're not going to grow them anymore if we're not hiring junior people
25:06 Brad So I've worked on consumer products for my whole career.
25:09 Brad I mean, I worked on consumer products. I worked on some B2B businesses,
25:14 Brad but they're always sort of been consumer products.
25:16 Brad But that's relatively high level.
25:19 Brad There still needs to be people to work on lower,
25:22 Brad you know, there's still people working at Anthropik and there's still people working at the layers below Anthropik or I don't know, working on GPUs or whatever and figuring out how to make these data centers. So maybe.
25:33 Brad actually the tech jobs actually just get pushed lower in the stack the tech jobs have been moving up and up and up the stack for my whole career and now they're like at the very top it's like oh you just talked to whisper what's this whisper thing called now whisper uh i can't remember you want to talk about oh
25:50 Scott No.
25:52 Brad I haven't tried it yet.
25:53 Brad I just heard about it this week. It's basically Siri,
25:56 Scott Uh
25:56 Brad you
25:56 Scott -huh.
25:56 Brad know, but it's a new version of Siri called Whisper. It's not by Apple.
26:02 Scott Whisper. Okay.
26:03 Scott Whisper. Whisper flow.
26:04 Brad Whisper Flow.
26:05 Scott Yes. Think it, speak it.
26:06 Brad It's just a much, much better like voice thing.
26:10 Brad And so now the programming has moved all the way up to where you're just talking to it. You're just saying, hey, Alexa,
26:15 Brad can you write me a fantasy football app?
26:18 Scott Right.
26:18 Brad to share with my friends i'm not making this up in the last 24 hours i've gotten two apps from two different friends one of them wrote an app for like baseball another wrote an app for golf they're little fantasy sports games and then these two one of them is a former computer programmer one of them has nothing to do with computer programming so now they're just saying hey alexa write me an app
26:41 Brad for my friends and I to play fantasy baseball right that's like we moved all the way up at the top of the stack but maybe the real tech jobs are actually going down the stack because somewhat something has to be underneath all this stuff or maybe the robots figuring that out are the robots going to write the chips like how's that all going to work Well, I mean
26:59 Scott okay,
27:00 Scott I would say like I'll say
27:03 Scott I agree that there's people with computer science degrees that really understand the kind of low level things like compilers and operating systems and kind of like really understand architecture.
27:14 Scott I think those people's jobs are more secure for longer.
27:19 Scott I think that's a given.
27:20 Scott But there are a lot fewer people in those jobs than people who are described as software developers.
27:26 Brad That's
27:26 Scott Right.
27:27 Brad right.
27:27 Scott I mean, I can't count the number of times.
27:29 Scott times in my career that I've had been working in organizations where I mean there's always people that are much more skilled and people that are less skilled.
27:40 Brad Mm
27:40 Scott And,
27:40 Brad -hmm.
27:40 Scott you know, you might recall during the time that you and I worked together at HealthTechCo,
27:45 Scott you know, it was a seller's market,
27:48 Scott like engineers were jumping ship all over the place. And there was just like this massive demand for engineers during that kind of that COVID boom and things.
27:58 Scott And I think that there were a lot of people that got hired as engineers who were not as talented as.
28:05 Scott the engineers that, you know, are the ones building these low level things, right?
28:10 Brad Mm
28:10 Scott Like,
28:10 Brad -hmm,
28:10 Scott I think, you know, we got people that were kind of coming out of code, you know, code boot camps and getting
28:15 Brad mm
28:15 Scott landing
28:15 Brad -hmm.
28:15 Scott jobs as, you know, full stack software engineers.
28:19 Scott That's the stuff that the AI agents can do.
28:21 Scott trivially now right so those people certainly are either going to have to rapidly up level their skills and that might be really hard or those those jobs are for sure going to get washed out that
28:33 Brad Yeah. Maybe there's just many fewer jobs in the tech industry,
28:37 Brad but they're much more technically demanding.
28:39 Scott would be my that i'm i'm an over on that outcome right i think that i think that's the more likely outcome
28:45 Brad But that doesn't account for.
28:47 Brad If you were interested in going into accounting,
28:50 Brad or if you were interested in going into any other of the 75% of the jobs in the economy that AI is going to get to.
28:58 Scott Yeah,
28:58 Scott I mean, I think there's a...
29:00 Scott bookkeepers, those jobs are easily replaced and accountants are less likely to be replaced,
29:06 Scott you know, and you know, there's like a difference there, right?
29:09 Brad A little bit. Yeah.
29:10 Brad I mean,
29:10 Brad I just use that as one example, but it seems like there are so many jobs that are going to be affected,
29:15 Brad so many industries that are going to be affected.
29:19 Scott Yeah, I agree.
29:20 Scott I agree.
29:20 Brad Hey,
29:21 Brad you know what's not going to be effective?
29:22 Brad I see the guitar on the wall behind you.
29:25 Brad Guitar repairman.
29:28 Scott Well,
29:29 Scott I guess so.
29:30 Brad Poor women.
29:31 Brad I can say repairman because I recently took a guitar to get repaired and it was all dudes.
29:38 Scott Ever tell you the story about that? I had to take my Ampeon one time back when I was in my regular playing days and my amp had all these like punk rock stickers all over it.
29:50 Scott And one of the stickers on my amp was from a band called.
29:56 Scott you know we'll have to keep the keep this clean but but i i won't be able to say the true band name but it was the band name was fs up right oh
30:04 Brad We swear on this podcast all the
30:06 Scott yeah
30:06 Brad time, by the way.
30:06 Scott okay the band name was fuck shit up and it was a sticker on my amp that said fuck shit up
30:12 Scott and i took my amp in to get repaired and they had to basically like unscrew something to get to the inside of the amp and and uh had to cut through my sticker but i wasn't there when i dropped my amp off and so when i got back i got my amp back they had like carefully with like an exacto knife cut just enough of the sticker to be able to open my amp up and then
30:37 Scott they repaired it as nice as they could and i was just like you could have just taken that sticker off and they were like well when i hadn't seen i didn't see who you were but you bring an amp here in with us with a sticker that says fuck shit up and i didn't want you to i didn't want you to fuck my shit up for ruining your your sticker
30:56 Brad It's a pretty punk rock sticker.
30:58 Scott Yeah, there you go.
30:59 Brad Well, we've digressed today,
31:01 Scott We
31:01 Brad Brad.
31:01 Scott have digressed.
31:02 Scott But there you go.
31:03 Scott We covered the rise and fall of security thanks to AI,
31:08 Scott the rise and fall of the economy thanks to AI,
31:10 Scott and the rise and fall of guitar amps thanks to punk rock stickers.
31:15 Brad That's right.
31:16 Brad All right. Well,
31:17 Brad with that, let's call it a day.
31:18 Scott Let's call it a day and another episode of...
31:22 Scott How many CTOs and what do we say?
31:23 Brad Smash that subscribe button rate
31:25 Scott Rate and review.
31:25 Brad review see you next week.
31:27 Scott See you next week, Scott.
31:29 Brad Bye
31:30 Scott This episode of How Many CTOs Does It Take Podcast was edited by Saffron Eptigal.
31:35 Scott Our theme song was AI-generated by Brad.
31:39 Scott Please follow us on all the social media platforms at at HowManyCTOsPod.
31:44 Scott You can find us on YouTube,
31:46 Scott LinkedIn,
31:47 Scott Instagram,
31:47 Scott Facebook,
31:48 Scott Threads, X,
31:49 Scott TikTok,
31:49 Scott and Patreon.
31:50 Scott Or at our website at HowManyCTOsPod.com.
31:55 Scott And remember to rate, review,
31:57 Scott and subscribe on the podcast platform of your choice.
32:01 Scott Thanks for listening.