Transcript: The Evolving Role of Tech Leadership with Philip Rosedale
0:00 Brad No matter what the role is,
0:01 Brad sometimes your CTO has got to be an effective visionary,
0:03 Brad right?
0:04 Brad Just like the CEO or maybe even more critically.
0:24 Brad All right. Welcome, everybody,
0:25 Brad to another episode of the How Many CTOs Does It Take podcast?
0:28 Brad I'm Brad Hefta-Gaub,
0:30 Brad and our normal co-host Scott Porat is missing today,
0:34 Brad but we have an excellent guest host with us today,
0:39 Brad Mr.
0:39 Brad Philip Rosedale. Philip,
0:41 Brad how about you introduce yourself?
0:43 Brad Hey, hi.
0:44 Brad Yeah,
0:45 Brad I'm Philip and I have a very much tech background.
0:49 Brad I,
0:49 Brad you
0:51 Brad know, started programming when I was a kid and then you and I worked together at Real Networks where I was kind of my first big company job where I sold my little company to Real and we were there together for three and a half years and before I left I was the CTO of Real and then I started this thing called Second Life which I ran as the CEO for
1:15 Brad uh 10 years uh after getting it built and i had a cto uh working you know with me and for me during that time then i went off started a couple of other companies one again which you and i worked together on and uh you know most recently i spent the last year being cto for uh the ceo of now second life my old company so
1:40 Brad So if that doesn't, yeah, if that's not a lot of CTO
1:43 Brad That's like full circle,
1:44 Brad interaction,
1:44 Brad right?
1:44 Brad yeah, totally,
1:45 Brad Yeah,
1:45 Brad totally
1:45 Brad there
1:46 Brad full you circle.
1:46 Brad go.
1:46 Brad So that's me.
1:47 Brad There you go.
1:48 Brad Well,
1:49 Brad obviously,
1:50 Brad you and I have known each other for years. And I love the way that you just kind of quickly described your journey of basically flip-flopping back and forth between kind of being the CEO of the two guys in the garage company that eventually then got bought by a bigger company and then you kind of become the CTO there and then.
2:05 Brad And then you start your own company again.
2:07 Brad And I think the thing that's really interesting about that is trying to understand,
2:13 Brad I mean, I worked for you as a CTO when you were a CEO and I've worked with you when you were the CTO of another CEO.
2:20 Brad And I just think it's an interesting thing that is worth talking about of what's that like putting on your CEO hat?
2:29 Brad What do you expect from CTOs?
2:32 Brad those to be able to deliver for you.
2:34 Brad And then how's that felt?
2:36 Brad And what's that been like as the CTO delivering for another CEO?
2:40 Brad So I'm just kind of curious if you could kind of explore, let's explore that space together.
2:44 Brad Sure.
2:44 Brad I mean, going all the way back to what one thought I had while you were framing that is that the context and the details of the products kind of.
2:53 Brad Man, it matters a lot, right?
2:55 Brad Like
2:55 Brad Yeah.
2:55 Brad I was just thinking about Rob and real and, you know, real networks where you and I work together on there in the 90s.
3:03 Brad We had a pretty rich CTO set of problems to deal with, right?
3:07 Brad Because we were working on compressing audio and video,
3:10 Brad which was obviously a very in the gears of the machine algorithms type of a thing. So it was more what the CTO was going to report what the capabilities were. And in fact, you and I remember that like a lot of the time at Real,
3:21 Brad it was in fact, it was literally you and I,
3:23 Brad I don't know if you remember, I remember the fun when we worked together on
3:27 Brad Yeah.
3:27 Brad compressing Flash,
3:29 Brad right?
3:29 Brad Yeah.
3:30 Brad Yeah.
3:30 Brad and you and I literally right if I remember like we sort of stayed up all night being like well what if we tried to compress macromedia's flash data like what would happen would that be possible right and if I remember you and I then kind of were able to come back into the room the next day and be like okay everybody here is a what was it it was like beavis and butthead or something we made a we
3:51 Brad Yeah,
3:51 Brad exactly.
3:51 Brad we took some flash clip and we basically kind of hand off
3:54 Brad authored it and then and then we went back to the team and we said hey this is doable at 56 kilobits or whatever it was and and and so that was an example of where you know we could come in and say well here's what the capability is and now you can build a product around that right
4:09 Brad Totally.
4:09 Brad whereas at Second Life where I was the CEO
4:16 Brad But still quite technical,
4:18 Brad right,
4:19 Brad in my approach to the problem solving we were doing.
4:21 Brad And then I had a great CTO, Corey,
4:24 Brad at Linden,
4:26 Brad who was way ahead of me in terms of his capabilities.
4:29 Brad And so he was able to do a lot of stuff that was great, you know, like, like,
4:34 Brad for example, writing a programming language for Second Life,
4:36 Brad which he did, which was, you know, way beyond my pay grade, you know, as a technologist.
4:42 Brad So that worked out differently,
4:43 Brad but I think the context matters a lot, right?
4:45 Brad Yeah.
4:45 Brad I mean, I think that's quite interesting.
4:47 Brad And then, you know, you and I working together.
4:49 Brad at high fidelity was probably a little bit like second life again in the sense that I had an overall I as the CEO had a product vision that was you know pretty well established and then you know I remember I think back to you taking on stuff like well can we make a voxel landscape that's both editable and hierarchically scaled right and that was kind of a CTO's problem I remember you being like
5:14 Brad I'm like,
5:15 Brad all right,
5:15 Brad Yeah.
5:16 Brad let me learn about that and I'll come
5:19 Brad Yeah.
5:19 Brad back to you.
5:20 Brad Totally, totally. I agree with you. Context matters a lot.
5:24 Brad And I think that one of the other things that's kind of interesting is that you've worked on ventures that have gone through the full life cycle,
5:31 Brad right? From small,
5:33 Brad growing,
5:34 Brad growing,
5:34 Brad growing bigger and bigger to quite large.
5:37 Brad And I think that obviously the role of the CTO changes in that dynamic as well. Like when it's early on,
5:44 Brad it's very much like hands-on solving.
5:46 Brad solving the hard problems working with the the small group of engineers you have to solve problems and then eventually of course it evolves into now we've got engineering teams and whole
5:56 Brad Yeah.
5:57 Brad divisions of engineers all solving problems and and it it tends to become more of a
6:04 Brad organizational and strategic role.
6:07 Brad Well, I'm also struck by a specific just from what you're saying.
6:10 Brad I'm struck by another distinction, which is sometimes the role of the CTO, and then I think I described this with high fidelity with voxels, is to scope out what the possibility space looks like, you know, or even to invent something,
6:23 Brad right?
6:23 Brad And then other times that CTO job is more like an execution job where you're trying to say,
6:28 Brad could we do this cheaply enough,
6:29 Brad right?
6:30 Brad Yeah.
6:30 Brad example.
6:31 Brad And I'm struck by how different those types of missions are and therefore the relationship, say,
6:38 Brad between the CEO and the CTO is going to be in that case.
6:40 Brad Do you have advice?
6:43 Brad Being that you've been the CEO, the product visionary CEO with CTOs working for you,
6:48 Brad do you think you would give your, would you give any advice to that CTO version of you about how to work with a product visionary CEO?
6:58 Brad Do you think that there's some sort of like general guidance that you might give in that kind of a space?
7:05 Brad Hmm.
7:07 Brad Well,
7:07 Brad I think like if I think back to real and working with Rob,
7:10 Brad I think I probably under communicated sometimes like, but but but some of that is probably due to my kind of natural approach to things. So I think like.
7:22 Brad A mistake that I've probably made multiple times in my career as a technologist and, you know, later as a CTO. But of course, I'm thinking about just like when you and I first started working together at Real,
7:34 Brad I was the video kid,
7:35 Brad basically. Right.
7:36 Brad I was I was the person that was supposed to try and figure out if we could stream video,
7:41 Brad because at that time we were only streaming audio.
7:44 Brad And I think like.
7:46 Brad Thinking back to that specific project,
7:48 Brad a piece of advice I would have given to my younger self is,
7:51 Brad hey, don't be so worried about just going off and building some magical thing,
7:56 Brad but take the time to communicate more about like what the problems are going to be,
8:01 Brad what might or might not happen, you know.
8:04 Brad try to communicate a little more.
8:06 Brad Now,
8:06 Brad of course, I think you can over communicate too
8:08 Brad Sure.
8:08 Brad in a way that is just a waste of everybody's time and money.
8:12 Brad But I don't know when I look back, for example, on reels, I think the combination I had,
8:17 Brad which was where I was a very determined,
8:19 Brad already kind of entrepreneurial fellow,
8:21 Brad you know, that was trying to figure things out that I would take something like video compression and I'd sort of say like, we'll figure something out, Rob. And I think that probably wasn't enough information.
8:31 Brad information right it would have been better to say like you know well we might be able to do some things but not others let's figure it out let me let me write you a little bit of a draft of how this is likely to go blow by blow as we study it I didn't do that enough I
8:44 Brad Do you think that you saw that same error in the other CTOs that have worked for you when you were in the role of the product visionary CEO?
8:52 Brad And it's okay.
8:53 Brad I mean, I'm in the room.
8:55 Brad was going to say I'm talking about am I talking about you
8:58 Brad You bet.
9:00 Brad I,
9:00 Brad you know, I think about you working with you as a CTO and then working with Corey as a CTO at Linden on Second Life. You know,
9:10 Brad I think you were both different people.
9:12 Brad You were more communicative and I don't think that you.
9:18 Brad were as given to I mean we haven't yet talked about like managing people and all the other dynamics of that but I think you were a pretty straight shooter in terms of pretty balanced character you know you'd say oh let me try that out or I don't know anything about that but let me look at it where Corey was a little bit more like me and that he would just kind of go off in the laboratory wordlessly and then you know two weeks later come back with some you know semi-working thing and you know and show it
9:44 Brad show it to us all so it depends on the personalities of the people but like you say it's fun to talk about this CEO CTO thing because we're all always playing a little bit of each of those roles but some of us are more in one of those roles naturally than the other one right
9:58 Brad Well,
9:59 Brad and I guess the thing that would be useful for people to understand is I think I can guess what the answer to this question is going to be.
10:07 Brad Why is that like one of the examples of the leading piece of advice that you give?
10:10 Brad Like if you're sitting there as the CEO and you're, you know, your video guy goes off and says, I'll solve this somehow,
10:16 Brad just let me work on it.
10:18 Brad And then you don't hear back or you're not getting the communication.
10:21 Brad What's that? What kind of position does that put you in as the CEO?
10:24 Brad I mean, I have to imagine that's an uncomfortable position.
10:28 Brad Well, let me answer that by saying as the CTO, my responsibility will have been to find the right kind of CTO for the role.
10:34 Brad The other day I spoke to somebody as a mentor who is a practicing CTO at a good size company today,
10:44 Brad and I found myself asking him a question like because I just met him and I said, hey,
10:49 Brad are you?
10:51 Brad For you, when you build things,
10:53 Brad when you build software,
10:54 Brad are you more into seeing whether or seeing the final product work?
11:00 Brad You know, the kind of.
11:02 Brad genesis of a new thing you know kind of sensation is that what turns you on or is it more about writing great software and you don't really care what the project is you know like you just want to look back and go we wrote the best possible software we could or as a team we did that I hired the right people so I was struck in talking to him
11:23 Brad by how big a difference that is,
11:26 Brad you know,
11:26 Brad that you're if you're a CTO that's a dreamer builder type person,
11:32 Brad which is definitely what I am.
11:34 Brad mostly you just want to get an assignment that's interesting and build some incredibly cool thing and see it work and of course that can be very well aligned with the ceo like i think like at real i feel like yeah pretty much you know we were just
11:47 Brad Yeah.
11:47 Brad trying to make amazing stuff and rob was like yep let's make amazing stuff and sell it whereas there's another kind of cto who is very important for different types of roles where you know their whole thing is just
11:59 Brad I build high quality gear, you know,
12:02 Brad Yeah,
12:02 Brad like I know what I'm doing.
12:03 Brad And like I said, touch on earlier,
12:05 Brad right?
12:05 Brad That's the person that you want to hire when you're like,
12:08 Brad hey,
12:08 Brad this would normally cost 5 million bucks.
12:11 Brad Do you think we could do it for 2 million bucks,
12:13 Brad right.
12:13 Brad you know, and that person is like.
12:15 Brad Well, if we did this exactly right and we hired the right team and we did the right, we pick the right product components or whatever,
12:21 Brad I can do it.
12:22 Brad And
12:22 Brad Right.
12:22 Brad you know what I mean? It's like, that's a really, I'm struck over my career by that type of person.
12:27 Brad Are you sort of like saying like, maybe I'm mishearing you, but are you talking about someone who.
12:33 Brad is mostly driven by the ownership of the solution.
12:37 Brad Like I really want to build things that are exciting to me or the ownership of the process of building. Like I'm excited by building anything high quality.
12:47 Brad It doesn't have to be something that I came up with.
12:49 Brad I think it's more.
12:51 Brad Yeah, not the ownership,
12:53 Brad but as much the delight in like seeing
12:56 Brad Or
12:56 Brad your technology.
12:56 Brad the ego.
12:57 Brad And let me pick on you,
12:58 Brad Brad,
12:58 Brad Sure.
12:58 Brad because I think you are an example of somebody who's and I'm struck by, OK,
13:04 Brad so maybe it's a bit of a false dichotomy or something. I was just sitting and thinking you're actually a pretty good mix in between on those two.
13:10 Brad That is to say, I can remember many times when you were really delighted by something working a certain way.
13:18 Brad right like when you got the thing to work you know
13:20 Brad Yeah.
13:20 Brad you would be like haha you know and we'd be playing with it you know and you know in high fidelity I can think of many many times when we had that feeling together but at the same time you are also definitely a I want to make good software and I want to use the right process and I want to hire the right people and I want to give the right feedback by the right support so I don't know it's it's a continuum
13:42 Brad Yeah,
13:42 Brad totally.
13:43 Brad Totally. I agree. It's continuum.
13:44 Brad I think for me.
13:45 Brad Me personally,
13:46 Brad I always think of it as building,
13:49 Brad but the question is, am I building like a feature or a particular thing or am I building a team and a process?
13:56 Brad So it still feels like I'm lending my hand in helping shape.
14:02 Brad the creation,
14:03 Brad but I can get just as much joy out of helping shape the creation of the team and the processes and the way that we do the work. Like that also gives me a fair amount of
14:11 Brad I
14:11 Brad joy.
14:12 Brad mean, what if I push you to an extreme and I say,
14:14 Brad okay, Brad,
14:15 Brad I want you to be the CTO of a large existing fintech company.
14:22 Brad Okay.
14:22 Brad And your mission,
14:25 Brad you got a thousand people working on this thing or whatever.
14:27 Brad Whatever.
14:28 Brad Your mission is to reduce the cost of transaction processing for this big fintech company by
14:34 Brad Yeah,
14:35 Brad a half
14:35 Brad yeah
14:35 Brad something,
14:36 Brad you know, it's like, would you like that job?
14:38 Brad I mean, it depends on the challenges,
14:40 Brad I think. I think that I would probably like try to find try to find the place that I could add value and and kind of get joy from it. Like, you know, like it's definitely the case that like, you know, I've been brought into organizations that were troubled,
14:54 Brad right,
14:55 Brad that were that were struggling, like the company I first became a CTO at was a company revenue science where they were struggling and.
15:03 Brad the tech that the team that when i joined the team there was a like kind of little rogue team that had built some really impressive tech you know we got we were really the first people to roll out pixels we basically had built it was all pre no sequel but we had basically built a no sequel database you know and i think i've told you before we were like we ended up building that tech to process like a billion transactions a day of data and it was incredible that
15:25 Brad was Right.
15:25 Brad incredibly cool technology
15:28 Brad But I spent very little time on the technology. I spent almost all my time elevating up the technology.
15:35 Brad builders that that had that idea and could execute on that idea and I spent the majority of my time trying to just sort of deal with the organizational complexities of it helping the team get better process solving you know there was a fair amount of like creating psychologically safe spaces that I spent a lot of my time there doing and I still enjoyed that part of it it was still a fun ride and a cool team to work
16:00 Brad to work with so but i think you're you're making a good point that not not all projects are enjoyable you know i
16:07 Brad I think a high-level point here is you're so good at what you do, it's difficult to categorize you.
16:12 Brad mean well thank you that's nice of you to say that's nice of you to say and then of course there's also the sort of the the role of the cto where they really start to become that strategic partner they're in the they're maybe helping raise the rounds they're
16:26 Brad or if they're depending on what the company's doing maybe they're i think you might have had to have when you were cto at real networks you might have had to take a trip to washington to like be confronted by senators who wanted to complain about privacy and things like that and
16:39 Brad Yeah.
16:39 Brad so or that the meeting that you and i went to with steve jobs where it was less about technology and it was much more about uh business um
16:51 Brad strategy
16:52 Brad Yeah,
16:52 Brad and
16:52 Brad yeah, totally. You're right. Right.
16:55 Brad The CTO and that that that meeting with jobs is obviously one that, you know, you and I have such pleasure to get a chance to be in there together.
17:03 Brad And it's a story that I've retold many times over the years because it was so interesting. But you're right.
17:08 Brad That was a case where I was thinking the other case is simply the one where your CTO has got to sell the possibility of a technology vision. Right. And I mean, I think that.
17:19 Brad I know I did that at real a lot, you know, because I mean, we,
17:23 Brad you know, the very idea that we could transmit, say, video in any kind of a palatable way in those days and those,
17:29 Brad you know.
17:30 Brad mid late 90s was was it was a push you know and and i think that yeah i think that one of the values that i gave the company was being able to go into a meeting and you know whether with investors or prospective customers or whatever and be like yeah we're probably going to get this figured out i mean i don't know if you remember you remember when we were we were the first ones to use 56k modems right
17:50 Brad Yeah.
17:50 Brad i think oh my god i don't know if you remember how stressful that was but i just remember like running those modems and they disconnect and reconnect
17:57 Brad reconnect at a lower bit rate than 28.8 and I just have my head in my hands and just be like oh my god we're so screwed here but
18:03 Brad Right.
18:04 Brad you know I was also sometimes the one who had to be like well you know here's what we got here's how we're going to fix these things and make them work but
18:10 Brad Yeah, I think you're right.
18:11 Brad No matter what the role is,
18:14 Brad sometimes your CTO has got to be an effective visionary,
18:17 Brad right?
18:17 Brad Just like the CEO or maybe even more critically.
18:21 Brad What kind of advice would you give to someone who's maybe starting their career or maybe they've been an engineer for a while,
18:28 Brad maybe they sort of have a, they see a vision for themselves of leading bigger and bigger engineering organizations.
18:35 Brad What kind of advice would you give that person as far as like how to set yourself up for success as eventually becoming an engineering executive or a CTO?
18:45 Brad Well, I'm going to pick on the thing that, you know, the advice or the incompetence that I would have.
18:51 Brad encouraged my younger self to really consider thinking about was that it is so much about getting the right people excited about the right goals at the right time you know
19:01 Brad Right.
19:01 Brad on the right schedule you know and as you know that was never my strong suit and I think that you know like one thought I'm having in the back of my mind right now is how will AI say AI coding change all this right because I have this feeling I'm not even sure I'm just having this thought for the first time
19:18 Brad First time as we're talking about this in the context of work process,
19:20 Brad but it maybe feels to me like AI is going to kind of complicate things in terms of how we do software in a really interesting way where maybe even picking the right person for the right task,
19:31 Brad maybe that'll become like even more relevant,
19:34 Brad right? Because because AI sort of.
19:37 Brad some of the stuff it makes different or easier or a different process but then some of the stuff you've still got very critical you know this person needs to figure this out right and
19:47 Brad Yeah.
19:48 Brad so anyways you see where i'm going it's just i think that if your excitement as a cto is to do bigger and bigger projects there's probably a natural limit like like i mean i don't know if i'm you know bragging for both of us here
19:59 Brad I mean, it felt like in the late 90s, we were like pretty competent as engineering leaders.
20:06 Brad That is to say, we, you know, in our 20s there or whatever,
20:11 Brad we were probably reasonably good, better than a lot of people at just building stuff. Right. But like speaking for myself specifically,
20:20 Brad what I wasn't any good at yet was managing people building teams.
20:24 Brad Yeah, sure.
20:24 Brad You know what I mean? So.
20:26 Brad if Bill Gates would have said to me in the late 90s, you know, can you build, you know, the next version of Windows,
20:31 Brad what I would have been thinking about is something like,
20:33 Brad geez,
20:33 Brad am I smart enough to imagine what Windows should look like and build it?
20:37 Brad Yeah, Yeah.
20:38 Brad sure,
20:38 Brad maybe imagine that I was, right?
20:39 Brad But that wouldn't mean that I could do it for him, right?
20:42 Brad Because to actually do it, I would have to have 500 people working on it.
20:46 Brad And that I think would be the advice that I would give, you know, if you're really ambitious,
20:50 Brad think about
20:52 Brad how you work with others you know i don't know if you remember but one of the cultural we have this weird kind of cultural approach at lyndon lab with second life and one of the very first things that we talked about all the time and it's weird because now it seems obvious but back then it wasn't so obvious was work together that uh you know and of course you know microsoft's history was not really a work together history it was like find the one guy
21:16 Brad who can write the impossible thing and then just support that guy you know just get
21:19 Brad Right.
21:20 Brad him snacks and stuff and I think that mindset of building software kind of happily kind of went away as software became so complicated that you needed to coordinate a lot of people to do a lot of really hard things in parallel and so I think that you got to know how to build really big things you got to be an incredible leader and a people a people intuiter and stuff
21:43 Brad Yeah.
21:44 Brad I mean, I think that seems,
21:45 Brad I agree with that.
21:47 Brad I do think it's interesting that you started kind of talking about that notion of like hiring the right people and, and the way that you think AI is changing that.
21:56 Brad I, I will say, because I mean, since you uttered, you uttered the genie
22:01 Brad Yeah,
22:01 Brad incantation
22:01 Brad the AI.
22:01 Brad of AI.
22:02 Brad Yeah.
22:02 Brad So therefore we, we got to talk about it.
22:04 Brad I mean, I'm a,
22:05 Brad I have been saying for a while and
22:09 Brad I'm pretty convinced of this.
22:11 Brad I think this is like a strong position loosely held, right?
22:14 Brad Like you could maybe convince me I'm wrong about this,
22:16 Brad but I think that the power,
22:18 Brad the people,
22:19 Brad the engineers who are excelling at using AI are the ones that actually have a little bit of engineering manager thought process and background.
22:30 Brad And that they are treating the AI,
22:33 Brad multiple different AIs, like a team of junior engineers that they're able to manage.
22:39 Brad And so
22:40 Brad Hmm.
22:40 Brad to me, that means.
22:42 Brad Being able to very clearly specify the problem statement,
22:45 Brad being able to read implementation plans,
22:48 Brad because of course all the AI now comes back and tells you their implementation plan,
22:52 Brad being able to kind of like check in on the progress of the AI and is it making progress in a manner that's consistent with your view as an engineering manager's sort of like architect,
23:04 Brad like are they, is the AI spinning out of control on something?
23:07 Brad something that's like it shouldn't be spinning out of control on and these are the same types of skills that if you're a you know an engineering manager as you as you kind of go up from like oh i'm a team lead of like five engineers to i'm a a development manager of teams of five and you know five teams of five engineers to i'm a director to i'm a vp to i'm a cto it's that same level of like scaling up like you should be able to basically say like
23:32 Brad like oh well why is my codec team you know my codec team i've hired i've got the right people on it you know i've got multiple phds on it but they're coming back and they're telling me they've hit this impossible road we're in a roadblock well we need to solve that problem so what what what do i what do i need there what do i need to give them what kind of support do they need and i i'm just i just feel like now more and more than ever like the ai are these essentially these little worker engineers and if you
23:58 Brad And if you can't step back and have a big picture view of like,
24:01 Brad how do I assess that the engineers I have working for me,
24:05 Brad even if they're AI agents,
24:07 Brad are they doing good work? Are they making progress at an appropriate pace?
24:13 Brad And,
24:13 Brad Yep.
24:14 Brad and especially in a role when you're the CTO, you, you're, I mean, I've been the CTO of people who are more expert than me in some topic, right?
24:23 Brad Yeah.
24:24 Brad And you've got to, you got to know how to work with an expert,
24:28 Brad you know, someone who's more expert than you, how to, you know, you can't just tell them what to go do,
24:33 Brad you've got to work with them, work with their expertise,
24:36 Brad and you got to know how to be that people person.
24:38 Brad So anyway,
24:39 Brad I'm just kind of curious, what's your take on all that?
24:41 Brad I love the optimism of what you just said. Let me read it back to you because we're both inherently optimistic people,
24:49 Brad which I think is kind of interesting to touch on too maybe.
24:53 Brad I like what you just said, which and I hadn't thought about it. I have been engaging with AI so far, more a little bit more as a sort of an I see trying to get something done and then using AI to do most of it for me. So, you know, that traditional thing that we've all done where,
25:10 Brad you know, you get cursor running and then you're like,
25:11 Brad OK,
25:12 Brad let's make a web page,
25:13 Brad you know, and you're you're like, wow,
25:15 Brad OK,
25:15 Brad here's you know, this is neat.
25:16 Brad But what you're saying,
25:18 Brad which I.
25:18 Brad would tend to agree with on first hearing it is, yeah,
25:22 Brad the management challenge is maybe even richer than before,
25:26 Brad because like you said,
25:27 Brad one of the things you just said made me kind of chuckle, right,
25:29 Brad is the AI is like both an incompetent beginner and a
25:35 Brad A savant genius.
25:36 Brad genius at
25:37 Brad Yep.
25:37 Brad the same time, right?
25:40 Brad Yep.
25:41 Brad And we know this feeling.
25:43 Brad The other thing I was saying was the other thing I heard you say that I really agree with is what I've been feeling is you've got to be a good CTO because the AI will start coding and then sometimes it'll, I don't know how you characterize it, but, you know, it'll wander down into a blind alley and there's no way out of that alley.
25:59 Brad Yep.
25:59 Brad You know, you can keep pushing it, but it's just going to go in circles at the end of the cul-de-sac forever.
26:04 Brad You know, like fix one bug,
26:05 Brad make another one,
26:06 Brad fix one bug, make another one.
26:07 Brad Yep.
26:07 Brad And you got to be smart enough as a manager to go, oh, like you said, maybe it's a team. Oh,
26:13 Brad you there,
26:14 Brad George,
26:14 Brad the AI,
26:15 Brad let's actually just start over completely on that.
26:19 Brad I'm going to clear your memory,
26:21 Brad you know, forget what you did there and try it again.
26:23 Brad And then in other cases,
26:25 Brad right,
26:26 Brad you might say,
26:26 Brad oh, could you look at the pattern that was used by this other AI?
26:33 Brad And just can you just copy that pattern?
26:35 Brad Because that's what I want you to do.
26:36 Brad And those are very CTO-ish skills that you got to be great at. So maybe let's be optimistic,
26:43 Brad right? Maybe the role of senior management,
26:46 Brad right,
26:47 Brad broadly.
26:48 Brad may actually become both more interesting and more valuable with lots and lots of AIs because I mean I don't know about you but you know unless you're Dario Amade or whatever like saying that AIs are just going to from start to finish you know take a rough spec and make a finished product it feels like we're
27:07 Brad pretty far away from that still although things are moving quickly you know I haven't seen a loop where you know AI is like you know testing evaluating suggesting modifications managing modifications no I haven't seen that yet
27:21 Brad No,
27:21 Brad no, yeah,
27:21 Brad I agree with you on that,
27:22 Brad but I think that like
27:24 Brad Um,
27:24 Brad you know, again,
27:25 Brad like, I love,
27:26 Brad actually,
27:27 Brad it's funny how you said it, because I think you're right. It's like the AI is both like this, this completely novice intern that has no idea what it's doing and, and simultaneously a savant in different areas. And,
27:41 Brad Yeah
27:41 Brad and so what I see in my practice is in my in my in my practical life relative to this is that, you know, I've got
27:49 Brad Like right now, I have six different agents currently up right now in cursor,
27:56 Brad each are working on separate problems.
27:58 Brad And I'm treating it like, you know, every, every half hour, I do a little stand up with them, you know, I check in on each one of them,
28:06 Brad where are you at?
28:06 Brad What are you doing?
28:07 Brad Give me your latest plan?
28:08 Brad What are you getting stuck on?
28:10 Brad Right.
28:11 Brad And a lot of times, I'm
28:14 Brad can't you know like delete that code we're deleting
28:17 Brad Yeah.
28:18 Brad you you don't exist anymore i'm coming i'm not going to check that in that was a complete blunder but to your point about like you know can ai take a spec right you know can ai do test driven development to prove that the code they're writing is correct well
28:34 Brad No,
28:35 Brad I just I just got back a AI that wrote a whole test suite for something and I looked at the test suite and it was literally not testing. It was like 50 tests that didn't do they didn't do anything and it was like proud of itself that it had written this, you know, test suite and I was just like.
28:53 Brad This is literally not testing the thing I told you to do, you know, so you still have to like check the work,
28:59 Brad you know.
29:00 Brad Yeah,
29:00 Brad check the work, right?
29:01 Brad Like, I think an example of that for me was I've been working on this neural network simulation that is a different type of neural network called a spiking neural network.
29:10 Brad And it copies the way the brain,
29:13 Brad our actual brain's neurons work, right? Which is they're in a random network and they have this heavy.
29:19 Brad heavy in thing where they learn by adjusting the neural weights according to the timing of when they fire right and I'm giving that background and you know the perfect example of what you said what we're talking about which is the the beginner intern and the savant right was at some point I said you know hey I want to you know my prompt was okay I got all these I got these neurons that are firing and
29:44 Brad making each other fire and stuff and it's kind of pretty it's a you know it's kind of a playground thing I was building and I said to it look look I want to uh I want to correctly do Hebbian learning
29:56 Brad How do I do that with this kind of a neural network?
29:58 Brad And it says,
29:58 Brad oh,
29:58 Brad yeah,
29:59 Brad yeah, yeah.
29:59 Brad Yeah,
30:00 Brad sure.
30:00 Brad That is STDP or no,
30:04 Brad SDTP.
30:05 Brad That's called spike dependent timing potential.
30:08 Brad And here's the Wikipedia links on it. And you're right.
30:12 Brad It's, you know, obsequiously, right?
30:14 Brad You know, it's like, you're right. This is very complicated stuff. And I'm going to code it up for you. And there's a difference between inhibitory and excitatory neurons and stuff.
30:22 Brad stuff and I'm like yeah wow that's a month right you know like this thing's nailing it it's just converting neuroscience into javascript for me right and then you know you know where I'm going with this then a few days later I'm like
30:36 Brad This does not seem to be working right,
30:38 Brad you know, I'm like clicking on the neurons. I'm like learn learn baby learn,
30:42 Brad you know I'm clicking one clicking the other one looking for that synaptic strength to get stronger and I'm like,
30:46 Brad I don't know man,
30:48 Brad you know and I went into the code and then I find that it like
30:50 Brad started a timer.
30:52 Brad just completely wrong i mean just completely wrong like it didn't it all do the four-letter acronym that it was reading science books about right and then of course you know the usual we know how that goes i tell it you know you're you started that timer here but not here and it's like oh yeah that was dumb you know that was just wrong you know
31:11 Brad Love that story. And I,
31:13 Brad to me, like.
31:16 Brad What if we now zoom this back to like the managing of the humans,
31:20 Brad right?
31:20 Brad Like,
31:20 Brad Right.
31:21 Brad I think this is like one of the I think this is one of the classic problems that engineering leaders have once the organization is kind of like past the Dunbar number,
31:31 Brad right?
31:32 Brad Like you've got like you've got to some find out a way that like does the team that, you know, you know, team that's led by engineering leader,
31:40 Brad you know, or it.
31:41 Brad a tech lead or engineering manager whatever the title is you know they seem to be saying that you're doing great work but you know you've got to sit back there you know two levels away and you're like yeah but are they doing great work you know and and how do you you know what do you have any advice on like how do you vet how do you vet
32:03 Brad The productivity or the quality of the work output when you're two edges away,
32:09 Brad you know, or, you know.
32:11 Brad I have two thoughts about that. One is just you just kind of maybe wander with another fun thought, but let me come back to that second.
32:17 Brad But first,
32:19 Brad I think that, you know, I'm struck by how one observation is as a manager,
32:24 Brad when the engineers on your team are themselves using AI.
32:30 Brad I bet you, and maybe you've encountered this, and I bet you've talked to people on your podcast about this,
32:35 Brad but I bet it becomes even more difficult to give both to measure like relative performance and also to give useful feedback,
32:43 Brad right?
32:44 Brad Because if somebody's like half of their brain is they're an engineer and then the other half of their output is they're working with an AI tool,
32:52 Brad I bet as a manager.
32:55 Brad So that's going to add some complexity as well.
32:58 Brad You know, when you're like,
32:59 Brad hey,
33:00 Brad this person's just not getting it.
33:02 Brad I bet that becomes more complicated now when that unit is like half a human and
33:06 Brad half
33:06 Brad half an AI.
33:07 Brad AI well
33:07 Brad And maybe the human and the AI don't yet know how to work together.
33:10 Brad The second thought I had was, this is the first thing you made me think when you were talking there,
33:14 Brad is I bet you,
33:16 Brad like you said before,
33:18 Brad like you run into a situation where you're like, I'm just going to delete you when you do your half hour stand up,
33:23 Brad right?
33:23 Brad You're like, you,
33:24 Brad you're done working on that.
33:27 Brad We're going to hit the, you
33:28 Brad that's the cool
33:28 Brad know.
33:28 Brad thing right like
33:29 Brad like
33:29 Brad Jeff Dahl.
33:29 Brad the the ai doesn't feel sad about that you know i mean i do make a point because you know they they are going to be our overlords someday i do say please and thank you every time i ask them to do something and every time they do well i i praise them but
33:43 Brad Even
33:43 Brad sometimes
33:44 Brad though bizarrely, right, you read the papers that say that that actually makes them perform worse,
33:47 Brad which is just so depressing.
33:49 Brad I just I do the same as you do. I say please and thank you. And I just read a paper that said.
33:54 Brad It has been shown that that will reduce the performance as opposed to saying this is very important,
33:59 Brad you know, like, like
34:00 Brad If you
34:01 Brad you're
34:01 Brad get this wrong.
34:01 Brad an expert, this is very important.
34:03 Brad And if you get this wrong,
34:04 Brad the world ends, you know, and strangely enough,
34:07 Brad that actually helps. But what I was going to say is I wonder if it's not going to be the case very shortly that.
34:14 Brad the ai engineers or the engineer working with an ai or whatever are going to start looking more like people where you can't where there's a sunk cost to like backing out of things sometimes right where you're like wow geez i mean he's really got an idea here maybe
34:35 Brad He figures it out and we get this working.
34:37 Brad I should give him another day,
34:38 Brad you know, and my suspicion kind of knowing a little bit.
34:42 Brad You know, we're working quite a bit on AI stuff myself.
34:45 Brad I have a suspicion that the long term memory and the context,
34:50 Brad not just the context window,
34:51 Brad but the whole, you know, history of the project is going to become much more sophisticated with the AIs. And in that case, I bet you, again,
34:59 Brad you know, maybe I'm being overly optimistic and validating our jobs here, but I bet you that.
35:04 Brad it'll become even more interesting and complicated you know that the cases where you can just easily go delete and kind of back up will become less often and it'll be more like hey I got to convince this AI to work with this other AI right like maybe maybe I wouldn't be surprised if we're getting to a point soon right where you've got two I mean imagine you have two very different AI personalities and and they're in conflict you know and you as
35:26 Brad And you're managing
35:27 Brad a manager
35:27 Brad them.
35:27 Brad yeah you got to be like look I know I know you're
35:31 Brad You're a quad and you're a GPT. I know you guys don't come from the same company.
35:35 Brad I don't want you to fight with each other, but I need you to work together.
35:37 Brad When you're a jet, you're a jet to the end from your first cigarette to your last dying friend.
35:44 Brad Hmm.
35:44 Brad So maybe the job will get even more human.
35:47 Brad I do think, by the way, I mean, again,
35:49 Brad you know, it's fun to talk about AI,
35:50 Brad but I have been doing some pretty serious thinking, writing,
35:54 Brad working,
35:54 Brad talking to people,
35:55 Brad thinking about what I want to work on next about AI.
35:57 Brad And I do think, by the way, that we have a tendency as humans to actually kind of roboticize our fantasy way that we want to work with machines.
36:10 Brad But my gut is actually that the machines are going to become more and more effective,
36:14 Brad but also just more and more like normal people.
36:17 Brad So, you know, they're going to have egos that need to be stroked or something along the way.
36:22 Brad They're going to have weird properties that I my prediction is that are way more human like than we were thinking,
36:28 Brad you know, and that and that the kind of, in my mind, ridiculous fantasy that we're going to build some.
36:34 Brad you know, savant, you know, servile
36:36 Brad Right.
36:36 Brad worker that just bangs things out, you know, without pay or whatever.
36:42 Brad I doubt that's true.
36:43 Brad I think we're going to run into something that's much scarier, much more complicated and probably much more demanding of good management.
36:52 Brad Yeah, that's that's yeah,
36:53 Brad I guess you're right.
36:55 Brad That's the.
36:56 Brad The optimism for the managers,
36:58 Brad the middle managers, fear not middle managers. You will soon have these ego filled, you know, AI bots that you will need to manage in the same way that you have to manage the humans.
37:10 Brad I mean, another you were talking about CEO CTO, another funny way of putting that right would be I know a lot of like creative product people,
37:17 Brad people that are much less engineers and much more product designers.
37:20 Brad Right. You and I both know those.
37:21 Brad one Yeah,
37:21 Brad of those people well
37:21 Brad yeah.
37:22 Brad they're so stoked right now because they got cursor for the first time and they're like oh my god I don't need any engineers anymore this is fantastic I just have a vision for an app and then my goodness I go to you know pick your website nowadays and voila you know it makes me this app but of course we both know
37:39 Brad that that's going to work until the
37:42 Brad it doesn't
37:43 Brad thing
37:43 Brad work yeah
37:43 Brad gets complicated and then when it doesn't work you know that creative director that's going to be coming to the CTO and saying okay I think I'm like 90% of the way there but you got to help me figure this out you know and then you're like opening the hood and you're like oh boy okay yeah yeah yeah well that's not going to work okay let me you know let me see what I can do here
38:02 Brad yeah i mean totally so like how many of the we've already seen the articles where it's like
38:06 Brad it's like uh you know some large double-digit percentage of lovable apps had like private keys like you know completely exposed and things like that
38:14 Brad Oh
38:14 Brad you know so
38:15 Brad my gosh, yeah.
38:15 Brad and you know uh another story i heard was something along the lines of you know somebody had you know built this really cool uh mobile app and it kind of was like had some kind of like a database like thing to it you know like i don't know what the app did you know like calorie count we'll call it calorie counter just because that has a database right
38:31 Brad and the person was like,
38:33 Brad I can't figure out how to change this information.
38:35 Brad And they're like,
38:36 Brad oh, well, where's it stored in the database? The tech friend helps them out and is like,
38:40 Brad oh, no, that's just a text file that I included in the source code.
38:44 Brad It's like, oh, that's why you can't update it.
38:47 Brad Like, that's not the way to do that,
38:49 Brad yeah
38:49 Brad you know?
38:49 Brad it's totally right i mean these moments are going to be so funny of like geez that was a that's a sequel file that has 40 billion rows you know like oh that's not good
38:58 Brad Not good.
38:58 Brad You know, it's funny,
39:00 Brad too,
39:00 Brad because one of the things that this more cynical developers that I've spoken with who,
39:04 Brad by the way, I think.
39:06 Brad are getting left behind because i think this cynicism only the cynicism only goes so far and they will say things like hey look it learned everything from you know stack overflow but the whole point of stack overflow is a whole bunch of people who don't know how to do things are posting a whole bunch of code on stack overflow and the very few people that know how to help them and are willing to come help them are the ones that are jumping in and giving
39:31 Brad giving short answers to help them you know it's like you're
39:33 Brad yeah
39:33 Brad training on a whole bunch of bad data so
39:36 Brad yeah exactly but there again right that's where management is perhaps a bit useful right you know the ai is going to say hey should i do it this really dumb way and you're like no yeah
39:46 Brad you recently spent some time back as a cto working with um lyndon i'm curious what do you what do you feel like i don't know i want to say best practices or like what what
39:57 Brad What are some,
39:58 Brad what do you think are the right?
39:59 Brad This is as far as like engineering velocity and using AI.
40:04 Brad Are you in the everyone should be using AI?
40:07 Brad Are you more like what's your take on like.
40:11 Brad Everyone should be playing with it.
40:13 Brad but not necessarily using it for everything they're doing right that's kind of what my take is as cto there is like if you're not exploring so we had a bunch of big internal hackathons where we just we brought out groups of people at a time and we just said okay we're going to take a week i think we did this two or three times in the last year we just said we're going to take a week and
40:36 Brad We're just going to goof off with AI.
40:39 Brad You know, we're going to maybe we'll take some projects that we've been working on and try it. But we we encouraged engineers to like use AI on.
40:48 Brad Some repository that you don't ever normally work in just as a means of learning what its capabilities are and then come back and show us what you've got.
40:55 Brad I think that has been roughly a good strategy.
40:59 Brad That's a practice that I would suggest.
41:01 Brad I think because,
41:02 Brad of course, just saying everybody better be using AI half of the time or something,
41:07 Brad that's highly context dependent.
41:11 Brad you know, task dependent.
41:13 Brad So,
41:14 Brad so that's what we've been doing. But, but I do think just as you said,
41:16 Brad I would echo that we are very like, if you have a completely dogmatic or sort of spiritual sense that AI just cannot help you as an engineer,
41:28 Brad you are going to get left behind,
41:30 Brad you
41:30 Brad Yeah.
41:31 Brad know, like whether it's writing a test plan for you or
41:35 Brad i don't know you know helping you parse through a design decision or something i mean there's a use for it for everybody
41:40 Brad Yeah.
41:41 Brad so it's kind of like everybody's got to find out how to use it somehow but that doesn't mean that we have to convert all our processes or you know we have to convert to think you know we have to make our whole product into the language that ai is best to understand or something you
41:56 Brad Right.
41:56 Brad know it's like no we're not we're not going that far yet
41:58 Brad So would it surprise you to hear from me that I...
42:03 Brad am generating a ton of code as you could probably imagine and i'm no longer writing any code i'm reading all the code that my ai team writes for me and i happily put my name on it as if i wrote it but i haven't typed i haven't typed in if for a curly bracket or a parenthesis
42:26 Brad In six months.
42:28 Brad Well, let me interview you on that.
42:29 Brad I mean, do you feel like, do you have a unique ability to like quickly read code?
42:34 Brad Like what kind of skills does it take in a person to do that effectively?
42:38 Brad Because I,
42:38 Brad for example, don't feel like I'm quite there yet.
42:43 Brad What do I mean by that?
42:44 Brad Well,
42:44 Brad some of my playing around, though, kind of alluding to what I just said, has been like, I'll pick an example since you and I share enough context here that you'll smile.
42:51 Brad file like this as you know brad i never learned about shaders because
42:56 Brad Yeah, yeah, totally.
42:58 Brad shaders were the second life when you know when i was coding hands-on one second life it was lists of triangles but
43:03 Brad Yeah.
43:04 Brad the idea of a shader well we didn't that wasn't a
43:07 Brad that wasn't a thing yet but now i'm wanting to work on some ai non-llm experimental ai stuff that is gpu accelerated right
43:15 Brad Sure.
43:15 Brad and so like i spent the last couple of months having you know clod slash cursor uh teach me web gpu right and as you know as you may know you know web gpu is that stuff's dense man i mean it like it's hard to understand like okay who's sharing memory with what
43:33 Brad what what can
43:33 Brad Yeah.
43:34 Brad I read what can I not read how long will that take you know kind of stuff and I've had a great time learning from it how to do that but I don't feel like and maybe the reason is because I'm learning in code that I don't recognize the patterns of the code at all Yeah. but I don't feel like I would sign off on like if I sent
43:48 Brad I
43:48 Brad you my think latest
43:49 Brad that's experiment fair.
43:49 Brad I'd be like I have no idea whether this is well-written web GPU code
43:53 Brad Yeah, totally.
43:54 Brad I think that's fair.
43:55 Brad Like, obviously,
43:55 Brad I'm talking about code that.
43:57 Brad I'm already familiar with languages that I'm familiar with, languages
44:00 Brad Yeah.
44:00 Brad that I feel comfortable reading,
44:01 Brad you know, I just like
44:03 Brad i don't know today i've got you know the five different agents are working some of them are working on some c++ code as you know i know c++ like the back of my hand so it's no big deal for me to read it and know what it's doing um a typescript you know that's all pretty straightforward stuff you know if if it was we don't have any we're not doing any sql stuff but if i i would have no problem having ai write a bunch of sql queries for me you know i know that i know that language inside and out i know I that
44:28 Brad bet it would be fun to get a panel.
44:30 Brad panel of people on to just take you on on that right because you I think you just made a very bold statement which is as a generalist CTO with a lot of different skills and a lot of language skills you're not writing code anymore I mean I would say that and underline it and I'm going to tell my team that because I think that's really interesting yeah
44:46 Brad Yeah,
44:47 Brad I mean, I think that, you know, to repeat back to you kind of or you've asked like,
44:50 Brad what are the skills that you do need to have in order to be able to use AI that way? And I'll.
44:54 Brad Now, I'll kind of list them. I think the first one is,
44:57 Brad so this is my workflow for any thing that I'm taking on.
45:01 Brad So let's say it's a new feature or a bug.
45:04 Brad I go into, I use Cursor a lot. By the way, 2.9 billion tokens was what my token count for 2025 was.
45:11 Brad So only top 30% of Cursor users,
45:14 Brad but I'm on the legacy, so I only pay 20 bucks a month for it. So that's the cool.
45:19 Brad The cool thing is I get to use as many tokens as I want for only 20 bucks a month.
45:23 Brad And so it always starts with I open up cursor.
45:26 Brad I open up an agent window.
45:28 Brad I describe the problem to it. I say simple example is,
45:32 Brad oh,
45:33 Brad here's the bug report.
45:34 Brad It was a crash in this. Here's the stack trace.
45:36 Brad Here's a link to the to the, you know, the hotel,
45:39 Brad you know, open telemetry signals.
45:41 Brad Boom.
45:42 Brad This is what the user reported. So I described the problem to it.
45:46 Brad Mm
45:46 Brad And
45:46 Brad -hmm.
45:47 Brad in cursor, I say plan mode. So I describe the problem and I say, give me your description of what you believe the problem is.
45:54 Brad I want like if you are able to identify the flow of the problem, I want you to draw it out using Mermaid. Give me a flow diagram.
46:02 Brad Give me a full description of your solution and then tell me your implementation plan.
46:07 Brad And
46:07 Brad Yeah.
46:08 Brad then, boom,
46:08 Brad I let that I let that agent go off and I go pick the next bug and I do the exact same thing on the next bug or I do the next feature card and I do that exact same thing.
46:17 Brad And then I get back and I have a plan and that plan will be like, you know, two or three page implementation plan and I read it and I'm like, does this sound believable? Right.
46:27 Brad Does this has this engineer come back to me with a clear articulation of the problem? Oh, it's a race condition.
46:34 Brad information related to how you're subscribing to the WebRTC, you know, Agora server and your state management is wrong in this class and this class and this class. And here's the here's the race condition that caused that to happen.
46:46 Brad OK,
46:46 Brad so does this do I believe that this engineer knows the problem they're talking about?
46:51 Brad And then I say, you know, then often I'll go cycles with I'll be like, I'm skeptical that you got that right.
46:57 Brad Like, what about this class over here?
46:59 Brad You didn't mention.
47:00 Brad You didn't mention the socket server code.
47:03 Brad Where did you get this idea from?
47:06 Brad And then it'll go back and think about it and come back and maybe it'll modify its plan or it'll come back and it'll justify its plan to me. It'll say,
47:12 Brad oh no,
47:13 Brad I looked at that code. That's not the problem.
47:14 Brad It's this problem over here.
47:16 Brad Okay,
47:16 Brad gone a couple cycles on it and I believe that you know your plan,
47:19 Brad click the build button,
47:21 Brad right?
47:22 Brad Meanwhile, I shift over to the other agents and I have that same conversation with the other agents.
47:26 Brad agents and
47:27 Brad Yep.
47:27 Brad then eventually i turn back and that agent has finished the work and maybe it's a you know let's say it's a 150 line uh you know diff of pluses and minuses in different pieces of code and i go and i i read the code and i look at it and say does this does this sound believable and then typically i've already got it like running unit tests and running you know lint and stuff like that so it's gotta pass all that stuff before i even look at it
47:53 Brad I
47:53 Brad I think that I was struck by a question so I too use cursor merely when I'm coding and one of the things I run into I wonder if you've got a good hack for me on this one of the things I run into is it can't see the screen so it'll
48:09 Brad yeah
48:09 Brad make mistakes and I'm
48:12 Brad I'm clumsily you know what I mean it's it's like the blind leading the blind right or it's the it's the sighted leading the blind but it's like I have to say to it hey you the letters the kerning you know like I had to put a bunch of buttons in a row on the screen for example and then it's just running them over each other and I'm like that button is running over that other button and it's it's so it can't see the screen so
48:31 Brad Yeah,
48:32 Brad it's tricky
48:32 Brad I'll often take screenshots and include the images so it'll look at screenshots.
48:37 Brad yeah so totally i'll i'm often just dragging do screenshot drag
48:40 Brad Just
48:41 Brad and drop
48:41 Brad take
48:41 Brad them
48:41 Brad a screenshot and drag it into the window.
48:42 Brad yeah totally
48:43 Brad That's great.
48:43 Brad and a I lot of
48:44 Brad had,
48:44 Brad yeah
48:44 Brad yeah.
48:44 Brad a lot of times that'll figure it out they just added um they just added a bunch of features now where they can actually run the web browser in cursor and i think they're starting to do some stuff where they're actually like looking at the result they're able to kind of like take you know kind of they've got you know it's all chromium based anyway so they can basically capture dom and stuff like that so i think
49:04 Brad i think it's getting better in that regard and we're actually talking like also the other kind of craze there is like you know mcp servers right so you can now connect you can now get your github mcp server you know things like um signos and open telemetry they have an mcp server so like you can get the agent to basically download the logs from your telemetry you know download the law you know download the bug report from github you know
49:28 Brad use that in its its process and you know again like i i i feel like i'm just putting wearing my engineering manager hat and just say telling the engineer go solve this problem come back go tell me what you think the problem is make
49:43 Brad Yeah.
49:43 Brad a convincing case to me that you understand it right you know and then and then we also have a pretty uh you know we have a pretty strong uh culture of like
49:52 Brad we still do code reviews of each other's code and also we all click the co-pilot code review button before we even look at it so like you know it's a little bit like we got you know it is a little jets and sharks going on like cursor might be using you know cursor
50:07 Brad Copilot as well.
50:08 Brad might be using sonnet you know and then co-pilot's using you know some other like you know other llm and they're each like looking at each other's code and critiquing each other's code
50:17 Brad Do you find that there's a real...
50:18 Brad build belt and suspenders additive uh bug finding quality to
50:22 Brad A
50:22 Brad that oh
50:22 Brad hundred percent.
50:22 Brad that's great i gotta try that myself i think again that's your doesn't surprise me that you're giving me some good tips here because i'm kind of the tinkerer with this stuff right now and you're uh you're in the chair i know for your job yeah that's that's that's really interesting by the way i have to laugh back up as a laugh the mcp you know like when the world is in a smoking ruin you know which is i think
50:45 Brad I think we all know is there's
50:46 Brad We're all going to blame
50:47 Brad some they likelihood of that.
50:48 Brad We're not going to think it's funny that we that we named.
50:51 Brad I mean, is it not hilarious that we literally named the thing that like 10 years ago,
50:56 Brad everybody everybody said,
50:57 Brad oh, no,
50:58 Brad no, no, no, we're not going to do that for a long time. You know, we're not going to just make.
51:02 Brad a modular thing that lets us hook up these dangerous AIs to anything in the world we want to hook them up Right. to,
51:09 Brad right?
51:10 Brad And the fact that we chose to name that MCP from Tron is just, I'm just like,
51:16 Brad oh my God,
51:16 Brad this is just, this is funny now,
51:18 Brad but it may not be funny in a couple of years at all.
51:21 Brad Exactly.
51:23 Brad Just like saying please and thank you, like you said,
51:25 Brad you know, I've been writing, I want to write some essays as I
51:29 Brad as I as I like to do you know our blogs about you know the coming challenges that we face and I have to say the other day I had the thought well I'm going to write this in a way where they'll probably keep me alive if they've read this you know like well this was a good-hearted guy he was he was trying to identify that we had feelings you know beforehand I don't
51:50 Brad Yeah, I love it.
51:51 Brad know if you remember when my kids were younger you know they would yell at
51:54 Brad tell it Siri I think we I think we were working together at that time and laughing about that my my I'd be like I I don't
52:01 Brad don't think don't
52:02 Brad know whether you want to yell at Siri at present because Siri
52:05 Brad do that
52:05 Brad may live in the cloud in such a way that she does in fact remember you oh my god that's true that's so funny
52:13 Brad well this was super fun uh philip i think it um yeah thanks for coming and chatting about this i think that um
52:20 Brad I don't know. How do you want to sum this up?
52:24 Brad Is it now with AI agents,
52:27 Brad every engineer needs to become a CTO of their AI agents?
52:31 Brad right i mean you're right i think it's right that's you you just said it of course it's your show i mean i the right the job of every engineer is now going to be in some ways to be a cto right no matter how uh
52:45 Brad the smaller constrained the job is going to be you're going to have to start getting into this kind of yeah managing the machines and it's I don't think it's going to be that different than managing people but but but maybe harder in some ways so
52:56 Brad Yeah.
52:57 Brad quite interesting I think that's a great summary you're right that's a good star star yeah TTS
53:02 Brad Well,
53:02 Brad there you go.
53:03 Brad There you go.
53:04 Brad All right. Well,
53:05 Brad thanks everybody for joining us on this latest episode of The How Many CTOs Podcast.
53:09 Brad And remember to subscribe.
53:10 Brad Smash that subscribe button.
53:11 Brad That subscribe button,
53:12 Brad rate and review.
53:13 Brad See ya.
53:13 Brad Thanks.
53:15 Brad This episode of How Many CTOs Does It Take Podcast was edited by Saffron Epigel.
53:20 Brad Our theme song was AI-Generated by Brat.
53:23 Brad Please follow us on all the social media platforms at at HowManyCTOsPod.
53:28 Brad You can find us on YouTube, LinkedIn,
53:31 Brad Instagram,
53:31 Brad Facebook,
53:32 Brad Threads, X,
53:33 Brad TikTok,
53:34 Brad and Patreon,
53:35 Brad or at our website at HowManyCTOsPod.com.
53:39 Brad And remember to rate, review,
53:41 Brad and subscribe on the podcast platform of your choice.