0:00 Scott The UI for the factory is the Kanban board.
0:20 Brad Buddy to another episode of the How Many CTOs Does It Take podcast. My name is Brad Hefta-Gaub, and my trusty co-host is Scott.
0:29 Scott Scott Porat, how are you doing today, Brad?
0:31 Brad I'm doing pretty good, Scott. What's the bee in your bonnet today?
0:35 Scott I got two questions for you. You can pick. Door number one, software factories. Door number two, reverse engineer anything. And door number three is going to be. vibe apps okay
0:47 Brad Hi, Baps. I'm not picking door number three. I don't know what's behind
0:50 Scott have
0:50 Brad that door. I like reverse engineer anything for 100, Scott.
0:55 Scott you done it have you used it played around with it okay
0:58 Brad No, no.
0:59 Scott that's
0:59 Brad But I have
1:00 Scott a boring
1:00 Brad talked to conversation a few people that did. Have you?
1:03 Scott no no no I'm thinking about it though I have a friend who is in the camera business her
1:11 Brad Okay.
1:11 Scott and her brother i should say have a camera store they sell cameras and things like that and they use some software a point of sales system at their store from the from the before times
1:24 Brad Yeah. Is written in PHP.
1:28 Scott no way way before that i mean i think it's probably a vb app it's from it's from way back when people it's a sense ancient it's like 40 years old And it's a point of sale system specifically designed for camera stores, has some features and capabilities in it that are very beneficial to camera stores, but unique for camera stores. So for instance. they track each camera like the serial when you buy a camera you're not getting a unit off the shelf you're getting a that specific serial number right of of camera um and they might have if they've got 10 of some nikon and they've got 10 zero numbers and they're selling you a specific one so just like that kind of stuff that's i'm sure there's some other businesses that are like that so they got this this ancient thing and she's asked me a few times like what should i do about this like it would cost
2:14 Scott because you know there's just like two guys on the east coast that run this thing and i don't think they have very many customers but like trying to shoehorn our business into a generic thing like shopify or whatever or whatever system people use these days you know we're not talking about square but i mean to say it doesn't really fit and then i heard a couple of weeks ago about reverse engineer everything and i thought oh Yeah,
2:36 Brad Yeah.
2:36 Scott that's interesting.
2:37 Brad Well, yeah. So what I've heard, so again, I haven't played around with it yet, but I did talk to one of the one of my star engineers that I love working with. He he went out and he pointed it at. he's a he's a fitness guy and he pointed at strava and he was really happy with the resulting app that he got from it and he pointed it at um zwift which i don't know if you know about zwift it's like a on it's basically like Zwift is a think of it as like an online game multiplayer online game where you're you're riding your stationary bike and it's like you're riding your bike in the Alps with all the other people that are on the same roads as you so that's what Zwift is
3:20 Scott Wait, wait, are you actually riding your bike
3:22 Brad yeah
3:22 Scott or is you're it like
3:22 Brad riding your stationary
3:23 Scott stationary
3:23 Brad bike
3:23 Scott so you're able to see or you're tapping the you're
3:25 Brad no
3:25 Scott tapping
3:25 Brad no you're
3:25 Scott the
3:26 Brad riding
3:26 Scott arrow keys as fast as you can?
3:27 Brad no you're riding your stationary bike you got a trainer in your in your your workout
3:30 Scott Got
3:31 Brad room it. so you're pedaling and it gets harder when you're going uphill and it gets easier when you're going downhill and and you got your computer monitor in front of you and it looks like you're riding your bike through the alps i
3:41 Scott Do you know what I was talking about when I was saying hit the arrow keys really fast?
3:45 Brad assume you were talking about some like a casual game or something
3:49 Scott There was a video game like when I was a kid, like on the Apple IIe.
3:53 Brad Yeah.
3:53 Scott It was called winter summer games and winter games and like you you run you hit the keys really fast to run and then you hit the space bar to jump or whatever
4:01 Brad I was playing the Winter Olympics game on my portable, we had a version of the Commodore 64 that was a portable version of the Commodore 64. And I remember I was playing on that when the Challenger. uh accident happened
4:20 Scott Oh,
4:20 Brad so
4:20 Scott wow.
4:21 Brad i always associate playing that olympics video game i think i was trying to do daffies or something like that by hitting the j key rapidly or something there
4:32 Scott We digress very far. I was an eighth
4:35 Brad you go that was a lot
4:35 Scott grade French class when that challenger blew up. My French went nowhere, by the way. Okay, so back to business. Your friend has tried emulating a couple of apps,
4:47 Brad yeah
4:47 Scott not emulating,
4:47 Brad reverse
4:48 Scott reverse engineering.
4:49 Brad engineering um and he basically said yeah he was really impressed he was like yeah i really did it and i was like yeah but nobody's i'm like how did how's that going to work with zwift like the whole point of zwift is you've got the other people there they they all need to run your reverse engineered version you know you're not you're not connecting to their service and he's like oh yeah that's a good point um i saw
5:08 Scott Out of curiosity,
5:08 Brad well
5:09 Scott how does it even work?
5:10 Brad so this is this is where the next teardown i'll tell you i saw somebody do a pretty detailed teardown of both the supposedly i've copied all of the adobe suite guy right and this guy was basically the teardown was done by somebody who is who's actually a very you know you know power user of the adobe suite and he basically kind of came back and get said yeah all the menus look the same and they mostly do the same thing but you start pushing down on it a little bit further you start to realize that a lot of the functionality actually did not get implemented
5:44 Scott Mm-hmm.
5:44 Brad And so that was the other teardown that I started hearing. So I saw some other people basically doing other versions of the reverse engineer anything. And what it really is, is it appears as if it's not so much that it's fully reverse engineering anything as much as it is. There's a difference between reverse engineering something and. doing a clean room implementation do you know what i mean by a clean room implementation so
6:12 Scott I don't.
6:12 Brad like this was a thing i don't know i don't know how often this happens anymore but i remember when i kind of first got into the industry there was this whole like notion of oh yeah you know you ip loss seemed a lot stronger and it was like well we can't reverse engineer this competitor's code but what we can do is we can have a whole bunch of people use the competitor's code and write a specification specification from it and then we can hand that specification to a group of people who've never seen the competitors product and then they they will implement the specification right that's clean room idea and that's really what it to me it appears as if that's what this reverse engineer anything really was doing it was really more just like it it it exercises the application to the best of its ability and writes a spec from that and then
7:00 Brad writes an implementation from that spec right
7:02 Scott That's
7:02 Brad which is interesting. pretty close to what you'd kind of want but it all depends on how good of a job you did in the how good of a job the robot did in exercising every single feature and functionality of the system to understand
7:14 Scott Well, and does one get an opportunity to review that spec?
7:20 Brad No, that's the whole point of it. This is like a it's like its whole its whole thing is to be a one shot, just like log it in and have it just run overnight and it comes back and gives you your implementation.
7:29 Scott Well,
7:29 Brad Right.
7:29 Scott it's interesting because if it just gave me the spec to read and then I could feed the spec into my robots that I already know and love, it might be okay.
7:38 Brad Yeah, and I haven't. And, you know. I don't know. Don't don't quote me on the Internet. Don't don't don't pile on in the comments on TikTok saying that I'm an idiot boomer because I haven't actually tried it yet. But I'm impressed by it, but I'm not impressed by it for the reasons that I think a lot of other people are impressed by it. I think a lot of people are impressed by it because they genuinely believe that it's doing full copies of everything.
8:00 Scott Mm
8:00 Brad I'm not impressed by that because that's not actually working. From what I can tell, that's not actually happening. thing um But what I but it is one more proof point that you know the systems are pretty good at writing a lot of code for you that's what this is that's why it's good that's what's impressive about it totally
8:17 Scott Give it another five years and see how it can do.
8:20 Brad yeah
8:22 Scott Okay. All right. So you got lessons number one and three. You've got software factories and vibe apps.
8:32 Brad Well, let's do software factories next.
8:34 Scott Okay. I think you've made yourself a software factory.
8:39 Brad Yeah. I mean, I have an agentically, I mean, I've got a bunch of teams working in SDLCs that are very highly agent driven.
8:49 Scott So earlier this week I was talking to one of the engineers on our team and he mentioned this thing that he had made a dashboard for his little his factory his agents I was like oh love to see that and he said oh actually you should go talk to Steve because his dashboard I haven't actually really worked on my dashboard that much you should go look at his it's way better.
9:09 Brad Okay.
9:11 Scott so i went he i flagged steve and i was like hey can you send me if stefan said this can you send me some screenshots i'd be curious to see etc and he did that i looked at it i was kind of it was pretty cool i mean it showed a lot of stuff it obviously showed you know tickets and what agents were working on and then where they were in the process it showed each ticket how many tokens did it use and the cost of each ticket and he then it had a like i don't know what he called it but like you know a page that was monitoring the whole system you know and giving him, you know, giving him statics, metrics and statistics about the whole thing, and it was a neat thing.
9:44 Brad Just to clarify, it sounds like he used the Kanban board view. When you said tickets and statuses.
9:52 Scott Yeah. So at a high level, like basically he had this like ticketing type of system that was, I think it's just like markdown files or something. There's some notion that you create a ticket that was a markdown file and then each ticket works its way through a series of steps, you know, like basically your kind of standard STD type of specs, right? They're not specs, I mean, STD type of steps. and then he had a view in there where you could sort of see like it looked to me like a kanban board which is like here's each step in the process and which tickets are in which step right to kind of get a bird's eye view of what was going on um anyway okay so i saw this i was like well that is cool in fact um and i thought to myself oh it'd be really cool if everybody was doing this and then i thought now wait a second
10:39 Scott he's built one and I know this other guy's built one I know probably half a dozen people who have built this kind of thing easily and I thought to myself should we be building this or should we be buying this and that's the question that is the question about this agent factory this nonsense world that we're in right now
11:00 Brad Yeah,
11:00 Scott if
11:00 Brad totally.
11:01 Scott everybody is like building their own thing and it's fun i think a lot of people are having a lot of fun doing it but it gets to this point where i wonder like should we be building this stuff or should we be buying this stuff
11:12 Brad Yeah, totally. Okay. I love it. I love this is the thread you're pulling on. This is way more fun than reverse engineering anything. I
11:20 Scott well
11:20 Brad don't know why I'm
11:20 Scott that's
11:20 Brad like that.
11:20 Scott why it was door number one
11:23 Brad I chose poorly.
11:25 Scott Wait till you get to door number three.
11:27 Brad Okay. All right. Well, okay. There's a couple of things about this that I think are interesting. The first one is. that I saw a post recently. I don't know if it was from the pragmatic engineer. You know, we just talked about the pragmatic engineers report recently, but I think it maybe was a post from him or might've been somebody else that's, you know, and they were basically saying that they were reporting on big tech companies who essentially have all built their own. factory uis and so a bunch of you know like the ubers and the and the whatever linears of the world are out there or the notions of they're all out there building their own um this thing so to your point
12:16 Scott We're all building our own.
12:17 Brad they're all building their own right um now you maybe you're saying like you've got six people on your team each building their own personal ones but um but yeah there's this question of like You know, none of the like the none of the big dev shops like Cursor, Cursor, for example, SpaceX, right? X.A.I. Their UI is still very much like single player mode, IDE-ish thing, you know. I mean, they've kind of pivoted away from the IDE to kind of be a little bit more like Claude codecs, you know, where it's just this like chat, you know, it's the chat, but the chat now is the front end on top of the individual agents or the agent, you know, projects that are running. And so, I mean,
13:03 Brad that's why I asked you about the Conbon thing. things because i was like you know i definitely feel like um the visualization of these things are there nobody's nobody nobody is selling one as far as i can tell no big player in the dev tool space is selling one that scratches the itch that your developers were apparently feeling because they've they've built something that's different than than what the big players are so i think it's possible that we are really kind of talking about like um there's just a gap in the market right now um now i might be wrong about that that it's just that there's a gap it also might be the case that everybody's sdlc is slightly different enough that maybe it's not
13:49 Brad landing as cleanly i mean you would kind of it seems like you would kind of be like well why hasn't jira and or um you know atlassian or any of the other people that are all about large-scale project management they would be the natural you'd think they'd be the natural leaders of being able to present this new tool that just has the agents doing all the stuff that's what I would think if you're going to buy something that's I'm assuming who you'd be buying it from right
14:15 Scott Yeah, it's interesting because at least in my head, the analogy was more of like the IDE vendors, so like VS Code or whatever, which is, you know, kind of the agent dashboard is like the new IDE.
14:28 Brad Yeah.
14:29 Scott It's not an IDE at all, but I mean, it's the new phrase differently. The IDE was the console at which you programmed by hand, and now the agent dashboard is the console at which you program by agent. And so like, to me, it seems like it fits more that use case. It is a tool for an individual engineer. to manage their work, you know, it's just showing the tickets that they have going on in their little version of the factory.
14:57 Brad Yeah, totally. And, you know. cursor kind of did that you could say that claude desktop and code and codex desktop are kind of their versions of it and they all kind of have a similar feel they kind of all are kind of that chat that chat sort of like threads based thing you know there was a little bit of a buzz that happened about a month ago which now i'm still not hearing that much about now is was was buzz came out and so then buzz and slack and a couple others were kind of saying like oh we're going to put this this front the front end is going to look like slack it's going to it's going to look like chat messages but it's going to be multiplayer you're going to have multiple users go into it and they're going to be able to instantiate their agent so so like we have kind of like i would just say that we have like there's a there's essentially like a you know a sort of like
15:45 Brad churn happening in the market where there's kind of different opinions about what this thing looks like and that might be one reason we haven't seen like a leader appear yet is because this is just like a thing that for some reason we're not coalescing on like what do we want it to be
16:02 Scott Yes, certainly your point about in terms of what do we want it to be has to do with variances in the software development life cycle. But even in that regard, at a high level, there are still commonalities between everybody's cycle might have variances in detail, but commonalities in overall structure, right? Something akin to like discover plan. implement deploy or something like that right to me what was whizzy no pun intended about this is like i saw a visual that showed me the status of the factory this wasn't the interface where where the engineer was typing and saying let's have the robot do this let's have the agent do that this was a layer on top of that which was like let's just look at the overall status of what's going on here
16:48 Scott And what agents are working on, what agents are not working on. I don't know. It's interesting. I would prefer to buy it.
16:57 Brad yeah yeah totally okay now i'm gonna continue to pull on another thread on that particular sweater called like what what is our factory ui um about gosh it must have been almost a month and a half ago at this point but it seemed like maybe just seemed like yesterday but it was probably like a month ago um some random guy on on x blew up um he he went completely viral on the on the um you know uh developer portion of x where he basically posted his homegrown factory right ui and it's exactly like what it sounds like uh this person steven on your team described or had built where it was basically kind of this combination of like a kanban board and individual tickets on the kanban you could see the little age you know how many agents who was working on it how many tokens they'd spent on it you know and it was this it was basically kind of
17:48 Scott a
17:48 Brad have a ui like this and this guy blew up on the internet because he just like show you know it's like a classic like you know uh whatever like mtv cribs you know it was like he was showing his workspace you know and rate
18:02 Scott Right my rig.
18:03 Brad my rig right and and it blew up and people were like wow that's so cool and then it it kind of blew up because a whole bunch of people basically started replying with here is my one-shot prompt to rebuild this guy because this guy did a video of his his thing and somebody basically like took the transcript of his video threw it into Claude and then made it a one-shot to like anybody can build this guy's rig with my little one you know with this with this prompt of based on his his demo
18:37 Scott That's
18:37 Brad and
18:37 Scott cool.
18:37 Brad so It was a, it was this moment in the zeitgeist, but these things go so fast now where I think everyone was saying like, yeah, anybody now can build their version of it, right?
18:53 Scott Right. And when I look at what these people have built, because the way I describe these is it's a workflow and a series of skills. There's a series of skills in a workflow, right? And those skills have a lot of detail in them, right? They're continuously improved over time. And so while everybody can build their own, you can build your own. I would rather have... I'd rather stand upon the shoulders of giants of somebody
19:30 Brad Yep.
19:31 Scott who's already spent a bunch of time working out the kinks of like, oh. You got to tell the review agent to do this.
19:38 Brad Yeah. And I think that like there's a whole bunch of those out there like you can you can go get a bunch of these in open source at this point. A bunch of people have put
19:45 Scott them The
19:46 Brad out there and the
19:47 Scott skills they put out there.
19:48 Brad skills and the and the and kind of the the you know the here's my open source version of the
19:54 Scott Yeah.
19:54 Brad of the UI to.
19:56 Scott I have a comment about the UI.
19:57 Brad Yeah,
19:58 Scott Oh,
19:58 Brad go for it.
19:58 Scott you were going to say, I think. Okay, go.
20:00 Brad No,
20:00 Scott Sure.
20:00 Brad I'm just I'm just think like, I mean, I hear I guess I'm just kind of a little bit like, meh, you know, like, I, I agree that it'd be kind of interesting to be able to buy something off the shelf. But I have a feeling that this is one of those things that's just so personal, that even if you bought it off the shelf, you'd still want to customize it.
20:20 Scott It's like an IDE, right?
20:22 Brad Yeah.
20:23 Scott I mean, there's, there's, I don't know, I'm just gonna, there's a dozen major IDEs out there. And but very few people are actually rolling their own IDE, they're getting Emacs or Vim and highly customizing it, or
20:34 Brad Yeah.
20:35 Scott VS Code and that's what has made the IDE so delightful is the whole extension model so people can customize
20:41 Brad them Yeah.
20:41 Scott and extensions are shared so they need to get further leverage there.
20:45 Brad Yeah.
20:46 Scott Okay, so I actually have an opinion as we've been discussing this and I've been thinking about it. I have an opinion. The UI for the factory is the Kanban board.
20:57 Brad I think so.
20:57 Scott Because you know why? This is, I'm just stealing from you. It was the Kanban board before. It was just humans, not agents. And that is how we tracked work in the factory. The workers were at cubicles out on the floor. right and they came into a room where we had sticky notes on the wall or put Jira up on the screen or whatever to reflect the status of that work and they told us in voice like oh no I need to move that ticket over to in review because I created a PR and I'm waiting for a response like that's all getting automated but the art the to me it was always that was the structure where we were trying to understand the status of what was happening in the factory
21:39 Brad I agree with that. Now, that is a nice segue to another question mark that I have in my mind about maybe I have an untested hypothesis about why maybe people aren't yet releasing the built version of this. And maybe that's because the Kanban board doesn't work at agent scale. It's a little bit like the pull request.
22:08 Scott Tell me more.
22:09 Brad Well, I don't know if you saw the news, but both Cursors, both SpaceX AI and Anthropik have both ditched Git. Git. They're no longer using Git, and they've both built their own version control systems.
22:29 Scott Oh, brother.
22:30 Brad And the reason they built their own version control systems is because the velocity of change requests is so great that it can no longer keep like your Git is always sending the diffs to every single. Copy?
22:50 Scott Mm-hmm, mm
22:51 Brad And
22:51 Scott -hmm.
22:51 Brad it's like the volume of diffs are now taking so much time and space just to transfer them that the fundamental notion of how Git stores transactions does not scale to agentic scale. Their factories are generating much more changes than can now be efficiently represented using Git.
23:15 Scott It's making me hurt.
23:17 Brad Yeah, like I saw some numbers. It's like they're like they're they're they're averaging something like 20,000 commits a day.
23:28 Scott Wow.
23:29 Brad Right. And so like so so my point would be like. It could be. I don't know. It could be that if you and I are thinking the Kanban board is the right way to think about the factory, then maybe our factories are going too slow.
23:44 Scott That is interesting.
23:48 Brad Does that make you queasy?
23:48 Scott I just don't even know how to process that, honestly, because it's saying, it's forcing me to imagine, like the framework I've been using to imagine AI. augmented coding as our friend mr beck taught us is we're doing the same thing but we're augmenting it and
24:11 Brad Mm-hmm.
24:12 Scott now what that is saying to me is like i don't know this isn't the same thing anymore this is something entirely different and i'm having a hard time it's you know there was that I think it was the author Isaac Asimov who said, or was it Arthur Clarke, I can't remember, said any advanced technology is.
24:37 Brad indistinguishable
24:37 Scott from approximate indistinguishable
24:38 Brad from yeah
24:39 Scott from magic uh and i guess it's getting into that space and i mean i don't know what i've never seen what they're doing over there but that's that's what's going through my head all of a sudden
24:50 Brad yeah and to be clear i mean i'm there's a part of this that i'm kind of like i'm a little skeptical of this like i have to imagine that microsoft or meta across their tens of thousands of engineers are doing 20,000 commits a day across the entire enterprise that'd be two commits per developer like that's no big deal at all right of course they're you know so there's a little part of me that's kind of like i don't know that that scale is all that you know impressive like maybe maybe you're maybe you know you know what i mean like i i'm i'm a little skeptical of the numbers but i But they did these press releases and they were like, oh, yeah, we had we had to do this because we were moving so fast, you know. Yeah, I mean, hey, it's October of 2026. And I'm going to say that for me right now, I think the Kanban board probably is the way my brain can function. And I think that that's a reasonable approach for today. And maybe I'm going to look back on this in.
25:54 Brad a month or six months or 12 months from now and go man cute brad was still thinking about
26:00 Scott Yeah.
26:00 Brad human
26:01 Scott Remember
26:01 Brad scale coding
26:02 Scott way back when we used to have Kanban boards? It's crazy. We knew the status of every change going into the system. Now you just say,
26:11 Brad let
26:12 Scott hey Siri, change my app for me.
26:14 Brad me challenge that for a second though as as a cto of a let's pick a you know the size of fast grab it right so you got like 100 ish engineers there's not a single kanban board that you as the cto are looking at that's showing everything that everybody's working on in the factory right
26:31 Scott No, no, no. Teams have their own. Teams have their own. own boards
26:37 Brad And isn't it the case, I mean, I'll just ask, I'll say for me, like as CTO of a team that's 100 or 200 people, I will admit like I always wished or struggled to see like, where's my fact review that shows me everybody, what everybody's working on there. I never really felt like there was a good tool for that.
26:59 Scott no you had to do it manually this that this is a little bit of a change of subject but the challenge with these tools and you know Jira has kind of become the industry standard the challenge with these things is in theory you can have all these layers and you can have it all roll up and everything it's now maybe if you're like building a 747 or something like that at Boeing like you can make it work but like there's just so much data that has to be so correct at so many layers that when you roll it all up into the top it gets really tough and we do that we use jira very extensively and our tpm team is very in tune with like using that generate all sorts of reports and data and we frequently in fact use ai to correct data or to now tag things what humans are having to do we can figure it out on its own but um
27:49 Scott It's still just so much work to get it all to roll up to be kind of accurate. I've
27:55 Brad Right.
27:55 Scott never seen it really work very well.
27:57 Brad Okay, cool. So I agree with everything you just said. Now, maybe that's the reason that the Kanban board's not the right answer for agent scale, because it's also not the right answer for 100 human scale.
28:11 Scott Yeah, I would say there's two different things going on here, though. The Kanban is the right level for, okay, for work, it's the right level for work unit scale. Like this unit of work, where does it stand in the work process, right? But when you're looking at it, whether it's 100 engineers, 100 agents, 1,000, 10,000, you're not looking at the units of work are different. They're no longer like. this specific change needs to happen to the system and I think this is maybe where it gets fuzzy right like this always drives me crazy when I see I just saw this like You know, I shouldn't say just software. When I see this kind of thing like on a roadmap or a Gantt chart,
29:00 Scott like let's just say like account improvements. Like what's that project?
29:05 Brad Right.
29:07 Scott When's that going to be done? Right. And underneath that is like, well, actually, there's like four or five different changes that we're making to improve account management for customers and they're sort of. If we don't want to have five rows in this list, because we have five rows, we have 100 rows, and it's too much to digest, so we've bundled them into a package called accounts improvements is what we're working on for the next month or quarter or whatever, and we'll tell you when the accounts improvements are done. And I don't know, maybe it's just the type A of me. I always want to know, like, what is the thing that's going to be done?
29:36 Brad Right, and... And but you also have like you have to be able to zoom out. You have to be able to like change elevation. Right.
29:44 Scott Well, yeah, that's what I'm saying is when you start to need to zoom out, you have to do things like account improvements, right? You can't have every ticket. And actually those would just be like maybe five stories and those stories might have multiple tickets, right? So like when you start to zoom out, you have to get into these abstractions and it starts to get more like art than science because
30:03 Brad Well, so I think what people are so now what people are doing is they're using the chief of staff agent to basically give them the summary of how everything is working in the factory. So like it could be that that's the answer is that the actual right way to have the view on the factory is to. Ask your chief of staff agent to summarize everything that's happening.
30:29 Scott You just gave me an idea.
30:32 Brad I don't think it's going to be an original one.
30:35 Scott It's only a little original. It might not be totally original. And it actually has nothing to do with a software factory. I tried to make a chief of staff agent for myself using Claude, and I wanted to be able to go into Slack. So I called my agent Isaac. And I wanted to just be able to go into Slack and say, hey, Isaac, could you check my email and see if there's anything that I should be doing? And I couldn't with Quad and everything, I couldn't quite figure out how to get it all working through Slack. And now to be fair to myself and all the listeners who are saying, you idiot, I spent like two hours on this. I didn't spend as much time on it. I could figure out how to get scheduled things to happen. Cohort does that very naturally in Quad, get scheduled things done. But I just wanted to be able to... throw a shout at my agent go get me a cup of coffee or whatever you know not really get me a cup of coffee not my what did i just have i just want to have a chief of staff i could talk to and it just occurred to me because here where i was going with that i was thinking to myself i kind of had this business idea which is like god in a few years everybody
31:35 Scott is going to have their own little personal assistant right they're going to call Isaac or Frank or Alice or whatever they're yeah
31:42 Brad It's not even first your first your buzzle's bubble Scott, but it's called it's called muse and it's called grok bot and it's called instinct and it just shipped and they're they're valued at a several billion dollars.
31:53 Scott this is before I did all this before those were released but it occurred so maybe maybe they're doing it it occurred to me that When I just think it might be this like factory model is the right approach for that, which is I need essentially put in a ticket, so to speak, for my agent to go do some work. And that work might be like checking my calendar where that's the that might be the approach for those things to work. And maybe that's how that wasn't. I've only used Muse once. I tried it out to test it with TaskRabbit. That wasn't my experience of how that worked. But maybe it is they're
32:31 Brad They're all using the chat model.
32:33 Scott using the chat model and they did a task I don't remember I think I could leave it to go keep on doing the task while I moved on to something else
32:43 Brad Look, I think that you're scratching at a really interesting question that, you know, certainly the last four years, this notion of text-based chat has just dominated. user interfaces for all all tasks and
32:59 Scott Yeah.
33:00 Brad i'm you know it's it's interesting like i i use it i i'm kind of you know like that's that's my coding interface nowadays is i mean i'm not using i use it in cursor you know i use the i use the chat and it's it's generating artifacts for me i look at those artifacts i read through them i interrogate those artifacts i um give it more i chat more with it you know um so and that's the way all these personal assistant um agents have been formed now like the the grok bots and the muses and the instincts of the world they're all just
33:34 Scott they've
33:34 Brad chat
33:34 Scott just
33:34 Brad -based
33:34 Scott spent more
33:35 Brad they're
33:35 Scott time with those things they're chat
33:37 Brad all chat you
33:37 Scott base and i guess they're starting to become more of a person like you might have heard me say a minute ago i named my guy i want
33:44 Brad named isaac
33:44 Scott to be able to say i wanted to be able to say hey isaac can you go do this for me Yeah.
33:48 Brad Yeah. I mean, the anthropomorphizing of them is definitely, it's a big piece of it.
33:55 Scott Actually, you are the person, you might not know this because I've never told you, you are actually the person who inspired me on this because a little while ago you were telling me about how you built your factory with a bunch of your agents and you gave them personalities, you gave them souls.
34:08 Brad Yeah. Yeah. Yeah.
34:10 Scott And that was kind of what sparked my interest. Oh, yeah, I want a person. but
34:14 Brad Yeah.
34:14 Scott I can like talk to not just a I just didn't want to just talk to the computer cloud I want to talk to a person and I want to talk to them in slack
34:23 Brad Yeah, totally. Yeah, exactly. And or over text them, right? Like
34:27 Scott Yeah.
34:27 Brad send them a text message. And Scott, if you had a if you had an executive assistant working for you, how would you how would you tell them to do things? How would you ask them to get stuff done?
34:35 Scott Slack them or send a text message or
34:37 Brad You
34:37 Scott maybe send
34:37 Brad send them
34:37 Scott them an email.
34:37 Brad show.
34:38 Scott Isaac, get me coffee.
34:41 Brad Exactly. So, I mean, again, like I'm still by the way, I'm still trying to answer your question of what's the right user interface for the fact. The factory and there's a part of me that's like maybe maybe the reason that nobody is selling the factory dashboard is because they think they're already selling it to you and it's called chat.
35:03 Scott big
35:04 Brad I mean, I'll tell you like with my fractional CTO clients, like I have them, you know, essentially hooked up to my agent and it's like I come in, I'm like haven't looked at Slack or email, you know, because I've been away from that one client for a couple of days. I'm like, hey, quick, give me a summary. What's happened while I was gone? You know, give
35:24 Scott Mm-hmm.
35:24 Brad me links to all the pertinent conversations that I need to catch up on. And it. Does
35:30 Scott Neither
35:30 Brad that happen?
35:30 Scott that work in.
35:31 Brad Yeah,
35:31 Scott Neither.
35:32 Brad yeah, totally.
35:32 Scott I failed that. I'm now pontificating this, what's the view for the factory question, and I'm going to make a callback, big callback, episode like three, types of learners.
35:47 Brad Yeah.
35:49 Scott And I think maybe the reason I'm fixated on the Kanban board is because I'm a visual person and I just want to see it. I don't want to ask someone what's the status of the projects. I want to look and see it.
36:00 Brad Yeah.
36:01 Scott And so maybe that's why just being able to ask on the command line is not as interesting to me.
36:08 Brad And you know what else is interesting is there's also I mean, this is going to. kind of tie this will tie back into the Kanban board you know one of the things that uh the other sort of like the zeitgeist of like how do you talk to your agents right is people are like I'm not reading the code anymore and I'm getting tired of reading all the words these agents say and so now that most of the skills are like give me your answers in html that include diagrams right so
36:35 Scott Oh.
36:35 Brad you're not alone in the you're not alone in the wanting to see To see diagrams like a lot of people are visual people to understand things
36:44 Scott Okay, I'm going to ask Gemini here, how many CTOs podcast? What episode did they talk about different? I don't know if you know this. One of my absolute favorite books is called On the Back of the Napkin.
37:06 Brad Huh?
37:07 Scott It's about visual learners. I'm sorry, visual explanation. And I think the author's name was Dan Rome. There's another guy in that space, Dave Gray, who I got to know a little bit, who was really individual explanation. We have in the Seattle area, we've got Lila Fever. Yeah, Lila Fever. And so all this visual explanation stuff. And there's a reason. You can boil it all down to this. There's a reason people say a picture is worth a thousand words because
37:40 Brad Yeah.
37:40 Scott it's true. It conveys so much, it can convey so much more information than words. You want to know what episode we talked about types of learners?
37:52 Brad Yeah. Did it come back?
37:54 Scott It did, as a matter of fact. Well, it didn't say what episode. It was released April 7, 2025. How did different learning styles impact team dynamics and decision-making in engineering teams?
38:06 Brad Nice.
38:06 Scott I'm going to ask it, well, what episode number?
38:12 Brad push push the push the agents push the agents or
38:16 Scott And so, and then that way our listeners will know which episode to go back and listen to.
38:23 Brad they could use our new website and they could search for it with the search
38:27 Scott Yeah,
38:27 Brad on the episodes
38:28 Scott season
38:28 Brad there
38:28 Scott one, episode eight. By
38:32 Brad you go
38:32 Scott the way, it says it has nine out of 10 stars. It's a highly rated episode.
38:39 Brad I don't even know how that's possible. That's pretty funny. I don't think anybody rates our episodes.
38:48 Scott I think my mom gives it five stars on every episode. Are you kidding?
38:51 Brad That's great. I'm glad. Let's get more moms listening to the podcast.
38:57 Scott Yeah, we're deep, we're deep, deep in senior citizens.
39:04 Brad You know, here's a funny shout out to the listeners. Somebody left on our Spotify comments. on the uncle bob episode where you said like oh we have 20 listeners and i said we're up to 60 scott we're up to 60 and somebody somebody somebody left on the the only thing they left was 61 and i and i didn't and then the next person left 62 and
39:37 Scott I love
39:37 Brad i
39:37 Scott it.
39:37 Brad didn't i didn't remember that i said that in that episode i was just like why did this this person leaves 61 and this other person leaves 60 i was like this must be some inside joke and then it took and then i like i you know when i updated the website recently and it like kind of read redid all the transcripts and stuff like that and i was like oh now i get it
39:58 Scott I love it. I love it. We're up to 62 listeners.
40:02 Brad A little bit more than that even, but anyway, it's fun that they're out there.
40:06 Scott It's really fun up there out there. Well, Brad, this has been a long and winding episode
40:10 Brad It's
40:10 Scott and
40:10 Brad been a long
40:10 Scott we never
40:10 Brad and winding
40:10 Scott got
40:10 Brad and we
40:11 Scott to
40:11 Brad didn't
40:11 Scott door number three.
40:11 Brad get to door number three and I don't want to go to door number three. We'll have to come back for door number three.
40:16 Scott That's fine. We will save door number three for next week or
40:19 Brad All right.
40:20 Scott some other future time.
40:22 Brad Or some other future time. So, all right. In the meantime, I'll see you next week, Scott.
40:25 Scott Bye, buddy.
40:26 Brad All right. Bye. This episode of the How Many CTOs Does It Take podcast was produced by
40:31 Scott the CTOs of the How Many CTOs Does It Take podcast.
40:31 Brad Produced and edited by Saffron Hep de Gal. Our theme song was AI-generated by Brad. Please follow us on all the social media platforms at HowManyCTOsPod. You can find us on YouTube, LinkedIn, Instagram, Facebook, Threads, X, TikTok, and Patreon. Or at our website at HowManyCTOsPod.com. And remember to rate, review, and subscribe on the podcast platform of your choice. More like your choice. Thanks for listening.