Kurt Schrader: Shipping AI at Shortcut

March 18, 2025  ·  31 min 38 sec

album-art
00:00

As AI transforms the way engineers build software, how is it changing the software that’s built for engineers? On this week’s Reqless, Paul and Rich welcome Kurt Schrader, the CEO and co-founder of the engineering-management platform Shortcut. Topics discussed include what it’s like to integrate AI into engineering-team workflows, why he thinks AI will actually force the engineering skill bar higher in the future, and building Korey, Shortcut’s forthcoming AI tool.  

Show Notes

Transcript

Paul Ford: Hello, I am Paul Ford, and I am the co-founder of Aboard!

Rich Ziade: And I’m Rich Ziade, the other co-founder of Aboard.

Paul: And Aboard is an AI-powered tool that accelerates software delivery for all kinds of businesses, and we’re going to show you all sorts of stuff really soon. But in the meantime, we’re having very exciting conversations about what AI means to the world of software. And that’s this podcast, Reqless, R-E-Q-L-E-S-S. Because who needs requirements anymore? You can just build stuff.

Rich: We should explain that more.

Paul: Well…

Rich: Whenever we say it.

Paul: I don’t know. People are smart and can figure things out from context clues.

Rich: That’s true.

Paul: So look, we have got a very exciting guest today. He has seen so much software get made and is wrapping his head around this whole new world we’re in and also helping his company adapt to it. We’re going to tell you all about that in a minute. But if you have ever made software yourself, trust me, this person’s seen more organizations make more software than you have. So you want to, you want to listen to this. Let’s play that theme song.

[intro music]

Paul: Richard, today on the podcast—

Rich: I can’t believe Reid Hoffman’s here!!!! [laughter]

Paul: Reid hasn’t made that much software. He made LinkedIn.

Rich: That’s true.

Paul: LinkedIn is—

Rich: Actually, that’s unforgivable what he made.

Paul: Oh my God. I paused for a second there, friends. Kurt. Kurt Schrader.

Rich: Welcome.

Kurt Schrader: Good to be here.

Paul: There we go. [laughter] There we go.

Rich: We’re off to the races.

Paul: Always a moment where you got to, like, tear open the can. Kurt, you are the CEO of Shortcut. Shortcut.com. And the reason I was saying you have seen all the code is that Shortcut is a developer platform for people to build software as organizations. I think, rather than me giving sort of the summary, could you give us the summary?

Kurt: Yeah, definitely. So I’m Kurt Schrader. I’m the CEO, co-founder at Shortcut. And we built project-management software for software and product teams to help them move faster and do less work about work, right? So if you want to capture your requirements, it can help you with that, all the way down to what’s actual work you’re going to do and who’s getting things done.

Paul: So you can throw that part of the product away because AI will do it now. [laughter] So wait a minute, you know, help, help the civilians out there in the world understand. Many people, if they’re working in development there, there are a couple big players in this thing, like Jira. Okay? What’s different about Shortcut?

Kurt: So when we started Shortcut, it came off of another company where I was CTO. We used Jira, we used some other tools, and everybody hated them. So step one was, can we build something that people will not hate? [laughing]

Paul: Yeah, that is real. That is one of the quality, that is one of the things about Jira that is pretty universal, and God bless us all, Atlassian, but man, people hate Jira.

Rich: It’s terrible.

Paul: I’ve worked with a lot of engineers and they just never have much good to say. So anyway, go ahead.

Kurt: So yeah, you know, how do, how do we make the engineers happy? How do we lower that bar? How do we move things faster? Just take a lot of the overhead and bullshit out of what you’re doing. And that’s where we started. It’s been about 10 years now.

Paul: Would you say the engineers are happy? I was never able to create that situation for engineers. [laughing]

Kurt: Well, I think some of your engineers used Shortcut. So I don’t know what to say. [laughter]

Rich: Thank you for that.

Kurt: Back in the day.

Paul: That was probably—

Kurt: Back in the day. No, it’s, I mean, will engineers ever be happy? I don’t know, but I think—

Paul: That was the happiest they ever were.

Kurt: We’re going to turn down the dial of unhappiness. Let’s just say that.

Paul: There we go.

Kurt: Like, continually, continually turn it in one direction.

Rich: I’m going to stereotype and you correct me. I mean it’s ticket management. Workflows for teams.

Kurt: Exactly. Yeah, it’s tickets, it’s tasks.

Paul: But really nerd-oriented. Really—because it wouldn’t just be tickets, it would be sort of, like, code commit connected to the GitHub repository.

Kurt: Yeah, I think we were the first ones who built GitHub integration that let you move stories and tickets around automatically. Like, all those different pieces in there.

Rich: Yup.

Kurt: So the less you can go in and have to deal with it, the better, on some level.

Rich: So your hand is on the pulse of a lot of software teams. Right? I mean, you’re seeing what’s going on out in the world and obviously—

Paul: Well, and give us a sense of the scope of Shortcut.

Kurt: Yeah.

Paul: Like, you work with lots of organizations. Like, how would you quantify?

Kurt: Yeah, we have, I mean thousands of companies use it, right?

Paul: And not just for, like, one piece of code. Like, they use it for all their, a lot of their stuff.

Kurt: Oh, yeah. Once it’s in your engineering organization, it usually saturates the entire thing. So you know, we have companies of two people. We have companies of thousands of people that use it to build things.

Paul: So thousands of engineers, thousands of organizations. So tens of thousands, maybe a hundred thousand. Like, lots of engineers.

Kurt: Yes. Yep, yep.

Paul: Okay, so you have a view into this world.

Kurt: Yep.

Paul: Okay.

Rich: I mean, the last, I’m going to say two years have seen a pretty fascinating shift in terms of tools and capabilities that everybody’s pitching in a thousand different directions at engineering teams, at engineers individually, et cetera, et cetera. What are you seeing out there?

Kurt: Yeah, I mean, and you guys have seen this, I’m sure. There’s just a lot of stuff out there. And I think, I think there’s an interesting early take, right? When we started this two years ago, let’s say, let’s go back two years, right?

Rich: Okay.

Kurt: You would get some kind of model—

Paul: When you say “started this,” we’re talking about AI-assisted coding.

Kurt: Yeah, like here, AI is going to help you build things, right?

Paul: Okay, yeah. So you get a model.

Kurt: Yeah, a lot of those early models were very bad at this, right? But they were positioned as, “This is amazing. Like, this is a, here’s this new scene—You’re going to replace your engineering team with—”

Rich: That’s been the promise. [laughter] Yeah.

Kurt: “With a bunch of models.”

Paul: I mean, let’s be clear. We’ve been trying to do this as an industry for 75 years. [laughter] Like, just anything—it never, it’s so sweet that you tried to make engineers happy because every other aspect of the economy is desperate to kill all of them as quickly.

Rich: Sure. Sure.

Kurt: We’ve been trying to do it longer than any of us have been alive.

Paul: Yeah, that’s right.

Kurt: And we’re all old. So, yeah.

Rich: Yeah.

Paul: Yeah. This is real.

Rich: It’s true. The promise has been relentless, even when it was not very good.

Paul: But, you know, it’s because of the magic trick. It’s because when it draws the picture, like, a month later, you look at the picture like, [laughing] “That’s some garbage by AI.” But the minute it draws it, you’re like, “I’ve never seen a cat with nipples on its head before. That’s exciting!”

Rich: Yeah, it’s, well—

Kurt: Yeah, can it do that again?

Paul: Exactly.

Kurt: You know, for us, starting with, “Hey, what do you want to build?” And then breaking it down, we see a lot of teams that are trying to push big tasks or stories, like the kind you used to write. Right? You can give an unbounded, undefined thing to an engineer and say, “Hey, build this.” Right? And you—

Paul: “Fix the API.”

Kurt: Yeah, yeah. “Fix the API.”

Paul: Yeah.

Kurt: And if you give that to any of the AI tools that I’ve used so far, which, you know, it’s probably different today than it was yesterday, so who knows? Just the output on the other end is just trash.

Paul: It’ll take a swing. Let’s actually characterize the trash, because I don’t think people listening, if they haven’t directly worked with this stuff, really understand what—because everyone’s been oversold, right? So when you say it’s trash, you’re an engineering manager who runs an engineering company as CEO. If an engineer came to you and said, “Hey, boss, here’s the code I wrote.” But there was, you know, what the AI did, if they threw it over the wall, what would you say to them about that code?

Kurt: I don’t even know if I’d look at more than a few lines of it before pushing—I mean, it’s, it doesn’t know how you write code. It doesn’t know, really, what the problem is that it’s trying to solve half the time. You know, what we’ve seen internally and across a lot of our customers and what we tell people is, “Hey, if you treat this like a very junior engineer—”

Paul: Okay.

Kurt: If you had a very junior engineer, what would you do? You wouldn’t tell them to fix the API. You would say, “This function doesn’t do the right thing. It’s not named correctly. Do the thing,” right?

Paul: Yeah.

Kurt: So what we’re working on is how do you take this big idea and start to break it down, break it down, break it down, and then say, “Okay, we’ve got ten things to do. These three are obviously human things.” Right? And then if we had seven junior engineers, could we give them these seven things and what’d they get, five of them right? Maybe that’s where we’re trying to get to.

Paul: You know what it’s pretty good at? It’s pretty good at dark mode. [laughter]

Kurt: Is it?

Paul: Yeah.

Kurt: So one of my test prompts is, “Theme this to give me a Miami Vice theme.”

Paul: Well, now you’ve gone too far. [laughter]

Rich: Yeah.

Paul: Now you’ve gone too far. Because what it does, what I’m finding when you do dark mode, so my dark-mode experience is very specific. If you’re in an environment and you’ve carefully architected an AI project, which means it has a component library, like, you’ve kind of made a lot of hard choices about the libraries and the systems you’re using, and you say “dark mode,” and it’s in that context, it will tend to pick the automatic dark-mode tooling that comes with that component library, and it will import kind of all the hard work that got done by somebody else five years ago.

Rich: Yeah.

Paul: And you’ll get your dark mode. If you ask it to implement dark mode from first principles, it’s chaos, right?

Kurt: Fair enough.

Rich: Just to zoom out a second—

Kurt: It’s not going to pull together a whole theming library.

Paul: No, that’s right. That’s right.

Rich: I want to go back to what you said, which is, I got to imagine somebody listening to the podcast that’s hearing about all these amazing app-building tools and code-generation tools.

Paul: [cartoon product manager voice??] “I’m a product manager and I made a game!” [laughter]

Rich: And you’re saying they’re trash. They were trash two years ago. Are they still trash today?

Kurt: So they are orders of magnitude better. And they’re getting better and better. I did see on Bluesky or something the other day where someone was like, “Ooh, the, the video game industry is dead,” because someone had made an Atari game with—

Paul: This is real. Like, they make like—

Kurt: What are we doing? What are these proclamations?

Paul: They make Pong, and then they’re like, “Well, that’s it for Assassin’s Creed.” [laughter] I will say, and just like, breaking news, the Ethan Mollick newsletter came out. We’ve talked about him before in the podcast. And he’s like, “Look, I’m using ChatGPT 3.7.” He got early access. He’s like, “It’s better at coding than ChatGPT-4.” Right? Or whatever—I got, I know the numbers—

Kurt: Claude 3.7.

Paul: That’s the new one.

Rich: Claude 3.7 is better, okay.

Kurt: Yeah, yeah.

Rich: I mean, they’re continuing to get better.

Kurt: But doesn’t, maybe model naming is a good example of how bad this stuff is. [laughter]

Rich: Yeah, yeah, yeah.

Paul: There’s very little product at these companies. Anyway, so, like, the new version of the thing—

Rich: Is it still a junior engineer?

Paul: You’re right. Actually, and just to be clear, Kurt, you’re totally right. It’s Claude 3.7.

Kurt: Yes.

Paul: It’s really good. So is it, yeah, junior—

Kurt: I think it is. I think you have to think of it as a junior engineer. I don’t know where that sort of frontier, that line you have to cross to give it something a little less well-defined. Right? Like, we’re making, I mean, we’re making a bet on this, right? Like, how do we boil this stuff down?

Paul: When you say we, you mean Shortcut?

Kurt: Shortcut.

Rich: Yeah.

Kurt: Shortcut is building a product—so let’s take one step back. We did the thing, first of all, that I think a lot of tools have in it, which is, like, a button that says “write this with AI.”

PaulPaul: Sure.

Kurt: Like, in the UI.

Rich: Yeah.

Kurt: And we already talked about engineers being unhappy.

Rich: Yeah.

Kurt: An AI button given, put in front of a bunch of engineers went poorly? Let’s go with poorly.

Rich: Is that true?

Kurt: Yeah, yeah. They were like—

Rich: What are you doing?

Kurt: Why is this here?

Paul: So you push this into the market—

Kurt: Well, no, we never released it.

Paul: This was internal.

Kurt: It was enough of the feedback, like, “How do I turn this off” on Notion, these sort of things, right?

Paul: I’m not going to ask you to respond to this in any way, but I have to imagine the engineers who go to work at the engineering-tooling company are a special kind of engineer, and I celebrate that. [laughter]

Rich: No, but did you see this with other tools? Like, did people have a bit of an uproar and, like, just sort of rejected it in Notion and Monday?

Kurt: Yeah.

Rich: And all these guys, like, what do you see?

Kurt: They want to see it separate. And there’s another problem for us because we, we have access to code. We have access to a lot of internal data. And people are like, “I don’t want my data leaving the bounds at all.” Right? So how do we—

Paul: I mean, this is real, like, people, you, you manage the critical assets for thousands of companies. That’s what differentiates them from their competitors, et cetera, et cetera. So they’re very vulnerable if that data leaks.

Kurt: Yeah, yeah. And—

Rich: Okay, so you backed off of that?

Kurt: So we threw that in the trash. [laughter] And that was probably a year, year and a half ago.

Rich: Okay.

Paul: But everybody was doing that at that time. So you’re like, “All right, we got to get in the mix.”

Kurt: Yeah. Yeah. And, you know, our board said, “What are you doing?” We said, “I don’t know. Everybody hated it.” They were like, “Really?” And I was like, “Yes. Everybody hated it.”

Rich: Let’s go there for a second. Are boards and advisors of startups or young companies asking right out of the gate, “What’s your AI plan?”

Kurt: Well, so—

Paul: You don’t even have to say asking. Maybe more like shrieking. [laughter]

Kurt: Yeah. You guys don’t have a board, do you?

Paul: No.

Rich: No.

Kurt: That’s nice, just the two of you?

Paul: We’re the board, man. [laughter]

Kurt: Yeah.

Paul: No, it’s great. It’s great for our employees that there’s no oversight. [laughter] It really is. It’s allowed us to—it’s allowed us to iterate in all kinds of ways.

Rich: It’s gonna be great.

Paul: Yeah. One day when we ship our product. Anyway. Keep going.

Kurt: Yeah. I mean, it comes up. We’ve taken a bunch of venture funding over the years.

Rich: Yup.

Kurt: Now all that said, we are profitable now. So we have a, I have a little bit of pushback on all of this stuff. And I said, “We’re working it out.” I don’t think boards—both of you have worked with lots of engineers, so you understand why an engineer would be like, “Don’t give me a button to, like, write a thing because I fucking hate it.”

Rich: Yeah.

Paul: Yeah.

Kurt: But our board, yeah, kind of has ramped up over time. “Hey, what are we doing here? We’re funding things. What’s going on?”

Paul: “We’re seeing growth over here.”

Kurt: Yeah.

Paul: Yeah. “Why aren’t you guys into it? Seems to make a lot of sense.” It’s not, which honestly, like we, we like to make fun of anybody who has power over our destiny, but that’s not the wackiest thing for a board to say.

Kurt: Oh yeah. Yeah. So we’ve started to build a thing off to this. I think the key piece for us was this, treat this like a junior engineer. Right?

Rich: Mmm hmm.

Kurt: We played with coding agents, things where you can give it a task and it’ll do things. And very quickly it was, like, this dark mode?

Paul: Yeah.

Kurt: Okay. You know, these small things, the more we can break things down, the more we can get specific about things, the more we can do small edits to, you know, what the thing needs to do, the better. And then as the agents improve, you’ll be able to get a little bit further away from that. But we want to start with the big idea and then have an agent—I guess that’s the word we’re using for everything now—help us break it down smaller and smaller into a bunch of subtasks and then hand them off to people to run with.

Rich: So essentially the currency in your, in Shortcut is the ticket. And the ticket is sort of the more, the most atomic element. And then you’ve got different ways of grouping them and context and epics and stories and all that. With this tool in flight, you could give it a top-level message or mandate and it will chop it up into tickets and organize them and send them off to teams. Is that the thinking?

Kurt: Yeah, exactly.

Rich: Okay.

Kurt: You flesh it out into the first set of things you want to build.

Rich: Yeah.

Kurt: And then break down, break down, break down.

Paul: So it helps it get more atomic. Which is always a challenge with tickets, right?

Rich: Yeah, yeah, yeah.

Paul: Like, getting everything into the smallest possible components so that people can just go in there and start checking stuff off.

Kurt: Yeah. But there’s still a bunch of pieces that are, like, these are people pieces. Right? Like, this piece is not broken down enough. I don’t know, I’m not smart enough to break it down anymore, and it’s probably not worth it. Like, even if I did that, it’s still going to get it wrong.

Paul: Sure.

Kurt: When I could just give it to an engineer, they spend their four hours on this and the, you know, like the, “change the name of the API endpoint” piece, right? Like, something else can do that. Nobody wants to do that work anyway.

Rich: Yeah.

Paul: All right, so, Kurt, let’s come back to the earlier part of the conversation. Everybody in the world telling me that you’re going to break this stuff down into these little tickets, and then we’ll just have our agents run around and write all the code. And it sounds like you’re telling me, no, I’m going to get to break stuff down in tickets, and then humans are going to have to get involved.

Kurt: Well, there’s two pieces there, right? One is humans need to be involved in the breakdown, because it’s not good enough yet.

Paul: There we go.

Kurt: And—

Paul: No, this is critical. And I think we keep saying it. You’re saying it.

Rich: Is it not good enough yet? It just goes a little off the rails.

Paul: It’s a junior engineer.

Kurt: You got—it’s a junior engineer, junior product manager, right?

Rich: I was going to say, junior product manager is what we’re talking about here, yeah.

Kurt: What we’ve got here, yeah. And you don’t want a junior product—like, it’s even worse if you have a junior product manager giving something to a junior engineer.

Rich: Yeah.

Kurt: Like, that’s just going to, what’s coming out the other side? [laughter]

Paul: My joke is, like, yeah, the food’s not great, but the portions are huge, right? [laughter] That’s—and I think, like, there is an element of, like, you know, it’s, lots of scrambled eggs is fine, like, for breakfast, you don’t want lots, just—

Rich: Not every day.

Paul: Sometimes you got to get in there and cook a little.

Kurt: Yeah. I mean, and you’re trying to avoid toppling the tower at some point, right? If you just build out this stuff on top of each other. We’ve all been in technical debt before, and I was, I was showing you guys earlier some code, on a website that’s clearly AI-generated that instead of a header, it created its own header out of an input tag that you couldn’t—

Paul: Yeah.

Kurt: And it’s like, this is clearly, like—

Rich: A mess.

Kurt: No one in the real world would ever do this.

Paul: What is the product goal? The product goal is to help teams move a lot faster, to make them smarter? Like, what’s your, what’s your goal?

Kurt: Yeah, it’s to help teams move a lot faster and take away all the, you know, you call this Reqless, right? Part of the reason people hate requirements documents, at least in our experience, is you’re just sitting there for an hour, and you got to get the format and the template and do the thing—

Rich: It’s annoying.

Kurt: Fill in the boxes and remember—yeah. [laughter]

Paul: It takes a long time for them to be written in very, a lot of the time is not actually writing the requirements, right?

Rich: No, it’s getting, like, the decimal-dot notation nesting in the—

Kurt: Well, and it’s not even clear if anyone ever reads them, right?

Paul: No, I’ll tell you, I have definitely earlier stages of my career, like, type hierarchy in Microsoft Word and getting the—

Rich: Oh hell yeah.

Paul: Right. Aw…

Rich: That’s the work. That’s God’s work, right there.

Paul: I just, part of me, that’s the most therapeutic thing you can do.

Rich: You know, funny story, when I was in law school, we were studying for the bar exam, and you buy these packets of, like, study materials, they sell them to you. It’s like, it’s a whole thing. But I wanted to create my own documentation, and I just was bored with the law, and so I spent so much time creating Word macros. [laughter]

Paul: This is exactly what—

Rich: My study-mates were like, “What are you doing, dude? This has got nothing to do with anything.”

Paul: The funny thing is, like, the true engineer mindset is like, I will configure everything until I no longer have to do any work.

Rich: Yeah.

Paul: And the amount of effort that goes into being that lazy is how we have a computing industry.

Rich: Yeah.

Paul: Okay, so you’re gonna make people fast.

Kurt: Yes. Faster. Just get rid of all the bullshit work, right? Like—

Rich: Low value, monotonous…

Kurt: There is a level of engagement—

Rich: Yeah.

Kurt: And, you know, the less of this you have to do over on this side of the line, the more you get to do on, like, actual interesting work.

Rich: I wanna look out into the future for a second, but just to put a pin in this, is this out yet? On Shortcut?

Kurt: No, it’s not.

Rich: Okay.

Kurt: And it’s going to be…

Rich: Is there a name for it so people can look out for.

Kurt: It’s called. It’s called Korey with a K.

Rich: K-O-R-Y?

Kurt: K-O-R-E-Y. Yes.

Rich: E-Y. Okay.

Kurt: Yep. I put a list of 800,000 first names into Namecheap, and that one was available for $80.

Rich: Oh, okay.

Kurt: And, you know, and it sounded pretty well, but actually, I think it is working very well. And I’ll say one more thing. We’ve been using some of these agents internally, and they have human names, right?

Rich: Yeah.

Kurt: Like, they’ve been anthropomorphized into, like, part of the team. And that’s helpful. But also, we, one of them is called Devin. We have someone on our team named Devin.

Paul: Mmm hmm.

Kurt: And people get confused about which Devin you’re talking to because it’s in your Slack channel. [laughter] So now we have Korey, but we also have Korey with a K.

Paul: Mmm hmm.

Kurt: Because there aren’t a lot of Koreys with a K that we found.

Paul: Yeah, yeah, that’s good.

Kurt: And you say, “Oh, send that to Korey with a K. Not Corey, the guy over here who is on the marketing team or whatever.” So yeah…

Paul: Rich has actually been a chatbot for over seven years now. [laughter] We just kind of…

Rich: When can people expect this to come out?

Kurt: We want to push it out to people in the next couple months.

Rich: Okay, cool.

Kurt: So at least beta, you know, if you’re a Shortcut user, you should have some access to it at some point soon.

Paul: So I think there’s a thing you have context on that I would really like to kind of close this out with. I want you to riff on it for a minute. So you’re talking about two things here. One is you’re building tools for groups of people to work with AI technologies in order to accelerate, break product management and engineering tasks down into smaller components so that lots of people can interact with those tickets as a result, right?

Kurt: Mmm hmm.

Paul: And that’s Korey. And then B, you are managing as CEO. Are you the CTO, too, or is there a CTO?

Kurt: We have a head of engineering, so.

Paul: Okay, so, there’s a head of engineering. So you, and, but it’s not a huge org. Like, there’s a couple layers.

Kurt: Mmm hmm.

Paul: And you’re managing and getting a lot of engineers and product people and designers to work together to use these tools and collaborate with them as they’re building your own product. Like, they have to call APIs, and they’re probably using some AI coding tools as well. I just kind of want you to talk, because very few people actually have this experience of—most AI is experienced as an individual. It will make you super powerful.

Rich: My assistant.

Paul: Sit down with Claude and here we go. What changes when lots of people are involved in using these tools at once and communicating them to each other and building culture around them?

Kurt: I think one thing we see is back to the junior-engineering thing. If people try to just throw it to the AI and then pass it off as their own work or, you know, “Hey, I wrote this thing up,” and it’s a bunch of bullet points and it sounds suspiciously like, you know, an AI wrote this. It’s like, oh, yeah, yeah… [laughter]

Paul: The sentences are always the exact same length.

Rich: Yeah, yeah.

Kurt: Yeah, yeah, yeah. There’s, I think we’re in a place sort of on the curve, right? Where if you just put AI-generated documentation or, you know, you haven’t done the work to say, “Is this right?” Just put it in front of people? You’ve lost a bunch of credibility already. Right? Like, we’ve had people internally, externally who say, “Hey, here’s a thing, like, here’s all the feedback. It’s summarized.” And it’s, like, what?

Rich: Yeah.

Kurt: Like, you didn’t do anything. You just sent that through the thing.

Paul: I mean, the name of the product is Shortcut. [laughter] But—

Kurt: Just loop it around the end here.

Paul: I think this is real. It actually doesn’t—there’s a thing about work, like, work is human effort that when I perceive it, like, there’s—we just had a conversation about marketing and I said words that I think 20 years ago I would have, like, can’t believe I said, which is, “I just really need a slick, well-put-together deck to tell me this story. Because then I can use that to go tell other people the story.” And so on. Like, I need some structure. I need the labor and the sort of the design work inside of this world in order to be able to do my job. And I think a lot of times people are like, “Well this looks kind of like that. So throw it over the wall and see what happens.” And we can tell.

Kurt: Yeah, well, I mean, and I think there’s a piece here, too, which is a skill-set collapse. Right? Because if you are a product manager and you know how to write up, “Hey, this is what we should build.” And then there’s a piece where you’re saying, “Okay, break that down into tasks.” You know, those used to just be build the backend, build the frontend.

Paul: Sure.

Kurt: Now you need to know enough to say, you know, “What does that mean? How do those pieces look? Where do the pieces come together?” Which implies some engineering skill. And then you have to look at that as an engine, with an engineering brain, computer software engineering brain, and say, “Okay, this makes sense, let’s give it to the agents.” They go off, spin their wheels for a while. Some of them come back and say, “Hey, I’ve finished it, it’s done. Here you go.”

Paul: Mmm hmm.

Kurt: Then you got to be able to look at that code and say, “Okay, does that make sense? Does it actually do what we wanted it to do? Did it break five other things over here?” I think the engineering skill-set bar is going to go up. I think, you know, a piece of it is design. Right? Like—

Rich: That’s interesting.

Kurt: What is that? How does that start to play out, right? So where do you insert people into the equation before you hand it off to the agents? And how do you check the work to say, this actually worked the way we want it?

Paul: Interesting. So you think people need more skills to work with these tools.

Rich: I said this, like, three weeks ago.

Paul: Yeah, but Kurt’s our guest. [laughter]

Kurt: I don’t think I listened—did I listen to that episode?

Rich: Yeah, yeah.

Kurt: Oh, I mean, there’s a, what’s the word? Generalist? But I don’t think it’s really generalist. It’s more specific.

Rich: Yeah.

Paul: Here’s what I, because I’ve been, I’ve been noticing this as well.

Rich: I think this is a fascinating plot twist, with all of it.

Paul: Here is my optimistic case for this technology, because the pessimistic case is pretty much a given, which is you just won’t need as many engineers, and we’re going to have to figure that out as it comes. But the code will do a lot of it, or the AI is going to do all the coding, and there we go. There are cases where that may be true.

But what I’m seeing more and more of is everybody’s getting all the software they need. If they’re kind of already a pretty good architect-style thinker, they can have all the software they want. And suddenly it turns out that you’ve wanted software that you never would have bothered with 10 years ago because it was really hard to build, but you can just go ahead and have it.

Now you have all that software, you’re going to need to manage the software, you might want to share it, you might want to show people how to use it, generating all kinds of interesting opportunities and things going on. I actually think that there’s a new set of skills, but that we might be entering this enormous software glut where suddenly everybody can have all their reports and all their tools and that one little clicker game they always wanted to make, and the database access thing that they wanted to experiment with. They’re going to have all that because this will accelerate them. But at no point do they get to escape from full knowledge of what they’re building.

Rich: Well, I want to close it with a question that pushes back against that very last statement. Do you think the day will come where you will generate this nicely organized set of tickets that, in aggregate, represent some business app that somebody’s building, and then that set of tickets will be handed back into the machine and it will produce the app. Is that day coming, Kurt?

Kurt: What do I think—

Paul: Well, because you have the platform—

Kurt: Well, it depends on what that—I mean, so here’s something we’ve noticed, and I’ll ask you guys this, which is we have a bunch of agents we’re playing with, right? Some of them are writing code. Ours is writing tickets. And if you are writing tickets and breaking it down and reading them, you can get to a 90% right case and it’s super valuable, right?

Rich: Sure.

Kurt: Like, it’s like, this is great. Like, I just saved four hours. I’m gonna go over there, I’m gonna go for a run, I’m gonna go drink a beer, whatever it is, right? Like, you’ve just saved a ton of time if you hand that off to, let’s see, we were using an agent that looked at your financial data, right? And said this is how much AR you added this week. This sort of thing and coding agents are similar, but the financial agent is a good example. If that gets something wrong one time, you’re like, “Fuck this thing.”

Paul: Yeah.

Kurt: Because you never trust it again. Right?

Rich: Yeah.

Kurt: Like, so we’re in a space where you get to the 90%, you’ve saved a bunch of time. Let’s do the work, you know, let’s write this code that implements this thing. Got it all wrong. I’m going to just try the next agent, or I’m going to go somewhere else. And like, it just doesn’t take a lot when you’re that, that binary true-false.

Paul: Yeah.

Kurt: To like—

Rich: And you just lose all trust.

Kurt: Yeah, yeah, exactly. Yeah.

Paul: So what I’ve noticed with programming with this stuff is that the impulse to laziness is unbelievably powerful.

Rich: Sure.

Paul: Because you fix it and it kind of fixes it and so on. But I actually find that when I’m really focused and leaned in and getting very good results, like, I’ve built a prompt toolkit that’ll get me to a pretty good API from a relatively complex schema in about a half hour. But I gotta be on for that half hour. I can’t actually turn away. And I got to really kind of look at the code. I don’t have to look at every single line, but I got to be focused. And if I’m focused, I can get the result I want. If I just keep slopping it together, it breaks real fast.

Rich: It gets worse and worse.

Kurt: Yeah, it’s trajectory, right? Like you, if you’re pointing at the moon and you’re off by two inches, right?

Paul: Yeah.

Rich: Yeah.

Kurt: You end up in the sun and you’re like, “How did I get here? It doesn’t matter.”

Paul: We created an accelerated toolkit for creating technical debt with these, with these technologies. You can create—

Rich: I think that’s the funny double-edged sword here. Is that used well, definite productivity boost. Used poorly, the technical debt that’s going to mount is going to be massive.

Paul: It’s instant! It’s instant technical debt.

Kurt: The first, first four lines of code are just wrong. And it’s just building on top of that.

Rich: It literally educated itself on mounds and mounds of technical debt.

Paul: That’s right.

Rich: [laughing] It’s actually based on technical debt.

Paul: All right, so let me, let me ask you a final question.

Rich: I thought I had the final question!

Paul: We’ll go for it. Mine’s very short.

Rich: No, I already asked my final question.

Paul: Okay, so my final question is very simple, which is, as you’re growing and hiring, are you looking for people, engineers, product managers, et cetera, who are specifically skilled and focused on these technologies, or does that not really matter yet? We’ll bring them in as they come.

Kurt: I… We’re not. We actually, my entire team was polled internally. Some of them work—one of them has worked for me for 10 years now, so, and we worked together at our last company. And I think the tool set is, do we think about how this is going to work with users? How do we build up—we’ve been saying product engineer—

Paul: Yeah.

Kurt: Come up for a while now, but, you know, is this going to work for users and get me where I need to go? And there is a big piece of, like, I’m a senior engineer and I can see the technical debt coming from a mile away. And, like, I’m going to back out, I’m going to back the truck up before we go down that road.

Rich: Yeah.

Kurt: Because what I don’t want is a junior engineer working with a junior engineer, and then you’ve got, you know, just a pile of stuff over here that you got to completely redo.

Rich: Right.

Kurt: So we have been, you know, I don’t think the technology itself is that complex. Right? Like, you got an LLM, you write some prompts, like, that sort of thing. Like, if I was building foundational models, I might be hiring for something else.

Paul: Yes.

Kurt: But, like, we just want to build great software that people can use and move fast with. And for me, that’s, you know, do you think about your users? How do you, do you define success as the users are saving time, they’re moving faster, and, you know, you can figure out the model stuff behind that. It’s nothing any different than stuff I’ve seen over the years.

Rich: Yeah.

Paul: Do you feel that your customers are around the same place, or are they, like, you know, you’re seeing a big swath of the coding world. Are they leaning into this more? Less? Around where you are?

Kurt: We don’t have a ton of visibility into that, right? Like, we see—

Paul: As it should be. [laughter]

Kurt: Yeah. There’s the rules and laws and, you know, internal security things.

Paul: Yeah.

Kurt: But I think people are really still figuring it out. Right? They’re still saying—

Rich: Poking around.

Kurt: What’s going to happen if I turn on Copilot? Like, what, you know, how does this work? Like, and half the team uses Copilot and half of it doesn’t. What does that mean for me?

Rich: Yeah.

Kurt: And it’s like, well, it means—

Rich: It’s a little Wild West.

Kurt: Yeah. I think it’s so early. Yeah.

Paul: It’s the classic internet quandary where everyone is telling you that you’re ten years late to the two-year-old technology, and it’s actually extremely early days and you should approach tentatively and figure it out as you go.

Rich: Totally.

Paul: So, Kurt, thank you for coming on.

Rich: This was great.

Paul: This has been a wonderful conversation.

Rich: Look out for Korey.com with a K.

Kurt: Korey dot…AI.

Paul: Ohhhhhh!

Kurt: Ahhhhh.

Rich: Sorry. K-O-R-E-Y dot AI.

Kurt: Yeah, the team’s going to be mad at me. I don’t know what happens if you go to that domain right now, probably.

Rich: Yeah, don’t go there yet.

Kurt: Yeah, we’ll figure it out.

Rich: In a couple months. All right. Thank you so much, Kurt.

Kurt: Thanks. Thanks for having me. This was awesome.

Paul: All right, Rich. Kurt’s seen a lot of code.

Rich: You know, it’s a recurring theme. The guests we’ve had on are figuring it out. They’re kind of like anthropologists.

Paul: Well, this is what I like about our podcast and why it’s so special and wonderful.

Rich: Right, we figured it out already.

Paul: No. It’s that, unlike everyone else in this misbegotten industry, we’re willing to say nobody knows what’s going on.

Rich: No, you know, my favorite is? [monster truck rally announcer voice] “120K ARR IN 30 DAYS BECAUSE OF AI!”

Paul: Yeah, I mean—

Rich: “I MADE A CHROME ADD-ON!”

Paul: You know, once I left Twitter, I stopped seeing those. You just, they just don’t happen—well, no LinkedIn a little bit.

Rich: No, I see them on LinkedIn, yeah. Anyway.

Paul: Anyway, if you need us. Hello@aboard.com.

Rich: Hello@aboard.com.

Paul: And really, that’s about it. Lots of big news coming. Similar to Kurt, we’re getting our stuff together and we’re excited to share it with the world.

Rich: Have a lovely week.

Paul: Bye!

[outro music]