"I think when you're getting these efficiency gains, you should always think about them as reinvestment opportunities where you can improve the quality of the product, you can explore more new possibilities for your product."
Is your company using AI to trim your budget, or to multiply your team's impact? We're joined by Matt Culver, a senior engineering leader at Super.com, to discuss why the common view of AI as a tool for cost-cutting is a misguided "accounting mindset" that ultimately destroys trust. He argues that leaders should instead see efficiency gains from AI as a powerful opportunity to reinvest in their teams. This conversation reframes the AI debate by urging leaders to look beyond the coding loop to improve the entire product development lifecycle—from ideation to delivery.
Matt explains that the key to successful AI adoption is aligning new initiatives with the developer's core incentives: removing friction and enabling the creative flow state that makes their job enjoyable. He provides a human-centric approach for channeling AI's power to solve upstream problems in product planning and market research, rather than just generating more code. Learn how to use AI not as a means to an end, but as a way to empower your developers, generate more value, and build a high-trust engineering culture.
Show Notes
Transcript
(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)
[00:00:00] Ben Lloyd Pearson: Matt is his senior engineering leader@super.com, where he champions the principles of data-driven servant leadership. He is recognized, he is a recognized thought leader in developer experience management and the science of measuring engineering organizations. Matt, welcome to the show.
[00:00:16] Matt Culver: Thanks. I haven't heard that read out before, but it sounds all sounds very impressive.
[00:00:20] Ben Lloyd Pearson: We, we like to hype up our guests here on Dev Interrupted, so, but hey, let's dive into, we, we, we had, uh, chatted a little bit. I'm really excited to, to talk about some of the things that you're here for. Some engineering leaders view ai, which is, you know, it's a topic that everyone is talking about these days. but a lot of engineering leaders view it as a way to like save time. Right? And as a result, um, a lot of organizations view it as a way to reduce headcount often, but you have a different opinion and you think that might be misguided in a lot of situations.
[00:00:52] Ben Lloyd Pearson: So let's start there.
[00:00:53] Matt Culver: Yeah, so my take on it stems from, um, something I've been thinking a lot about as I, [00:01:00] as I kind of study and admire other companies that have become ultra successful. In particular, NVIDIA really focuses on a metric of revenue leverage per employee, and they have one of the highest proportions of this in the world.
[00:01:13] Matt Culver: And so
[00:01:13] Matt Culver: when I started thinking
[00:01:14] Matt Culver: about the value that AI brings to our organization, um, and hearing a lot of the conversations from, uh, different people interested in the value of AI generating, or improving efficiency or, allowing us to produce more lines of code, that kind of thing was, this shouldn't be a conversation about, um.
[00:01:34] Matt Culver: Tuning for efficiency. This should be a conversation about, uh, how we can take the time that we gain back and reinvesting it or how or how can we leverage AI to make one developer have a much higher value generation. Leverage essentially is the way I think about this. Um, and every time people get myopically, um, sort of focused on the development loop, um, I immediately start to think.[00:02:00]
[00:02:00] Matt Culver: What about all the other, other inputs to this process? Uh, because the value generation loop, uh, the end-to-end product development lifecycle, if you will, in those software engineering companies, starts with a human writing an idea down, uh, articulating that idea, uh, or, or looking for an opportunity in the market for something they should be
[00:02:19] Matt Culver: creating. And that process is time consuming and that process is complex and that process is often wrong. There's a lot of opportunity, I think, to apply AI there, considering again, the whole value generation cycle. Um, and some of those problems are a lot easier to solve too. like
[00:02:35] Matt Culver: again, we can make, you get, you get like just a, a straight one time, um, step function improvement.
[00:02:42] Matt Culver: Like developers are 20 or, you know, get 20% of their time back and that's great and that is valuable, but is that as valuable as, hey, we cut the entire. Total product lead time from having a notional idea about a product to being able to write a spec, to be able to get that spec ready, broken down [00:03:00] into tickets and estimated.
[00:03:01] Matt Culver: Then start writing the code if that process took us 20 days before, and it takes us 10 now, that's crazy valuable as well, right? Because that means that, that it, it, it can start realizing its revenue potential earlier. And so
[00:03:13] Matt Culver: when I look at this problem, I'm really still trying to consider it as a holistic problem where, we have to just throw away the old paradigm and reinvented it in a new AI first paradigm where those, those, the, you know, your, your product.
[00:03:26] Matt Culver: Developer or owner, you're your developer writing the code and your product
[00:03:31] Matt Culver: designer
[00:03:32] Matt Culver: are all sitting down collaboratively now. Uh, starting with AI first to generate an artifact they work from, instead of working in, in, you know, parcel, uh, work streams, uh, through Jira tickets in that traditional way that we have for, however many foot.
[00:03:48] Matt Culver: How long has Jira been around? Like, I don't even wanna know, but yeah.
[00:03:52] Ben Lloyd Pearson: Yeah. And I, and I think you, you're hitting a, a really great point 'cause, you know, with ai, it, this is, this is a, a, a top a theme. We've, we've hit many [00:04:00] times on this show. Uh, our listeners are probably sick of hearing it at this point, but with AI, you can do more than ever before. And if you're doing the right thing, that's, that's amazing.
[00:04:08] Ben Lloyd Pearson: You're doing more of the right thing than you've ever done in the past. But if you're doing the wrong thing, we're doing something that doesn't have value to your organization. You're
[00:04:14] Ben Lloyd Pearson: just doing.
[00:04:15] Ben Lloyd Pearson: more waste. Than you've ever done before. and really what you want is for that more to result in more impact coming out of the engineering organization.
[00:04:26] Ben Lloyd Pearson: So, um, you
[00:04:27] Ben Lloyd Pearson: know, if you're, if all your engineers are doing
[00:04:29] Ben Lloyd Pearson: is using AI to, to migrate some deprecated system that nobody, none of your customers use anymore, then, then, then why are you even, why are you even using, using AI to begin with? You know?
[00:04:40] Ben Lloyd Pearson: So I, I really appreciate that you have like
[00:04:42] Ben Lloyd Pearson: a more of a, a focus on like making sure that that AI is being.
[00:04:46] Ben Lloyd Pearson: It's helping you reinvest
[00:04:47] Ben Lloyd Pearson: into the organization in a, in a more productive way.
[00:04:50] Matt Culver: Yeah, like I, I pe people keep asking for, I, I think a really, uh, interesting thing that maybe we should be a lot more vocal about that we're coming to realize is, [00:05:00] um, our journey as a company, has been one where we have you see what our, uh, one of our original founder founders, um, uh, talking about this all the time.
[00:05:11] Matt Culver: AI has been in our, it has been in our ecosystem, and we've been the like aggressive bleeding edge or early adopters. And as a result we've been able to ship more and more product growing our company every year by this huge, huge margins. But, and we have not been growing the number of engineers doing that work every year.
[00:05:29] Matt Culver: We've been relatively static as a headcount for the, the past several years. Not, not as, not as a strategic cost saving decision, but, because we were able to get more and more and more value generation leverage because we were getting smarter at like our entire tooling pipeline and removing friction for developers and using tools like LinearB for developer enablement and
[00:05:51] Matt Culver: Speeding up
[00:05:52] Matt Culver: that, that iteration loop and getting stuff out of people's way.
[00:05:56] Matt Culver: AI is just another tool having arrived on the scene [00:06:00] that allows us to further optimize that.
[00:06:02] Ben Lloyd Pearson: Yeah. And, and let's, let's dive into that. 'cause you've emphasized that, you know, any change that, that you implement that impacts your engineer ultimately needs to,
[00:06:11] Ben Lloyd Pearson: to improve
[00:06:11] Ben Lloyd Pearson: developer experience, you know,
[00:06:13] Ben Lloyd Pearson: that should always be
[00:06:14] Ben Lloyd Pearson: your goal.
[00:06:14] Ben Lloyd Pearson: So how do you see AI reshaping the way that organizations prioritize
[00:06:18] Ben Lloyd Pearson: developer experience?
[00:06:20] Matt Culver: There's actually
[00:06:21] Matt Culver: kind of a hazard
[00:06:21] Matt Culver: here that I've spent a lot of time thinking about this week, which is. When I, what I'm hearing and what I'm seeing is that the interaction surface area between a developer and the AI in the, in the development tool chain right now, uh, isn't making developers jobs more enjoyable and more interesting, isn't It isn't in many cases.
[00:06:43] Matt Culver: In fact, uh, yeah. It's, it's not making their job better. It's actually making it worse. And we have, uh, a lot of cases where this is true, for example. A really common hazard that I'm seeing all over the industry is someone will take ai, [00:07:00] they'll point it at their code base, and they'll ask the following question, what's wrong here?
[00:07:06] Matt Culver: Right? Uh, and, and you know so much, and you know, and you know what the AI's gonna do, the, what you just accidentally did was set a constraint. What you said was you have to give me an
[00:07:18] Ben Lloyd Pearson: yeah.
[00:07:18] Matt Culver: and you better give me a plausible answer, and so it's gonna give you a plausible answer and you're gonna spend a lot of time trying to validate it.
[00:07:25] Matt Culver: Coming
[00:07:26] Matt Culver: to realize only later that if you had approached this conventionally, you would've produced a much higher quality, uh, engineering result. and it's not to say that AI isn't gonna get better at this, it absolutely certainly will, but these kinds of behaviors are gonna still be there. Um, and so, uh, something that I'm having to constantly coach people on is what you should have asked was.
[00:07:47] Matt Culver: Here is what here is what I'm trying to do based on some investigation I've done ahead of time, or help me investigate how this behaves. Okay, I have a notional idea of how I'd like it to behave that I think is better. What do you think? Great. [00:08:00] And that, like you, you kind of use it as a, as this like interrogative and collaborative process to enhance your intuition, your, uh, direction, not.
[00:08:11] Matt Culver: Give me the answer, give me the answer. Gives you The dataset it was trained on, which is going to give you a globally mediocre average answer in most cases. Um, and we see this with AI code generation, right? Like it's 4x or 10x uh, test writing. You're gonna use this everywhere. But then you look at the tests and you're like.
[00:08:30] Matt Culver: Did you instrument every single condition of testing all of those endpoints separately instead of just
[00:08:36] Matt Culver: writing something that
[00:08:37] Matt Culver: was reusable logic? Uh, and, and the AI's like, yep, I sure did,
[00:08:42] Ben Lloyd Pearson: I mean, if you can generate the code in 10 seconds, like why not
[00:08:45] Matt Culver: Yeah. There's no cost to it. There's, there's no cost to it, except for, you know, the global depletion of all drinkable water, but.
[00:08:50] Matt Culver: We'll get into that later. But yeah. Uh, like it's, and I think fighting against these patterns is gonna be something interesting, um, in how we like train people to [00:09:00] use the tools effectively to remen remember to be rigorous, uh, to remember that some of those fundamentals are st are, are gonna be even more important than they were in the past.
[00:09:08] Matt Culver: Right? Like, uh, uh, peer review of the code that you write is now going to be a harder problem because AI might write four times as much and you weren't in the loop when the logic was being generated.
[00:09:20] Ben Lloyd Pearson: Yeah. And, uh, you know, sort, sort of tying, you know, some of the points we've been making together, it's like, not only do you want to focus on reinvesting time into making sure engineers have more impact, but if you're overwhelming your, your engineers with AI slop, you know, for the, the phrase that, you know, a lot of people apply to this.
[00:09:38] Matt Culver: Yeah.
[00:09:38] Ben Lloyd Pearson: and making their job worse. You're, You're,
[00:09:41] Ben Lloyd Pearson: You're
[00:09:41] Ben Lloyd Pearson: actually potentially making your system less efficient rather than more efficient. So by doing more, you're actually slowing things down effectively. Um, and I think, and I think that's a, you know, a lot of organizations don't always account for, for the, the new bottlenecks that they, they will create by not [00:10:00] properly, like adopting AI within their organization, And, and speaking of adopting ai, you know, it's not just about tooling, you know? Um, and, and we've been saying this a lot, like, you know. You, you shouldn't seek to adopt a tool with ai. You should seek to solve problems and then find the AI tool that solves that problem and apply it appropriately.
[00:10:20] Ben Lloyd Pearson: Um, and adopting ai, it's, it's, it's a very deeply human process. It affects humans, humans have to use it. So what are some practical steps that you think engineering leaders can take to, to make sure that you have buy-in from your engineering team? That you have alignment across the engineering, um, particularly when you're rolling out like AI powered workflows.
[00:10:40] Matt Culver: Yeah. Um, we were chatting just a little bit before, before, uh, this interview, and, and I was, I was sharing that, like I came to this realization, um, having been here for ELC, that, um, it's the, it's still the same, uh, problem to solve, or it's still the same methodology because there's [00:11:00] still human beings involved and if you don't align incentives to the developer.
[00:11:04] Matt Culver: Primarily as your goal in all decisions you make at the organizational level, whatever you're trying to do will fail no matter what. It's, because you know what drives organic adoption and what sticks as culture is when you're like, Hey, I removed annoying friction for you so you can enjoy your job better so that like more of your day can be spent in the, in the like wonderful flow state
[00:11:29] Matt Culver: dopamine loop creator. That makes us all happy to.
[00:11:32] Matt Culver: do our jobs.
[00:11:34] Matt Culver: Anything that that works against that is going to, is going to hurt you. And eventually you'll just have like a, a slow degrade of resistance where people are like, oh, this has some novel use, but it's not gonna stick. Right? Like, they're not gonna keep coming back and using it as a tool if you haven't aligned those incentives.
[00:11:50] Matt Culver: And so when we first started measuring organizations for things like cycle time, we had to make it a non-punitive process. We had to say like, Hey, this, this metric is representative [00:12:00] of. when you as a, as a, an engineering manager or as an IC get into a healthy pattern, and we had to show, you know, quantitatively because they're like, everybody who writes code is somewhat of a, somewhat of a math nerd.
[00:12:13] Matt Culver: We had to show quantitatively like, you know how, what, you know how you're enjoying yourself, and I have data that shows that you're enjoying your job more and you feel more empowered. Well, guess what? At the same time we got, like, those loops got shorter. And that's not an accident,
[00:12:26] Matt Culver: right? Like, uh, happy developers make great stuff and happy developers are generally more productive.
[00:12:33] Matt Culver: Uh, same thing here, right? Like if we, if we force every developer to become. Just a, like a, a line cook that makes engineering specs that an AI writes the code for. I'm not sure anybody's gonna want that
[00:12:45] Matt Culver: job.
[00:12:46] Matt Culver: Uh, so we really have to think harder here now. Like we're at a pivotal point where, um, we need to think about what the, the human interaction, uh, mode is for working with ai, but we'd better align with the people who are gonna use it before anybody else, right?
[00:12:59] Matt Culver: [00:13:00] Like, don't align with your accounting team first. Align with your developers.
[00:13:04] Ben Lloyd Pearson: Yeah. So I'm just curious, you know, 'cause I know that you are, uh, you know, you are an organization
[00:13:08] Ben Lloyd Pearson: that really does
[00:13:09] Ben Lloyd Pearson: look heavily into
[00:13:11] Ben Lloyd Pearson: you know, both qualitative and quantitative metrics. So what, what have you seen as an organization while you've been on this AI journey? Like, has there, has
[00:13:18] Ben Lloyd Pearson: there been changes to your metrics that
[00:13:20] Ben Lloyd Pearson: have been expected or unexpected or that have surprised you?
[00:13:24] Ben Lloyd Pearson: Uh, how, how has that been?
[00:13:25] Matt Culver: probably the most interesting thing is our, our. Learning velocity has gone up. I wouldn't say like there's not been a, like, because we were lucky, we did turn on our AI adoption happened very fast, a very short period of time.
[00:13:41] Matt Culver: So there is a pretty clear demarcation point where there was like the pathway and the new way. What does the data look like between them and other than net lines of code going way up. which does not equal, like, that doesn't mean we generated more
[00:13:55] Ben Lloyd Pearson: usually terrifying
[00:13:56] Matt Culver: might,
[00:13:56] Ben Lloyd Pearson: And that happens,
[00:13:57] Matt Culver: you know, it, it, it keeps me up at night for [00:14:00] sure.
[00:14:00] Matt Culver: But we
[00:14:01] Matt Culver: were able to, to, uh, increase our experimentation velocity a little bit. We were able to handle more,
[00:14:07] Matt Culver: uh, discrete work streams.
[00:14:09] Matt Culver: Work
[00:14:09] Matt Culver: streams per quarter
[00:14:10] Matt Culver: related to a product or feature change. And so. We've just gotten a lot better. if you look at our experimentation velocity as an organization, the only way we're able to do that is we have so many intelligent little tools all throughout our process that are assisting us so that I don't need to go mind, you know, the thing that's currently running, I just get like a push notification and this is where like automation generally, but AI enabled automation is, is really starting to like help that process.
[00:14:38] Matt Culver: Um, and I think, yeah, like I, I wouldn't say, and I think. Anybody who tells you otherwise is, is probably telling you a lie. Uh, but over and over and over as I talk to every other leader, the, the consensus is we're still all really figuring this out,
[00:14:53] Ben Lloyd Pearson: Yeah, absolutely. Everyone is, yeah. So you
[00:14:57] Ben Lloyd Pearson: you've mentioned this accounting mindset a couple [00:15:00] of times now. We've kind of dabbled in it a little bit. So I want
[00:15:02] Ben Lloyd Pearson: to, I wanna address
[00:15:03] Ben Lloyd Pearson: it a little more directly, you know, 'cause, ' there's a few mindsets about how to measure, like developer productivity or developer experience, that, that kind of thing.
[00:15:11] Ben Lloyd Pearson: Um, you know,
[00:15:12] Ben Lloyd Pearson: efficiency
[00:15:13] Ben Lloyd Pearson: always comes up. Like if you're lucky, your organization also thinks about quality as well. At least you should. But let's think about this accounting mindset. Like why, why is that approach doomed to fail, and what should you do instead?
[00:15:28] Matt Culver: it's doomed to fail because it, it, it, it's a fundamental misunderstanding of the value that, that, um, AI is generating. But, but, but previous to ai, it's a fundamental misunderstanding of thinking of a developer generating code as a cost center of any kind. Um,
[00:15:46] Ben Lloyd Pearson: yeah.
[00:15:46] Matt Culver: they generate the capital asset that generates revenue, right?
[00:15:49] Matt Culver: The sooner it becomes real, the sooner that revenue generation potential can happen.
[00:15:53] Matt Culver: and our
[00:15:54] Matt Culver: job as. software engineers and as companies that make software products
[00:15:58] Matt Culver: is to get
[00:15:59] Matt Culver: the, [00:16:00] you know, the successful and the useful product into our customer's hands. I think when you're getting these efficiency gains, you should always think about them as reinvestment opportunities where you can improve the quality of the product.
[00:16:12] Matt Culver: You can explore more new possibilities for your product. you generally are improving the, the value generation leverage of the engineer on your team, and you might be able to have slightly
[00:16:22] Matt Culver: smaller atomic unit teams, which would be that, that certainly will make people who think of. The human capital cost as a, as a, as a problem. Uh, wherein like now, because we were able to deliver so quickly where, where you would have
[00:16:37] Matt Culver: four developers,
[00:16:38] Matt Culver: and 1:00 PM and one product designer, that new atomic unit I, I think, might shift towards a team that's more like two full-time engineers because they're able to do a lot more.
[00:16:48] Matt Culver: We've removed all this friction, but thinking of it as like. I can fire 20% of my staff if I'm 20% more efficient isn't true because they, they also like the,
[00:16:59] Matt Culver: they're [00:17:00] all doing
[00:17:00] Matt Culver: things that interlock. So if you remove one person, you've just hobbled the team. In reality that, that was,
[00:17:09] Matt Culver: that was
[00:17:09] Matt Culver: a poor choice for two reasons.
[00:17:11] Matt Culver: Um, and I think the other reason is like we have to be like, I'm a very human centric leader. I, I. I really think that building trust is hard and destroying is super easy. So a really great way to destroy that trust is to start, is to start firing people because, uh, especially the software developer, because you think AI's getting you some efficiency, I guarantee you it hasn't.
[00:17:31] Matt Culver: There's a tax in there that you haven't accounted for.
[00:17:33] Ben Lloyd Pearson: Yeah. Yeah. And, and I, and I've, I've shared this, this anecdote a lot on, on Dev Interrupted episodes, but I I was once on a team that was measured by the number of commits and the number of lines of code that we changed. Like that was, that was our KPI, you know, and, uh, it's, it's amazing how easy, even before ai, it's amazing how easy it was to just fake that metric.
[00:17:52] Ben Lloyd Pearson: And in the, the age of ai, it's, it's so much easier now. Like you don't even have to think about it actually. You can have agents that just do it automatically.
[00:17:59] Matt Culver: Yeah, [00:18:00] we'll just name the agent, Goodhart's agent and it'll just sit there in the background and make a hundred billion tiny nails.
[00:18:06] Ben Lloyd Pearson: yeah. Yeah.
[00:18:08] Ben Lloyd Pearson: you know, so given that, you know, LinearB is like the, the, the sponsor of Dev Interrupted, and I know
[00:18:13] Ben Lloyd Pearson: that you're a customer of LinearB, I'd be remiss. if I didn't ask
[00:18:16] Ben Lloyd Pearson: uh, about how your journey with LinearB has, has been, because, you know, I know that, uh, you know, we've been talking a lot about metrics.
[00:18:23] Ben Lloyd Pearson: LinearB has been a big part of your journey with AI and, and just even beyond that, you know, so, how is that? Played a role with within,
[00:18:31] Ben Lloyd Pearson: like, both AI
[00:18:32] Ben Lloyd Pearson: adoption, like both before
[00:18:33] Ben Lloyd Pearson: and after
[00:18:34] Ben Lloyd Pearson: in developer experience as a whole? Like what, you know, how has the, the metrics and the AI automations, like, how has that played a role within super.com and all the changes that are happening this year?
[00:18:44] Matt Culver: Yeah, I mean, pre ai it was a, a use, use of, you know, de developer productivity tool. Uh, LinearB in particular was a huge cornerstone of the culturally built around, uh, developer led self-improvement. [00:19:00] So we have, uh, we have like a metrics review process. That is a coaching session with the developer and the product manager.
[00:19:08] Matt Culver: That happens completely exclusive of the performance management process.
[00:19:12] Matt Culver: It is like, you know, it, it is meant to be, non-punitive. It is purposefully decoupled for the purpose
[00:19:19] Matt Culver: that,
[00:19:20] Matt Culver: when we got people to buy into it as a, as a like a foundational cultural element, like I am doing better at my job. If I'm great at applying the resources measured as.
[00:19:30] Matt Culver: You know, capacity, accuracy, and planning accuracy. And if cycle time becomes a very important thing to my team, enabling me to do stuff like, Hey, this is a little bit long running, that is an opportunity for me. Like, you know, you know, WorkerB comes into the channel and is like, Hey, this thing's long running that isn't meant to be a punitive performance.
[00:19:49] Matt Culver: Uh, trigger that is meant to say, do you need help? How can I unblock you? This is a learning moment. Do you wanna pair? Um, and when we really ingrained the, the, how, the, the like [00:20:00] quantitative measure related to the behavior that made you better at your job and, and be able to deliver faster and be able to be more effective as a contributor to that team, it was a sea change.
[00:20:09] Matt Culver: Like we were able to go from, when I started the data within the organization, we were cycling on average in like a hundred plus hours for any given piece of work. And today, the org as a whole is sitting at, I think we're sitting at 38 hour cycle times. In a world of AI that's still, that metric is still gonna be indicative of our value generation cycles.
[00:20:33] Matt Culver: And I imagine we'll just see, like, you know, again, we'll be able to shrink it a little bit more because all the, uh, all of the inputs to cycle time or, or total product lead time just become more efficient. So, um, we're, as we introduce new stuff and things are changing very fast with the AI tools, we're able to see that, that we're getting that kind of like additive value to that process and those cycles happen faster.
[00:20:56] Ben Lloyd Pearson: Yeah. And, and what I, what I really, I think appreciate most about, you [00:21:00] know, just everything you've shared with us today is that, I think really what's sort of at the center of the culture you build is this high trust environment. You know, as You said, trust takes a lot of effort to build, really easy to destroy. Um, metrics are probably one of the most effective weapons at destroying trust within an engineering organization if they're wielded
[00:21:18] Ben Lloyd Pearson: wrong, you know,
[00:21:20] Ben Lloyd Pearson: um, but when you have buy-in, when you have alignment, when, when everyone understands why they're being applied and, and, and
[00:21:27] Ben Lloyd Pearson: feels empowered. To be a part of using
[00:21:30] Ben Lloyd Pearson: them to improve, I think they can be a
[00:21:32] Ben Lloyd Pearson: a really powerful tool and, you know, and AI shouldn't have to disrupt that effort, you know, so it's really great to hear success stories too, because I've, I've heard so many stories that go the opposite direction as well.
[00:21:44] Matt Culver: Yeah, I mean I, I like, again, I think, I think. What you have to do is look at this problem of saying like, where's the, where is my constraint? Uh, understanding where your constraint is and your end-to-end process, not just writing code, like everything that is involved. [00:22:00] Focus on that constraint and then ask, what's AI doing here?
[00:22:02] Matt Culver: Uh, and that's like you, you've immediately got your priority order of like, What's my constraint? Is there an adjacent
[00:22:08] Matt Culver: possible applied solution that somebody's already got out there that I can, that I can experiment with myself that's AI powered. And then do that. Uh, and you'll find that it's probably not code generation.
[00:22:19] Matt Culver: You'll find out that probably your constraint today is like, it's probably your decision making process for what you're supposed to do. It's probably product ideation. It's probably product planning.
[00:22:29] Ben Lloyd Pearson: I recently heard a vision generation, like coming up with a vision for the next thing. You know
[00:22:35] Matt Culver: Yeah. Yeah, yeah. Like imagine if we all lived in a beautiful world where it had big full funnels of like data vetted ideas.
[00:22:42] Matt Culver: Um, because like something we are looking at is like, Hey, can we just, can we have AI write
[00:22:47] Matt Culver: Feature changes
[00:22:49] Matt Culver: as smoke tests where we don't really care if it's fundamentally broken, but it gets a signal
[00:22:55] Matt Culver: on, like
[00:22:55] Matt Culver: what that feature might be if we fully implemented it. That's pretty low stakes and the [00:23:00] complexity of what has to happen in the background for that kind of a change is, is pretty low, and if that feels really adjacent possible.
[00:23:06] Matt Culver: Um, same thing for like market research. We start writing PRDs. I have
[00:23:10] Matt Culver: a
[00:23:10] Matt Culver: bot that just like scans the PRD. I, I'm gonna, I'm gonna turn this on in the next few weeks actually, but we've had fairly good success with like having it just crawl the documents and do everything from, like, look for
[00:23:23] Matt Culver: its standards compliance for, uh, the copy that we're going to use to look at
[00:23:28] Matt Culver: all of our different competitors.
[00:23:31] Matt Culver: See if they have related features and then compare them and then add notes about it in the document that I produced. And I don't need to have a person go out and read that anymore. And that seems like a great use of this powerful tool that's really has a great aptitude for that. Where again, it's like. Please, please, please fix this complex architectural multi-service thing. Is, is like, it, it's just, it's gonna give you garbage. Um, and, and we know that like, like anything that isn't [00:24:00] greenfield or on the like testing side, things haven't gotten really a lot better in the last two years.
[00:24:07] Ben Lloyd Pearson: Well, Matt, thank you for joining us today.
[00:24:10] Ben Lloyd Pearson: Um, if our.
[00:24:10] Ben Lloyd Pearson: listeners out there wanna follow you after, uh, tuning out from the show, where's the best place to follow you?
[00:24:15] Matt Culver: Yeah, I mean, um, shoot me a line on LinkedIn. Um, I'm doing a better job of, of like sharing some of the research that we've been doing inside the company lately, so I'm sure you'll see me on some more
[00:24:25] Matt Culver: podcasts. but,
[00:24:26] Matt Culver: um, yeah, just drop me a line on LinkedIn. You shoot me an email, matt@super.com. I'm real easy to get ahold of.
[00:24:32] Matt Culver: I'm, I'm kind of notorious for like, every time I find an interesting person and an interesting, uh, leader or somebody doing some innovation somewhere else, I'll just add them on Slack and we just. Like, I think that's kind of, it's a, it's a lovely world that like the tech community becomes very small when you're like, yeah, I'll just, I'll message Brian on Slack and
[00:24:49] Ben Lloyd Pearson: Yeah. Yeah. Yeah. Awesome. Awesome. Really cool. Well, thanks again for joining us. It's, uh, really great to get your insight today.
[00:24:54] Ben Lloyd Pearson: And to our listeners, thank you for tuning in this week. If you're not subscribed to our substack, make sure you head over to Dev [00:25:00] Interrupted dot substack.com. Give us a review over wherever you listen to your podcast, and we'll see you next week.
