#422 Managing Developer Teams in the AI Era: A Conversation with Will Blackburn

[00:00:00]

Mehmet: Hello and welcome back to a new episode of the CTO Show with Mehmet. Today I'm very pleased joining me from the US, Will Blackburn. Will, thank you very much for being with me on the show today. The way I love to do it is I give it to my guests to tell us a little bit more about, you know, [00:01:00] yourself, your journey and what you're currently up to.

Mehmet: So the floor is yours.

Will: Awesome. Well, thanks so much for having me. I'm excited, excited to be here. Uh, so yeah, my name is, is Will Blackburn and I am currently a co founder of a startup that's located here in the Southeast United States, uh, called DevClarity, and we are creating a platform for software development leaders to help them better.

Will: Manage their team. And so, uh, to give you a little insight into why, uh, we're tackling that problem, uh, can go back a few more years. About eight years ago, I was a founding employee of another software company. Uh, day one employee, and I was the sole software developer. And then over the course of about seven years, I grew into the CTO role and went from the only software developer in the company to eventually managing, uh, a software and product team of about 35, um, that was across the United States.

Will: And, [00:02:00] and so learned a lot through that journey, uh, made a lot of mistakes. Um, and really came to the conclusion that while there's a lot of tools out there for software developers, and you always have developers building tools for developers, there's not a ton of tools out there for. Dev leaders. And, um, and so a lot of that journey I went on was a trial by fire, jumping into the deep end and learning how to swim.

Will: And so now dedicating my time, um, building a platform to help. Cool.

Mehmet: Uh, and, you know, thank you again, Will, for being here with me today. So, you know, it seems like your journey was, you know, from the beginning in, in software development. Just out of curiosity, I ask this sometimes to my guests. Um, you know, why did you choose this, uh, this path, I would say?

Mehmet: And what was the, you know, I would say, the reason? The thing that attracted you the [00:03:00] most for software development?

Will: Yeah, uh, that's a great question. Um, I, uh, really fell in love with software when I was at university, uh, college, as we call it. Uh, and Through that, through that journey, um, just the, the, the first thing that I learned was that, um, with software, I could just build something from scratch.

Will: And, um, it was this amazing feeling to be able to sit down at my computer and then a few hours later, a few days later, have an application that I've built that I can put out in the world and that other people can use and it can solve their problems. And so, um, Really, the first time I did that I went to a library.

Will: I wasn't sure where my career was going to go, but I'd always loved kind of tinkering with technology, um, and sat down with [00:04:00] actually a PHP and my sequel for dummies book and, uh, sat there and went through the tutorial and I kind of modified. The tutorial for something that I cared about in a problem that I knew that people had.

Will: And, um, from then on, I've been, um, really kind of addicted to technology and finding ways to solve other people's problems, problems with it. Um, and, and so as I've seen technology advance, um, over the past 10, 15 years, It's only made me more excited about what we can do and the problems we can solve and how we can push the world forward with the technology.

Mehmet: Amazing. Um, I want to go back, you know, you just explained about what you do at DevClarity, but I want to also figure out from you, Will, you know, of course, being a developer yourself, maybe you've seen some of the, and you just explained that I know, but when was the moment when you said, [00:05:00] okay, this is something that is big enough.

Mehmet: A lot of people, probably are facing, you know, these issues and we need to sit down and build something for that. So what, I would say, what was the trigger for you to start, uh, you know, building DevClarity?

Will: Yeah. Um, really, I felt a lot of that when we were building our last company. Our last company was a FinTech company, financial technology company.

Will: And, um, as Really? So, so I was a first time development leader and first time CTO, and then we went through COVID and the market went crazy. Developers were getting sniped left and right for two times as much as they were making prior. Um, and, uh, it was, it was a crazy time to be a dev leader and to really learn about software development.

Will: And, um, I was just [00:06:00] shocked at the lack of help and tools that I had, um, in that process, I was doing things in Excel and, you know, OneNote and, uh, it was relying on kind of generalist HR tools that just didn't work. Didn't understand software development and how software developers are different and the technical nuance that comes with it.

Will: And so even then I was, then I was frustrated and that's always good motivation for a software developer to, to start to tinker and build with something. And I didn't have time then, but, um, our company sold in, uh, December of 2022. And so. When the company sold, that's when I could really like turn back to that problem and, um, start talking with other people.

Will: And as we went out and talked with other development leaders, we heard the same stories that it's, uh, difficult to manage already. It's probably even more difficult to manage software developers because of the technical nuance [00:07:00] and how skills and system expertise kind of hides beneath this fog. That is a technology.

Will: And so that's when my co founder and I really, really knew that. Um, this is an area where we wanted to spend our time.

Mehmet: So if I want to come before I shift, you know, to, to some other topics as, as a, as a leader, as an, you know, maybe I would believe, you know, you would ultimate customer or ultimate persona would be engineering managers, probably, you know, um, You correct me if I'm wrong, like probably maybe even VP of engineering.

Mehmet: I'm not sure if you know, they would be interested. So when I would use your tool will or your solution. So the direct benefit that I would get if you can explain that to us a little bit more.

Will: Yeah, absolutely. And you're right. It's essentially engineering managers, but then also [00:08:00] VPs, directors, and, uh, we eventually want to serve, you know, um, true CTOs of larger organizations.

Will: Um, yeah, the, the, the main benefit that they're getting is, um, some structure and help around. What are kind of the core management functions of a software development leader. And so, uh, at kind of the lowest level for engineering managers, that starts with kind of developer specific one on ones. And so you can take developer specific one on ones, you can track action items that developers are working towards.

Will: Um, but then as you look at what a VP and a CTO cares, cares about, they care a little bit more about creating the structure for their, for their Managers to be successful in helping them. And so that's where we come in with helping them, uh, understand, uh, what skills breakdown looks like across the [00:09:00] organization or, uh, who has the system's expertise, how do those play into the conversations that the engineering managers are having with their software developers?

Will: And, um, kind of always goes back to, uh, you know, are your developers happy, uh, Unblocked and productive. And so if you can make sure your developers are that, then, uh, it's going to be a win win across the board.

Mehmet: Cool. Now, you mentioned something about how, of course, and all of us, we're witnessing and personally, I witnessed that fast change in technology, but when it comes to software development, uh, we're like, what are like the biggest one that, that you are, you know, really experienced?

Mehmet: And now we are experiencing, of course, I know like AI would be the one, but how AI is shifting, you know, the way software development is done today? Yeah.

Will: Yeah. Um, yeah, it's, it's [00:10:00] changing so much. It's, it's, uh, kind of hard to understate how much it's changing, um, software development. And so, you know, I think the first place people go is that it can increase productivity and, um, I think that that's the right place to start.

Will: It's really a leverage multiplier for software development organizations. And, um, so what we've seen in, uh, both our own experience, but also being able to talk, and since we serve software development leaders, we get to hear their stories and we get to hear what they're doing and what's working and what's not working, um, uh, almost every company we talked to today has rolled out some sort of.

Will: Um, AI coding assistant, uh, most frequently it's something like copilot inside their editor. That's that's helping their software development team. Um, just write code and take away some of the. Kind of more boring [00:11:00] monotonous work that developers have to do. And, um, that's where it's starting, but then we're also starting to see it creep into other parts of the software development life cycle.

Will: So whether that is, um, you know, more on writing requirements and getting help with that and making sure those are well defined before it gets truly assigned to a developer or the developers involved in that. process, or whether it's on the other side through writing tests, um, or performing tests. Um, it's, it's really starting to change so much and it's moving at a rapid pace.

Mehmet: So can we say it's, it's working on, on the both sides, um, of the productivity and the creativity? Can we, can we say this?

Will: Yeah, I, yeah, absolutely. And that's how I personally experienced it, um, as well. And so I think, yeah, a lot of people like to kind of put it in the box of just, okay, this is [00:12:00] co pilot now it's inside VS code.

Will: It's inside my editor and now we can, you know, write some lines of code for me that, uh, are, uh, Highly predictive, but it really can do much more than that. And that's where I think that if you look at companies today, most of them have checked that first box of, okay, we figured out security and compliance enough, uh, or we're happy with what these companies are saying, what Microsoft's saying to where now I've got co pilot, um, helping out my developers and, um, what those companies are starting to look at is how can we help out in the other areas.

Will: And one of those is in, um, kind of the creative problem solving. I think that's where it can, uh, continue to help developers move a little faster and create higher quality, um, products by not having it just do that all on its own, but sitting there alongside a developer or someone who's responsible for writing those requirements [00:13:00] and helping them out.

Mehmet: Right. If I want to ask here about, you know, from writing code is okay. How, I'm not the expert here, so I'm asking you, Will, how good AI is today in putting, let's say, requirements for a project? So, I mean, in another sense, you know, like, like, You need to write the code and you might ask the AI, chat, GPT, co pilot, whatever.

Mehmet: Okay, write me, for example, a code that does blah, blah, blah, blah, blah. But is it good now enough? So I give a requirement and then it decides like, okay, what exactly need to be built? So it start to structure, you know, the things in a proper manner. Is it? Is it reaching this phase or not yet?

Will: Yeah, my answer is it can be.

Will: And this is where, like, I would, uh, turn to the development leaders and say, and that's your responsibility to get it there. And so what I mean by that? Um, I'm not suggesting that you should be training it. Your own [00:14:00] large language models. Actually, the models we have today are great. And so what I'm really talking about is the other parts of the picture.

Will: So when you think of like what AI really is and how we experience it, you've got a model that you're talking to. So again, like the models are good enough now to be able to do what we need them to do. And then you've got your prompt. And so that's the next area we go to, which is. Um, are you writing good prompts?

Will: And then a piece of that, but I like to separate it out is the context you provide, um, in, in that prompt. And so, um, when, when I see people complain about the output of AI today, as it's related to either requirement writing or even code writing, most often they're getting the prompt in the context piece wrong.

Will: And so, um, when, when you think of. An experience inside of an [00:15:00] editor, and maybe you've like written a comment inside your code, asking it to do something for you. Um, that's, that's like level one of, of doing this well. And so you're not going to get the best output just as if you open up a chat, you BT new.

Will: Conversation. You say, hey, I want to add this functionality to my code. Um, we've got this request coming in. Can can you write the requirements for it? Uh, it's probably going to get you kind of like B level B minus level requirements back. But if you've set up a workflow, um, where. The A. I. Has the ability to read your code base.

Will: Maybe it can use retrieval, augmented generation rag and find the correct files it needs for context, or you can just give it to it directly. And then, if on top of that, not only can I get the right code base, but it can understand the principles that you care about and the patterns that you've [00:16:00] established in your code base.

Will: Um, then that's where I have been blown away. At the high level of quality that the models can, can output. And so, um, like how can you do that today? Uh, a tool I love is called cursor. It's an another editor that many of you probably have heard of at this point. It's a fork from VS code. So, uh, very similar.

Will: But they've gone a step further and, um, there's a few other tools out there like this. Another one's called windsurf and, uh, they make it really, really easy to give it the context that you need to be able to tag the files that you care about. And then in cursor, for example, you can create what's called a cursor rules file.

Will: It's a dot file. So, um, It is essentially becomes a system prompt that's included in every single prompt that you write to AI. So every time you open up a chat window or you do an inline prompt request, [00:17:00] it has this system prompt. And that's where you can define your language, the stack that you use, the version numbers that you're on, uh, the patterns and principles that you care about.

Will: And when all of that setup is done correctly. You get really good output,

Mehmet: right? You mentioned copilot, you mentioned cursor, you mentioned, you know, of course, there's plenty of other tools that are emerging every day, but of course, like copilot and cursor are the most ones that, and I think cloud, it has some good capabilities in generating quotes also as well.

Mehmet: I've tried like something and I was surprised, honestly, although like I'm not a developer, but at least, you know, I can understand, you know, the deepness and I was surprised by the way. Actually, it writes the commands for you, so kind of a documentation for the code, which was, you know, also good. What, like, some of the challenges you've faced?

Mehmet: You've seen, or maybe you have heard from, from other people about these tools.

Will: Yeah. [00:18:00] Well, it's been fun. So our startup is about a year old now. And, um, uh, we actually sold my last company right as, you know, chat, GPT was coming out and co pilot, uh, beta was first available. And, um, and so I didn't get to make the decisions for that engineering work.

Will: Um, but I get to make the decisions for our smaller startup now, and we've leaned really heavy into AI development. And our thought is, um, you know, what we've done is, is we use all of the developers use cursor, we've got Slack channels dedicated to talking about the wins and the challenges with AI. So we can just continue to try to learn and do things like update our cursor rules file to, to make it better.

Will: I think like. The, um, the, so the biggest issue I see out there today is, um, a lack of good context. And so we kind of just went over that and talked about the solutions there. But I [00:19:00] think that's where everyone has to start, which is, have I given it the right context in terms of code base and then instructions through like a cursor rules file.

Will: Beyond that, there are other challenges still. And the biggest one I see, um, and I don't know if there's a term for it, But it's essentially when the AI starts to go down this rabbit hole. And, um, that is where it's really important. Uh, to have someone who understands the language, understands the framework, understands the application to catch AI quick because it can start to waste hours and hours of your time when it goes down the wrong rabbit hole towards a solution.

Will: Um, that, you know, either there's a much simpler solution. And it started going down this trail and because the way these generative models work, once those tokens exist in the history, then it's going to continue to, to, to go further. That's the biggest, I've, the [00:20:00] biggest issue I've seen with it. Um, and so I'm a big fan of like prototyping out changes quickly and then stopping and then, uh, you know, as a human and, Hopefully, as an expert or someone who's knowledgeable about the framework in the stack, you can figure out if it's going down the right path or if you need to kind of reset at this, um, this early level and change, change it up.

Will: So, uh, Andy Grove. Uh, famous Intel CEO, uh, had, had a famous principle about catching things at its lowest value stages is what he called it. And so he used the analogy of if you're making an omelet and serving it to a customer and you've got a bad, bad egg, you want to figure out that the egg is bad before you even crack it.

Will: Um, but hopefully it's before you've cracked it and you've combined it with other eggs and you put more ingredients in it and then you've served it to a customer who gets angry at you. [00:21:00] You want to figure out if it's bad much sooner than that. And so you can apply that principle here with AI and kind of check in early with it to make sure that it's going down the right path.

Mehmet: Right. Um, one thing in I've seen in general with the large language models, not necessarily again, I don't talk from an expert perspective here when it comes to generating code, but I tried it with code also as well. Um, and this is exactly to the point that you just mentioned. So if you keep, you know, the conversation, of course, using the prompts after a while, especially recently, I don't know why all of the models, at least the one that I used, Like Gemini, OpenAI, ChatGPT, and Cloud.

Mehmet: So all of them, what I feel is like after, I would say maybe seven, eight iterations,

Will: so

Mehmet: they become lazy and they start to act weird, which is okay. Not weird in a [00:22:00] sense, but you understand my point. So they don't deliver. And then I think the same thing from the last time it gave me a kind of a good answer.

Mehmet: I put it in a new chat. You know, I started a new prompt and then I start to get like better answers over.

Will: Right.

Mehmet: Yeah. So, so it's like, I mean, I have to give it a pause or a break and then, or I have to bring some another expert look into it. So I'm not sure if this is valid.

Will: Yeah, no, I've seen exactly the same thing.

Will: Um, and, and so, uh, yeah, like, and, and I think that's, again, you know, as it has these new tokens and like, it, it starts to not forget, but you know, we're going to use human words here. to kind of explain these things. Um, so yeah, it's, it starts to forget what you care about or what the original goal was, and then the performance goes down and that's, that's again, where if you've got a good system set up that allows you to hop into a new conversation and have the right context again, quickly, um, [00:23:00] you don't feel like, oh, I've got to stay in this conversation because it has all the context it needs.

Will: And if I start a new conversation, it's going to lose that you're able to kind of quickly hop into, um, Uh, a new conversation and continue from there and then,

Mehmet: yeah,

Will: go ahead. Sorry.

Mehmet: No, I just wanted to say like, this is also, if I want to do like the comparison to your point, like using human words, so like, I remember if I used to have back in the days, a very complex project and you know, all the team used to be lost.

Mehmet: We say, okay, you know what, let's erase the board and let's start from scratch. So it's like something very similar. Right.

Will: Or like go for a quick walk, like go refill your cup of coffee and, uh, see, see what happens when, when you come back. So, um, yeah, yeah, absolutely.

Mehmet: Um, one thing, which, you know, maybe I gonna take you back to the.

Mehmet: management, uh, position here. [00:24:00] So how is AI, you know, shaping the role of, of leaders, you know, and engineering managers? Yeah. Are there like some new skills that they need to be acquired by them? Is there anything that they need to do it different than they used to do before? Like what's the shift for them here?

Will: Yeah. Yeah. We think about that in, in a, in a couple of ways. Um, one just with AI in general, I think. Every leader needs to understand that AI is going to lead to increased productivity, you know, hype aside, whatever 10 X or two X, whatever it is, it's going to be, there is going to be increased productivity.

Will: And so that means like, individual developer impact is going up. So developers are already the most highly leveraged employees at organization. A single developer can impact your company in ways that a single sales rep or a single. HR [00:25:00] employee or a single operations employee couldn't dream up because of the way they can write this permissionless code that then runs and gets executed a million times over.

Will: So now with AI, if that productivity is increasing, developer impact is increasing, and that means that managers and development leaders are in a position to have the greatest impact that they've ever had. Their leverage as managers is increasing and so, uh, they, they need to be thinking about this and paying attention, understanding and enabling their developers to use these tools and to really lean into it.

Will: So that's the 1st piece that the other way we think about this. It's also creating a new opportunity for leaders to, um, have more tools and have more help. And that's what we're leaning into with Deaf Clarity because so much of people management is more on the qualitative [00:26:00] side of things as opposed to the quantitative side of things.

Will: You know, developers spend 10, 000 hours talking to computers and giving computers instructions, and then they wake up one day and they've been promoted to a manager and now they have to give people instructions and humans have like hidden compilers. So you can't be sure that what you've said to them.

Will: Set to a human gets interpreted the same way every single time. So it's so different than, than, uh, talking to computers. And so now with generative AI, we're able to create tools to help leaders make sense of and understand some of the more qualitative things like. Taking one on one notes and then having that as, uh, an insight that you can marry with quantitative data.

Will: So maybe someone's commits have dropped over the past two weeks, uh, but maybe now actually the data exists to understand [00:27:00] why that happened and we can just automatically connect those two things behind the scenes. And, um, so you can get a better sense of what's going on at your organization, where you need to help out and where you need to, you know, Um, you know, leave your developers alone.

Mehmet: Yeah, absolutely. Now, doing the part which is a little bit fun, and I know, like, the question, it's hard to ask, and no one has a crystal ball, right? So, but, from all what you have seen, Will, and because, you know, you, you, You are in a, you know, I would say, uh, position where you, because of this, of, of dev clarity.

Mehmet: So you are interacting with a lot of, you know, managers and of course, developers and so on. So what is the future like? I mean, when I say the future, you know, before we used to ask after 10 years, five years, I know we cannot predict anymore like that long, but how [00:28:00] are you, how are you seeing things moving with the AI wave?

Mehmet: Taking us places now.

Will: Yeah, yeah, um, I think, uh, while a lot of the future is unclear, I think what is clear to me is, um, whether we want it to happen or not, more development is going to happen through prompting. And that's just, it's just clear to me that there's advantages to that. And so, uh, I think when you look 12 months out, you know, we will be well on our way, uh, towards moving most programming, uh, to happen through prompting and, um, And, and, and so that's going to have a significant impact, uh, on the development world.

Will: I, um, I think developers are here to stay. And I think that, uh, you'll need developers to make [00:29:00] the best use of these new technologies and these new tools that make programming easier than ever, because you're able to use human language now, kind of one step further. And I think I would like encourage people to think about the history of it.

Will: Talking to computers and software development and think about like the spectrum that we've been on where in the early days we wrote machine language code that really told the computer exactly what to do at the machine level and since that point, we've been slowly abstracting towards human language and, um, you know, we got to the point with these high level programming languages.

Will: Now we're so much of the, uh, Of what's underneath, we don't have to deal with anymore. And we can write these classes and these objects, uh, that encapsulate what we actually care about and abstract up. And so, in many [00:30:00] ways, um, you know, using human language to create what eventually commands the computer to do something is a natural step.

Will: It's, uh, it's, uh, Different because you can use AI to write a lot of code now in a way that has never been done before, but it's also a natural step up. And so my prediction is that more and more code will be writing with prompting. The good news is that you'll have more and more control over that through giving good context and using things like cursor rules to tell the AI exactly what patterns you want written, to establish the templates that you care about, to have the test coverage that, that, that you want.

Will: And, um, so developers will become more focused on problem solving. And be able to move quicker on, uh, the computer code.

Mehmet: Absolutely. I agree with you. Will, [00:31:00] um, as we are coming, you know, to an end and we're almost closing, so, uh, what this is, I ask all the time, so what piece of advice you would give to other, to be founders or like new founders or even CTOs, you know, who are in, in now today, maybe listening or watching us, uh, especially when it comes to ai.

Mehmet: Final piece of advice.

Will: Yeah, I think my final piece of advice on top of everything I've already said is, um, use the tech, just like lean into the tech and use it and then go one step further and form opinions. I think something that took me too long to learn in my first journey as CTO. Is that, uh, like I wanted to, you know, get everybody's thoughts on things and come to a consensus about things and, uh, try a lot of different things.

Will: [00:32:00] And you've, you've got to do all of those things, but you also have to find a way to form opinions as a team and then align people towards. Certain efforts. And so, um, how do those come together? One, use the technology and then that will give you the information and the knowledge you need to form opinions on things.

Will: And then I think if you do this right, you're going to form some opinions that using AI is the right thing to do, and you're going to lean into ways. Um, that, that you can do it and you're going to end up with an organization and a team that can move faster while still writing really high quality code.

Mehmet: Right. Absolutely. Well, finally, where people can get in touch with you and find more about DevClarity.

Will: Yeah. Uh, so you can go to dev clarity.ai and hear, uh, all about how we help [00:33:00] development leaders and VPs, uh, engineering managers create structure that helps their developer teams be happy, unblocked, and productive.

Will: And then you can find me on LinkedIn. That's where I'm most active. Um, and, uh, I'm sure we can post a link, but it's just, uh, will h Blackburn

Mehmet: Sure.

Will: Find me.

Mehmet: Yeah, I will I will take care of that. Well, don't worry So this is for the audience all the links that will just mention they will be in the show notes if you are listening On your favorite podcasting app or if you're watching this on youtube, they are in the description Will thank you very much for you know, being with me here today a lot of insights great, you know, also advice to to fellow CTOs and founders about leveraging the technology hundred percent on that and also like what you're currently doing is I think something that every, uh, you know software development manager should be looking at because You know time is money and you you help them in saving [00:34:00] time and you know increase the productivity.

Mehmet: So thank you for sharing You know your story and of course, you know more about dev clarity with us today Um, and this is for the audience This is usually how I end my episodes. So if you just discovered our podcast, by luck, thank you for passing by. If you like it, please give us thumbs up, share it with your friends and colleagues.

Mehmet: And of course, come back to listen to a new episode. And if you are one of the people who keeps coming back, thank you for sending me your comments, your suggestions, your questions, and keep them coming. Read all of them and I try as much as I can to reply as fast as possible Thank you very much for tuning in and we'll meet again very soon.

Mehmet: Thank you. Bye. Bye. Thank you