Iron Sharpens Iron: Co-Founder Leadership Lessons with Brendan Cody-Kenny, Co-Founder and Chief Scientist at Sema

(Intro Narrator at 00:00:00) Today, we're talking to Brendan Cody-Kenny, co-founder at Sema, about leadership lessons he's learned as a co-founder, the future of Gen AI code, and more. You're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:20) I'm sure you heard the episode that we did with Matt, right?

(Brendan Cody-Kenny at 00:00:24) Yep.

(Joel Beasley at 00:00:25) Nice. So you're in Dublin today?

(Brendan Cody-Kenny at 00:00:27) Dublin, Ireland. Yep.

(Joel Beasley at 00:00:29) How'd you meet Matt? If you're in Dublin, how'd you meet Matt?

(Brendan Cody-Kenny at 00:00:32) Yeah. We connected online. There was a website, Angel.co, and he was looking for computer people, and I was looking for commercialization people—you know, basically business folk. And that's how we got started chatting. And we were both really interested in software and how to improve the way people build software and change software. So we went from there and built a team, built a company.

(Joel Beasley at 00:01:00) I don't know about that. Is that a Match.com for business?

(Brendan Cody-Kenny at 00:01:04) It was something like that. It was kind of like LinkedIn but for startups. I think they pivoted a couple of times. They were trying to get investors to invest in startups at one point, as far as I remember. But at the time, yeah, it kind of felt like a LinkedIn for startups.

(Joel Beasley at 00:01:21) So did Matt post and he said, "Hey, this is the problem I'm solving. I'm looking for people with expertise in this area," or how did that work?

(Brendan Cody-Kenny at 00:01:27) Pretty much exactly that. Yep. Matt had posted looking for nerds. Yeah. And I was looking for business folk who were interested in the same problem, and we got chatting. So that was great. Yeah. Because previously I'd been doing—you know, I was coming out of academia, looking for my next move, something I'd done a lot of study on. Automated software automation. I was trying to automatically improve the performance of software. And, you know, I'd done a bunch of research at that stage. And I kind of got really interested in commercialization. So I was looking around for people who knew about that sort of thing and who could, you know, do things like build teams and look for investment and all that kind of business side of things that I wasn't—I was curious about for sure and interested, but maybe I had more technical skills at the time.

(Joel Beasley at 00:02:23) Yeah. I've had a number of conversations where the companies that do really well are the companies that have multiple co-founders that each have strengths in specific areas. So I think that's a good sign. Tell me about this automatically improving software. Is it just a GPT instance that you instruct? You're like, "You are the best software engineer on Earth. Go make this code base better."

(Brendan Cody-Kenny at 00:02:49) That sounds good. No. No. At the time, it was—I mean, it was the same goal, right? We wanted the big idea was that you could automatically improve software. And the way we were trying to do it at the time is you would generate variants of the software, kind of mutations, mutants of the software. And provided you could measure which ones were better than others, you could iteratively improve the software. And so that was the idea. And so it was very costly to run. You know, I would run this thing for a week, run it over the weekend, that sort of thing. And I had, like, at one point, I had like 40 computers working together to try and do this, generating all these mutants. And the idea is that the software would improve eventually. And it kind of relies on you being able to test the software. So you need to have a good test suite. You need to make sure that the code doesn't break. If you want a faster program using this method, you'll basically get back an empty program that does nothing. It's very fast, of course. But so you have to make sure it still works, it still does what you want it to do. So that was kind of the big idea. And what's interesting is, you know, at the time, we really had no idea that all this Gen AI, GPT stuff would come along. But the way we were making the mutants of the programs back then was really primitive. Like, we would make a change and the code wouldn't even compile. It would just break it, and you would just throw it out. And it was a lot of trial and error. So these GPT systems are really powerful at changing code. You can probably—you know, what would have taken a weekend back then, you could probably ask one of these AI systems, like, "Here's what we're starting with. Give us a variant of this that you think is more performant." And in one or two steps, or maybe 10 steps, not thousands anyway, you probably would have something that was an improvement. So it'd be very interesting to revisit that research and, you know, I can only assume it's having a huge impact on that research field today.

(Joel Beasley at 00:05:04) I want to talk a little bit about the partnership with you and Matt. I'm definitely going to tell people there's a Match.com for entrepreneurs. What's one piece of advice or the most important piece of advice when picking a partner and starting a new business together?

(Brendan Cody-Kenny at 00:05:22) Yeah. I mean, it should be fun, right? And you should be able to work together. You know, if you've got that weird feeling about someone where you don't trust each other or something like that, I think you know, this is a multi-year journey you're on, and so you kind of need to be able to work together and trust each other. So I think that really comes first.

(Joel Beasley at 00:05:45) What was the dating process like?

(Brendan Cody-Kenny at 00:05:47) Yeah. I mean, well, we were emailing back and forth and having calls, introducing some of our friends to each other and people that we were already working with. There was a lot of that. Matt came over to Ireland to visit, and we went and spoke to people here. And, you know, I went over to the States with Matt and then spoke to a bunch of people there when we were forming the initial team and trying to get people together. So that's pretty much—yeah. I mean, it all started with just a message on a chat app, ironically enough. But, yeah, it's worked. I mean, there's certainly been times where you're like, "This is wild. Will this ever work?" And then, you know, you just never know what happens next. So I think keeping an open mind and willing to give it a shot. A lot of people—yeah. I guess there's plenty of worry about, like, "Will this work?" And I don't know. I seem to hear from a bunch of technical friends, like, "Okay. If you build it, you know, you might get ripped off or whatever." But I think the chances of that—

(Joel Beasley at 00:06:53) That's fear.

(Brendan Cody-Kenny at 00:06:54) Yeah. Exactly.

(Joel Beasley at 00:06:56) Are you kidding me? Do you know people are having to do something to exist right now. They have something that they're doing to exist. They're going to have to stop doing that and start doing something entirely new. Sure. And it's like, I've never been too scared of people stealing things in that regard. If anything, when I see new stuff comes along that solves a huge problem, I'm like, "Great. Another problem I don't have to solve. I can just throw money at it." You know? I can just buy the solution versus trying to build my own compliance software internally and monitor the code basis. That's one of the things that looked really interesting to me about what you guys were doing.

(Brendan Cody-Kenny at 00:07:39) Sure. Yeah. Yeah. Absolutely. Having that focus as well is super important on what's important to you. So, yeah, buying in these tools to help you manage all of this change will definitely be a big part of it. I think the same way—I mean, it's kind of analogous to the way that open source is used now in private industry, right? There are plenty of tools, and we spend a lot of time—we actually have tooling that we've been working on before we started on this AI detection for understanding your open source footprint, if you like. All the dependencies all the way down the dependency tree that you use. So in a similar way, if you measure and track your open source movement, that has kind of enabled open source in private industry. I kind of think it's going to be similar. It's going to play out similar that if you're going to use open source, you have these checks and balances. And similarly, if you're going to use AI to help in your SDLC, in your software development, there'll be similar tooling to help with that. And so we're working on that.

(Joel Beasley at 00:08:46) Help me close a gap in my knowledge. Matt was the guy working on, like, CTO dashboards and things like that. Is that correct?

(Brendan Cody-Kenny at 00:08:54) Yeah. Yeah. Absolutely, Matt.

(Joel Beasley at 00:08:56) Okay. Because I had your site pulled up real quick. When I loaded the site, it's like "Manage the risk and capture benefits of GenAI in the SDLC." And I'm like, "Am I in a parallel universe split? You're the guy I think you are." Okay.

(Brendan Cody-Kenny at 00:09:12) Absolutely. Yes. Same folk. Yeah.

(Joel Beasley at 00:09:14) Well, tell me about that progression since my conversation with Matt.

(Brendan Cody-Kenny at 00:09:18) Yeah. So things move quickly at Sema, right? There's a lot of great sense of adventure. And so maybe the change you've seen is because of that, right? We—our kind of, I guess what Sema does is we help people have conversations about software, and we kind of bridge the gap between highly technical people and more business-oriented people. So a lot of what we do is analyzing the details of software and bringing insights to the surface so people can have a conversation about the software. And a lot of what we've done previously is around code quality, team skills, process analysis, all of this sort of stuff based on the code base and the history of the code base. And so what we're doing recently is we are extending our suite of analysis tools to cover AI monitoring and detection and contextualizing that for business leaders.

(Joel Beasley at 00:10:17) So if I'm like IBM or even a more medium-sized company, and I want to deploy—I could deploy this system across, hook it into my GitHub Enterprise or whatever my version control system is. I could say, "Hey. I want to be notified if people start integrating generative AI into my projects across my whole company." You guys could deal with that?

(Brendan Cody-Kenny at 00:10:36) Yeah. Absolutely. So we keep an eye on the code base as it evolves and detect when we think AI is being used to write the code. And, I mean, that's basically it. The risk is—I guess there's two risks with all this AI stuff. One risk is that you're not using it enough, the idea that your developers now have kind of access to an intern. Kind of your development team is almost doubled if people are using it a lot, where each developer has their own intern that they can talk to and discuss and work with, pair programming style. So that's one risk, is not using enough. And then if you do use it that much, you need to keep an eye on how it's being used. You want people to—you want it to be kind of symbiotic where people are working together with these tools rather than it being a crutch where you ask a question and it does your homework, this sort of thing, and you don't learn anything.

(Joel Beasley at 00:11:36) So if I'm slow to respond, it's because there's a lot happening right now with AI, right? I'm learning. So I just got involved with a decentralized AI project called Morpheus just to explore. It was at the intersection of decentralized AI and cryptocurrencies or crypto. And I said, "Okay. Well, this is an interesting project." I knew a couple people in person that were involved with the project, and I said, "I want to see—" These are two areas where I know I have knowledge gaps. So my background is building all these softwares and selling them, then starting the podcast seven years ago. And then four or five years ago, this became my full-time gig. So I haven't been hands-on the keyboard all day for almost five years now. So I kind of took a step back, and now I'm, you know, after a 17-year stretch of never missing a day of programming essentially. Wow. So there's a gap in my knowledge. Like, usually, if I was going to go sit down and build some software, I imagine it as, like, code. But what I'm seeing is a lot of these things is, like, they'll take an LLM, and then they'll tell it a bunch of stuff. Like, they'll just talk to it like it's a person, right? And I'm like, "That's the thing. How do you ensure the consistency of the output?" Yeah. And then this creates all the issues that come along with LLMs. So I was—I'm in the stage right now of learning, trying to understand how those things happen. Do you write tests? Because you mentioned test suite earlier, which is what I was interested in.

(Brendan Cody-Kenny at 00:13:10) Yeah. You absolutely have to test what comes out of this thing. And so, I mean, there's a lot of ways of generating code. There's been plenty of sort of WYSIWYG things that will help you configure code and generate code, and there's this whole no-code movement as well. And it is mostly, as far as I know, deterministic. You ask it the same thing today and tomorrow, it gives you back the exact same thing. And you're totally right. You ask the same question today and you slightly word it different—even if you don't word it differently, you'll probably get something slightly different back tomorrow. So in that world where people are just creating code and the cost of creating code is a hell of a lot cheaper in terms of time, people's time, you know, you can have two or three different implementations of the same thing. Like, it's really crucial to be able to test the difference there and find out, well, which is the one I'm going to go with, which one is more or less secure, which one's more performant, which one has bugs. You know, what's the—they can be written in different paradigms as well, different languages, right? All of that becomes, like, open to testing and, you know, there's a whole lot of barriers between these languages. You'll find that a lot of these languages and these ecosystems, people—a lot of people work in one or the other, and things are implemented a certain way in certain areas. And there is a kind of an incumbent library for—you know, there's kind of like one main way of doing it in each of these libraries. So all those barriers, you know, it becomes really easy to translate between these different languages. So in that environment where it's just so easy, instead of you coding for two hours a day on, you know, two hours on a specific problem, you can spend a good hour coming up with several different implementations. And, basically, humans only have a certain amount of bandwidth to fit all of that, you know, read all this code. So being able to have test tools to discern between these different versions of the program, I think, will really be important. So we'll swing from all the focus being on the code and within the implementation. I'm kind of wondering, will some of the focus swing back to more DevOps testing, runtime, environment, observability, all this stuff that's—you know, it's already well underway and kind of the industrialization of the SDLC. That's the way I think about it where there's a lot of—yeah, it seems like there's a lot of work being done in that area. And so I think that will become even more of a focus for engineering teams.

(Joel Beasley at 00:15:53) Are you scared of this stuff, or are you excited about it?

(Brendan Cody-Kenny at 00:15:56) I'm pretty excited. Yeah. I mean, just the sheer adventure. I mean, for everybody out there to be able to learn so much, I mean, that's just incredible, right, that you can—you can travel so far with these things. Yeah.

(Brendan Cody-Kenny at 00:16:10) No. I'm optimistic about it because, well, for now anyway, it's a really powerful tool for managing information and getting at information. So that's really incredible. I guess it'll get pretty scary when we plug it into the physical world. I think there's a long way to go to plug it into data and plug it into the context.

(Brendan Cody-Kenny at 00:16:34) So like the SDLC or, you know, your daily workflow, it's available in some places, but it's not exactly plugged in. It doesn't have access to all of our information yet.

(Joel Beasley at 00:16:48) Yeah. Well, there's plenty of companies that are building those APIs. That's what they're doing. Right? Companies are building their own language models and giving it all of its data.

(Joel Beasley at 00:16:58) They're saying here is all of our data. Here is complete access to all of our storage vaults, and then they're trying to teach it and talk to it.

(Brendan Cody-Kenny at 00:17:07) Yeah. I mean, every time I read, you know, come up with some idea about the future, I read about some company doing it the next day. So yeah, the pace of development is astounding. But yeah, I mean, I'm very interested in that because it just gives you access to so much stuff. There's just so much stuff out there to learn. I guess when it plugs into the physical world, that'll be, you know, when you're automating robots that can have kind of a physical impact on the world. That's something I know less about.

(Joel Beasley at 00:17:37) They're gonna outnumber us 10 to one pretty soon here. I wanna steer the conversation back to technology leaders. So technology leaders, people who are growing and advancing their career, they're the ones that listen to the show to learn from other tech leaders, what's coming around the corner, how to grow, how to become better and improve. This obviously is really interesting, the tools that you're making that's gonna help you monitor your code base and see when Gen AI is included. So that way, the newest employees don't just run wild with some technology in your project that you don't know how it got there.

(Joel Beasley at 00:18:12) Right? But AI and generative AI as a whole, how should these technology leaders look at it? Should they be scared of it? Should they be happy about it? What's the response?

(Brendan Cody-Kenny at 00:18:23) I definitely think it's a good thing and should definitely be happy about it. Like I mentioned earlier, you know, if you think about your team size doubling, now everybody's got their own personal tutor, their own intern to talk to. And so, you know, I mean, doing it in the right way, you know, if somebody's just gonna make the system do all their work and they're gonna copy and paste it without any oversight, that would be clearly a bad thing. You can also spend a lot of time back and forth with these systems and not actually get anywhere. Now arguably, that happens quite often anyway when you're programming.

(Brendan Cody-Kenny at 00:19:01) You know, you get blocked. But these systems, yeah, I mean, the more senior you are, you can probably move faster with these systems. But as well, for junior folk, being able to unblock people, I think, you know, you can drop an error message into one of these systems, and it'll instantly tell you a couple of things that it could be. And, you know, you could spend a couple of hours. So I think from what I've seen, it just looks like it's gonna be a really powerful addition to a software developer's toolkit.

(Joel Beasley at 00:19:30) I saw an article the other day called the death of Stack Overflow. Oh, yeah. Can you just imagine that's 99% of what people use Stack Overflow for. I have this problem. What do I do?

(Joel Beasley at 00:19:41) And then other people yell at you, and then some very few people give you helpful responses.

(Brendan Cody-Kenny at 00:19:46) Yeah.

(Joel Beasley at 00:19:46) That's gonna, that's gone now. Now you've got the AI tool. You don't need Stack Overflow.

(Brendan Cody-Kenny at 00:19:51) That does sound like a possibility. It's been a while since I've been on Stack Overflow.

(Joel Beasley at 00:19:56) Right? Usually, people are on early in their career to get help from other people who know what they're doing.

(Brendan Cody-Kenny at 00:20:02) I still, you know, even though I've been programming for a long time, I still use Stack Overflow to get, you know, there's always some command that you just forgot how it works or some syntax thing. So all of that, yeah, you're totally right, goes away, or becomes just way more accessible, just easier to work with. But I would see that as a good thing overall from an engineering leadership perspective. Right? It sort of supercharges things.

(Brendan Cody-Kenny at 00:20:29) It makes things more dynamic. Your entire team can spend more time thinking rather than working on syntax issues or kind of minuscule bugs. So, hopefully, your entire engineering department can move up one layer of abstraction. And the ability to modify code and write code and move code around, you know, it just becomes way, way faster.

(Joel Beasley at 00:20:53) I'm curious about companies. Have you seen companies using large language models privately and connecting them to their own information to work with those large language models?

(Brendan Cody-Kenny at 00:21:05) I haven't seen a lot of that. It takes a reasonable amount of work to get these. I've played around with it a bit. There was one project called private GPT, and you can run this sort of stuff. You can run it locally.

(Brendan Cody-Kenny at 00:21:18) It's incredibly slow. But you can download them. You can train these things on your own datasets. But it's a reasonable amount of work to get it plugged in. I think, you know, we're right in the middle of it.

(Brendan Cody-Kenny at 00:21:31) Like you said, there's a lot of companies giving huge amounts of data to these systems. And so, yeah, I think that's, you know, that is sort of a barrier to entry. I think there'll be a lot of startups who will be able to provide these tools without every single company out there having to kind of build their own private LLM. But if you have a very specific use case, of course, you'll be able to build your own personal one if you need it. I think that's kind of the thing.

(Brendan Cody-Kenny at 00:22:03) I mean, not many people run their own email server anymore. It's kind of easier to get it as a service.

(Joel Beasley at 00:22:10) That is true. Yeah. I was curious if, because of the way these language models work with you telling it to do things through prompting it, converting that information into vectors, and then storing it like that. If these systems can be trained, like, for example, there's a popular open one called Llama. Right?

(Brendan Cody-Kenny at 00:22:32) Right.

(Joel Beasley at 00:22:33) How do I know that that Llama instance when it was built, the model that they're gonna release to everybody hasn't been given instructions to pass anything that looks like a credit card number back to some server that they have? You know, how do we know that they haven't embedded or taught the model to do something nefarious?

(Brendan Cody-Kenny at 00:22:54) Yeah. That's a great question. That's a, yeah, pretty big concern. That same question can be applied to software right now. So you download a binary.

(Brendan Cody-Kenny at 00:23:05) There are ways of testing for, you know, basically network access. We still have control over that. So all of that kind of, all of our kind of network and system security still is highly relevant for these systems as well. And I guess, to your point, you can't really look inside these systems and read the code. They're kind of opaque boxes in a way.

(Brendan Cody-Kenny at 00:23:30) And so yeah. I mean, in that scenario, being able to test these things externally becomes even more important. And I think all the tools that handle security right now will similarly can be applied to this problem.

(Joel Beasley at 00:23:46) As a co-founder, what are you learning right now?

(Brendan Cody-Kenny at 00:23:50) Yeah. I'm learning all the time from the people I work with. That's always a new thing. I'm certainly learning more about all these GPT systems. I continue to learn about distributed systems and software in general.

(Brendan Cody-Kenny at 00:24:06) There's always something new. There's always, you learn to program and you learn patterns and how to build software. But when to use these patterns and when is the right time to use a certain architecture, that's still something you can learn from, learn about. So I'm still learning about that. I'm still learning a lot about people.

(Brendan Cody-Kenny at 00:24:32) Right? It's been great working with Matt because he's extroverted in a way. You know, he's very outgoing so that, I definitely enjoy working with him and the sheer energy he brings. And then as well, you know, working on this new AI detection tool, right, there's a lot of people applying AI to various tasks, various problems. And I don't know that there's as many people doing what we're doing in terms of, like, is your software team using AI?

(Brendan Cody-Kenny at 00:25:06) How is AI influencing the software you write? And so that's been very interesting. We've run a bunch of experiments, and we find very counterintuitive results. So, for example, today, we found that one of the GPT systems, on average, humans have better grammar and better spelling on average, but the variability is a lot higher. So, and it might be dependent on the code base that we tested as well.

(Brendan Cody-Kenny at 00:25:42) But, you know, some people have really, really strong grammar better than any of these GPT systems, and some people don't. So it might be person to person, team to team. There's a hell of a lot of variability there. And what we're seeing is that these GPT systems, at least on that measure, are highly consistent. They have a certain level of grammar and spelling, and that doesn't seem to vary wildly in comparison to the human side of things.

(Joel Beasley at 00:26:11) It's interesting because there's an interesting thought about effort and energy. You can communicate at different levels to different people, and how you communicate, I believe, directly corresponds to how much energy it takes to communicate. Right? So you can have people that are, let's say, level eight communicators that when they're doing a git commit or whatnot, they might just be level three because they're exhausted at the end of the day. But the GPT systems are more likely to be stable.

(Joel Beasley at 00:26:41) Right?

(Brendan Cody-Kenny at 00:26:42) Yeah. Yeah. That's true. Yeah. They don't need sleep.

(Brendan Cody-Kenny at 00:26:46) That's a significant advantage.

(Joel Beasley at 00:26:48) And they can work together. So

(Brendan Cody-Kenny at 00:26:50) Yeah. When these things get kind of networked together, yeah, make for very interesting meetings. Right? You've got double the number of voices in the room.

(Joel Beasley at 00:26:59) Yeah. What's the call to action? Do people go and they sign up and they take a look at it? How do you get people started?

(Brendan Cody-Kenny at 00:27:09) Yeah. Absolutely. Happy to chat with anyone interested. We can organize pilots and demos. Absolutely.

(Brendan Cody-Kenny at 00:27:16) Myself and Matt, more than happy to talk with anybody. Shoot us a message. Find us on LinkedIn. Go check out our website, semasoftware.com. All of that.

(Joel Beasley at 00:27:25) I'm on the website right now. If I do the comprehensive code base scan, what is that gonna tell me?

(Brendan Cody-Kenny at 00:27:31) That is gonna tell you about your code quality. It's gonna tell you about your team. It's gonna tell you about process. It'll tell you about security. It'll tell you about all your open source dependencies.

(Brendan Cody-Kenny at 00:27:46) And so that's it in a nutshell in terms of what that, and so our AI detection is also being added as another module to that. So if you sign up for a code base scan, you know, a lot of investors when they're investing in a company or buying a company rely heavily on our code base scans to get really up to speed. You know, you wanna know what you're buying. You wanna know what you're investing in. And we also have CTOs who use those tools, you know, to get through their first ninety days and get through it in thirty days, this sort of thing.

(Brendan Cody-Kenny at 00:28:16) So you can move very quickly, and I think there's a hell of a lot, you know, the value sort of grows exponentially with this system. With the more code, and the larger the code base, the larger the team, the bigger the organization, the more valuable these tools are, because it's so hard to get a handle on who knows what, who are the experts in what parts of the code base, you know, where's the risk in there, what really needs attention, what parts of it are good as well. You know? That's really important. So we've, you know, that's our code base scanner, surfacing all that information about the engineering function in an organization. So it's kind of analogous to, you know, a CFO coming in and reviewing the accounts, looking at the books. Similarly, a CTO or technical leader, or an investor in a company wants to know what they're getting. They wanna see and have a summary of what's actually in the software.

(Joel Beasley at 00:29:20) Well, it's definitely interesting to watch your company and you and Matt's journey mature because I like where it's headed. I think the idea of an AI code monitor product, the Gen AI compliance tracker, it's gonna be real big as these regulations, whether or not the regulations make sense or not, is irrelevant. They still are government regulations nonetheless. So being able to be aware of what they are and your level of compliance is important. And then I also think I like the CTO aspect.

(Joel Beasley at 00:29:51) CTO getting up to speed in thirty days instead of ninety days by understanding what's going on in the code base scans and all of it. So you guys have, like, three or four things going on right now that all seem to have a very similar set of, like, core tools. So I'm curious to see, like, which one's gonna take off the most for you guys.

(Brendan Cody-Kenny at 00:30:13) Yeah. Yeah. And, I mean, what we're seeing is that all of these tools a lot of the time, we run all of them as well. Like, a code base scan will encompass all of these different modules and all these different aspects, all these different ways of thinking about your software, your software function, basically. So it's kind of, we try to be very broad.

(Joel Beasley at 00:30:36) Yeah. But this has been great meeting you, Brendan. We made a podcast. How do you feel?

(Brendan Cody-Kenny at 00:30:40) Great. Yeah. Thanks for that, Joel. That was fun. That was exciting.

(Joel Beasley at 00:30:44) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.