#567 Engineering Creativity: Peadar Coyle on Scaling AI Audio Infrastructure

Mehmet: [00:00:00] Hello and welcome back to a new episode of the CTO Show with Mehmet today. I'm very pleased joining me, Peadar Coyle, He's founder of AudioStack. As you know by now, I don't like to steal much of my guest time. I prefer to give it to them. So Peadar, like thank you first to be with me here today. Tell us a little more about you, your background, your journey, and tell me a little bit more about, you know, why you started AudioStack.

So the floor is yours. 

Peadar: Cool. Thank you for having me. Um, AudioStack is an AI native solution for audio production, so that's quite broad. But our customer base is largely media and entertainment companies and brands throughout the world. I have two co-founders, Bjorn and Timo, and uh, three of us got together about nearly seven years ago now to explore the impact.

Of, um, you know, you know, synthetic voice and synthetic speech and synthetic, uh, music generation and, and various adjacent technologies and what that would have. So we started [00:01:00] quite early with quite a technological focus. And myself, I'm a technologist. Um, I was open source, uh, contributor to IMC three, spent a lot of time in the Python ecosystem, spent a lot of time, uh, doing evangelism, but also, um, working as a.

Professional data engineer and data scientists. Um, and lately I've been, uh, delving quite a lot into AI agents and very excited about what that means for the future of, uh, software development and seeing a lot of that in our own internal teams. 

Mehmet: Great. Yeah. And thank you again, Peadar. And you know, we're gonna talk about what, you know, AI a lot and of course, um, what you're building.

The way I love to do it with all founders or co-founders that I host on the show. Uh, although like maybe I have the answer, but this is just to clarify a few things to the audience. So back in the days, you know, when you started AudioStack in 2019, right? So what kind of problems in the audio production felt [00:02:00] so broken that you thought like it is worth to build something for this from scratch?

Peadar: Sure. Thanks for that. So I guess like to go back to like what kind of underlying problems there were. I think one thing is just like time to value, right? So I remember like even looking into this, like before I set up this company and like the, the, the time it takes to, you know, to get an ad produced or to get some audio book produced or whatever, you know, involves a lot of kind of manual processes.

And I think also that, um. Uh, it's also quite interesting when you, when you get, uh, things to like a human level and a human level kind of opens up like new use cases and stuff like this. So that was kind of the underlying, I wouldn't call it a thesis because I don't think we were that, uh, sophisticated about it at, at the time, but uh, that was kind of like the underlying, uh, thoughts at the time.

Mehmet: Now [00:03:00] people often think that. You know, audio is, is, is a creative medium first and a technical one second, how we, you can help us, you know, and when it was a moment for you to realize that actually audio production was a systems and infrastructure problem more than, let's say creative, uh, medium. Um, 

Peadar: I, I, I still don't think it's.

I think the creativity aspect is still very important. So like we see that time and time again with our customers. The whole, you know, and, and even internally, there's like a, there's a creative aspect to it, I think. What, um, and that, and that I think is one of the things that. Doesn't go away. I think ai you can, you can automate tasks, but you can't automate, you know, human creativity and, uh, and, and you know, the ability to think through, you know, something quite end to end, you know, you know, even like subtle stuff like, you know, what sounds would represent this [00:04:00] brand better and, you know, and how do you get this kind of this.

This message across. And that's, you know, um, when you're working with stuff like that, you're often working with creative teams and you know that those kind of details are very important. What I think. Well, I think what we realize is that like just accelerating that time to value, like just allows people to experiment faster and by experimenting faster they can, you know, they can do things that are more personalized or they can do things that are more localized, but, but also it just allows, you know, you know, that that allows like kind of, you know, even an acceleration of creativity.

Because what we've seen is that there's kind of like. Quite manual process is put an inherent bottleneck on the ability for people to, you know, to iterate faster. And I think that's one of the things that I've seen. Across kind of, I think this is one, like one of those meta trends about ai. I think AI is [00:05:00] changing work, you know, for, for all of us.

But it's also like on, you know, it's not getting rid of that kind of like, you know, the, the things that you. Cannot be put into an algorithm. So I think that's, you know, like, and that's like, and that's taste, discretion, the ability to take risks, the ability to be creative. And I think absolutely that's, you know, you know, that's, that's one thing like that.

We see it time and time again. People go, oh my God, we could do this so much faster. So what does that mean now? And, oh no, we were thinking about this before, but we were inhibited by budget time, whatever. You know, that kind of like unlocks, you know, you know, kind of latent your abilities, you know, that, you know, and it's fundamentally empowering.

So I think, I think that's, that's a part I gets most, me most excited that, not this kind of dystopian viewpoint, but you know, a, a very much a human empowering viewpoint. And I think that's ultimately the most exciting thing about what we're working on, what we are working on as part of a much wider set of [00:06:00] entrepreneurs, uh, academics, et cetera, you know?

Uh, who are all trying to, you know, try and move the needle, uh, forward at the moment. 

Mehmet: Perfect. Now I, I always like, and to hear it from. A technologist like yourself, Peadar, like if, if you want to describe to us a use case, right, like a real scenario where customer will benefit from what you're building. If you can like, just, and feel free to talk about it as much to describe it.

Like, you know, uh, as I was telling you, there's no time limits here. So just because I want to audience to understand. Sure how exactly you help them. And I know, like for a fact, uh, you shared this with me before, like even you have some use cases here in the UAE also as well. Yeah. So if you want to shed some light on that part, and maybe later I would follow up with some maybe deep dive questions.

Peadar: Sure. That's no problem. Thank you. So one example that we have in the UAE, I'll, I'll just use that example, uh, is a financial, uh, services provider and they wanted [00:07:00] to do, you know, uh, three ads a day with, um, different security. So gold, Bitcoin, Tesla, et cetera. Um. And, uh, for each of the seven Emirates, right?

So they wanted to do, you know, good mornings. Mm-hmm. Um, good morning, uh, Dubai, good morning, Abu Dhabi, et cetera, et cetera. And they wanted it to be all, you know, different for each, uh, kind of like, uh, Emirate and, uh, you know, a variant based on, uh, you. Uh, price data and stuff like this. So we linked up to their API, we pulled the price data.

We pulled, you know, things like time of day. We, and we produced hundreds of ads that went out, um, in a localized format throughout, uh, you know, UAE, you know, which allowed them to tailor their message to, you know, a much more personalized experience and also. Fundamentally, you know, drive downloads, [00:08:00] which was the metric they cared about, but also like, you know, there's an element of novelty there.

So, you know, the, the inhibiting factor was, you know. Doing hundreds of these, uh, ads was too cost, you know, you know, too costly, too, too monotonous. And we just sort of like allowed them to unlock, but also an element of kind of creativity there by, you know, you know, them being able to, you know, pull kind of live data that they weren't really able to do before.

So that was a, you know, that that's, you know, kind of very much a UAE specific case, all those cases, but I think it's the most exciting one for your listeners in the uae. 

Mehmet: That's awesome. Uh, Peadar, like now when, when you know, you pull, you try to pull, um, live market data, you know, to trigger content generation, and of course, as you were mentioning, you need some time to specify it for different cities, cultures, and especially the UE is interesting because, you know, like, uh, [00:09:00] like we have.

200 plus nationalities plus like, you know, the ee itself, like, uh, culture, it's diverse also as well. So where, where the, you know, from, from AI perspective of like, let's me call it like all generation perspective and the data where, where the complexity shows up and, you know, any, you know, things to, to, to overcome these complexities, like whether it's by orchestration, automation, or maybe something else.

Peadar: Yeah, so the question was, you know, what, what challenges were there to be overcome? Yes. Um, yes. I think like a fundamental thing was, um, you know, having checks and balances to make sure the right, uh, things went out and, you know, the systems were orchestrated correctly. Um, we also had like, you know, fallbacks that we had to build into the system, if, you know, you know, if data didn't come through correctly.

Um. Uh, there was a [00:10:00] bit of like optimization to make sure that things happened in a fast enough period. 'cause latency's quite valid, quite important there. Um, but, um, I think like, and also just like a more, just from a software point, point of view, um, I think there was a challenge going from like quite a.

Rough proof of concept to a more robust solution that, you know, would work with, you know, you know, hundreds of variants and, and you know, and allow us also the ability to do, you know, some sort of internal quality assurance because it's quite important that, you know. Uh, you know, that that didn't go wrong as such, you know, because the, you know, the, the, the media that you're producing at the end is actually the important output.

Um, so there was quite a lot of things like this that emerged in that is, is that kind of what you're at after for that 

Mehmet: question? Yeah. Yeah, yeah, yeah, yeah. Absolutely. Now, you mentioned something [00:11:00] interesting, Peadar, which is actually I was preparing to ask you about like, checking on the ai, right? So, um, how do you think about.

You know, the observability of AI generated, uh, audio, um, like how we can know or what are like the, the measurements we can take to know, like it's knowing the right thing. And sorry to make it like a bit of loaded questions because you know, recently I'm discussing with a lot of, um, CTOs and, you know, VP of engineering, uh, at different startups about keeping the human in the loop.

So how much. Of automation is too much and how much of also like, keeping the human in the loop, it becomes that kind of, you know, uh, a, a, uh, a, a uh, a way of slowing you down, I would say. Yeah. Act like a bottleneck. 

Peadar: Yeah. Um, I think the, the bottleneck is an interest. I think. Like, frankly, I think a lot [00:12:00] of this space is still new.

Right. So you are like. So I think that we're seeing the emergence of things like evals and like evals of like large language models is seems to be a relatively new term and we're seeing an evolution of like, of tooling and observability there. One thing that I think is interesting whenever you're looking at it is how used case specific.

Um, some of those kind of like validations or scoring are so like, that feels like a lot, like, you know, it's almost taking the human's discretion, you know, you know, because we have like a, you know, internal team who will do a lot of like, like internal qa, both by listening to or observing some sort of metrics that they have.

So some of that's automatable, but some of that's more, you know, taste specific. And I think there's like. There's, and, and, and of course there's also things like, you know, using, you know. LMS as judges and stuff like this, which, [00:13:00] which can work on things like, you know, like around kind of like script generation.

Um, but even that, like, has a fundamental stochastic aspect to it. So I think, like I still, you know, I, it feels still very early and I think, like, you know, I've seen, I've seen the emerges of things like, uh, rag pipelines. I've seen the emergence of like, you know, like. Better kind of like test data. Like, you know, there, there still feels like a, you know, a, a, a need for kind of human and lu.

I think that's where the kind of emergence of four deployed engineers comes from. You know, because config can go wrong and deployments can go wrong. Um, and there's an element of like also managing perceived risk on the customer side and, you know, you know, to make these systems work as well as possible.

Um. I think we'll probably see some emergence of some kind of better standards over time, but I think we're still. Like [00:14:00] figuring out a lot of the, you know, like there's no, there's no solution you go to, that's a one click thing that observes everything. And you know, like I, we don't have like a, uh, like a well established kind of DevOps like.

Like kind of set of, uh, principles, um, you know, kind of like Google kind of SLAs, um, like the kind of SRE book. I think we're still seeing the emergence of that, and I think that's where like, like discretion, taste, you know, and also like spending time with the domain experts to try to pull that into some sort of automatable, uh, solution in itself.

Mehmet: Right Now I gotta ask something. And by the way, because I've seen also, uh, in, in your background, uh, you studied, uh, uh, I think, uh, philosophy, correct me if I'm wrong, right? Yeah, correct. Yeah, yeah, yeah. So, so might, it might look like a philosophical question. Now. I, I know you work on the, you know, from, from audio [00:15:00] perspective, maybe on, on the media side of it, what I say, like maybe generating ads, but you know, when it comes to.

Music maybe. And, you know, so there is a debate between people like, about this and about like AI getting, um, like even the voices. I, I remember there was a case where someone, uh, tried to sue because, you know, her voice was similar. Like what it comes to these things, Peadar, like from a technologist and also from a human perspective.

How do you think about it? The reason I'm asking you this, because I know also. You want as a technologist, as a CTO, you want also to get the best. At the same time you have to put in mind, like, we need to do it in a way that, you know, people will accept it. How do you approach, you know, these controversials if the, the word might be correct here.

Peadar: Um, yeah, thanks for that. So I think, um, [00:16:00] my understanding of the question is like, how do you think about like, uh, trust and safety and, you know, kinda. In, in terms of AI technology and copywriting? A little bit, maybe, sorry, copywriting. Yeah. Yeah. So I mean, yeah, so fundamentally copyright, so that, that's very much, uh, like, you know, you know, vendor management, uh, legal agreements with your, your, your suppliers.

Um, looking at things like data protection. I feel like a lot of that is relatively. Well established, um, in the past also avoided certain vendors who we were a little bit concerned about their, their attitudes to these sort of things. Um, so, you know, that, that, that's I think part of, um, that, you know, provided a customer, uh, trust kind of thing.

Um, there's also like an element of, um, you know. You know, producing better guardrails, um, [00:17:00] you know, staying on top of like the, you know, different regulatory expectations and also like customer. You know, demands like are getting more sophisticated. So like I'm seeing more in like, you know, kind of like procurement processes more, you know, in depth discussion about like where data is stored and, and, and, you know, and, and you know, and how, how you audit these systems.

And I think we're seeing the emergence of things like EOM, like. Uh, you know, compliance standards for, you know, for AI systems in particular, you know, like ISO 4 2 0 0 1. Um, in much the same way as we, you know, I mean, we're SOC two compliant, but we also, you know, that's very much a SaaS like compliance standard.

Um, so I think, I think you have to be, you have to realize that you're a partner to. To customers. And if you're dealing with, um, kind of like large media and entertainment companies, you know, you know, their, their IP [00:18:00] portfolio is important in itself. So to do anything that would, would be, you know, somewhat shady, would, you know, kind of invalidate that kind of relationship.

Mehmet: Great. Um. I gonna shift gears a little bit. Uh, sure. You know, Peadar, and ask you more about the engineering part of, of what you do. So you spoken publicly about shipping fast and early. Now at all, stack, what does it this mean in practice? Is it like code production? Is it custom value delivered or is it something else?

Um, 

Peadar: thank, um, so I, I interpreted that question about like, the importance of, um, shipping fast. Yeah. Um, I think it's very important to like create a cadence and like, and also unblock that cadence. 'cause sometimes that cadence can slow down. I think when that cadence slows down, you know, we ship. We ship at least once a day.

We ship several times a week. Um, we have like a sprint demo [00:19:00] session, you know, every Friday. And when people showcase what the, you know, what's being delivered, uh, we try to make that as customer f. Facing a customer focused as possible. Sometimes, you know, sometimes some stuff is like infrastructure's a bit more detached.

So it's not always perfect. But, um, I think it's, you know, I think it's a very interesting question to ask. You know, what happens, you know? You know, why, why did this process slow down? And what often emerges is things like, oh, the CICD system is not sufficiently well developed. Or the testing is harnesses are not built in there so that people don't feel comfortable delivering.

Or, or, or there's some sort of fear about like, um. You know, because I think the ultimate fear I have is like, is code not shipped and building up, uh, you know, just fundamentally just builds up risks. That's why shipping small [00:20:00] increments is so important. That's why, you know, shipping large prs with lots and lots of like changes, you know, fundamentally increase.

Increase the risk that, um, that you can have. And I think also like developers like to deliver value, right? So I think it's a very important kind of cadence that's important and it's very important part about being in a startup. And I think it's also one of the things that helps you retain talent and also attract talent from, you know, shall I say it?

Slower moving organizations. 

Mehmet: Excellent. Now. And if I want to think about that, um, Peadar, from customer point of view, for you as, uh, as a CTO, uh, because I know like you talk also a lot about customer obsession, right? Um, so how do you balance something that, you know, the customer might ask for, maybe a feature and, you know, [00:21:00] making sure that.

You deliver it fast, but at the same time it doesn't, I'm just giving an example. Maybe it could be something else. Uh, maybe building this feature fast would cause probably, maybe kind of an imperfection, maybe a technical depth, maybe something else. So how do you balance these two when, when the requirements come?

Peadar: Um, so I guess the question was about how do you balance, um, you know, the need for quality with the need for speed? Yes. Um, I think like, I think someone, it depends on like everything. The answer is always, it depends. Like it depends on like the kind of like. The, I guess the maturity of that feature set. So like some features are, you know, and how, you know, how, you know how much impact they have throughout your wider software stack.

Um, [00:22:00] you know, you know, some security features or permission features or examples of that. So, you know, those take a bit more planning and a bit more kind of rigor, um, for kind of smaller things I think. I think, you know, really looking at, you know, how do you produce the smallest increment of, of, of customer value while also, you know, also, you know.

Where we also do it is we are kind of looping in our product teams to have a more holistic look at it. Right? So not like customer asks for some sort of configuration, you know? Is that a, a wider set of problems, you know, that affects a wider number of customers, or, or is this just like a kind of like a one, you know, a custom build for, for, for one customer?

And I think there's also an an oscillation of that sometimes. What appears to be custom for one customer becomes, you know, more impactful, um, for others. So I think [00:23:00] there's like a, a bit of just like, like, like managing that pendulum, you know, and, you know, and, and managing that risk and, and, and also thinking of, and trying as much as possible to think of things in like a more product mindset.

Mehmet: Great. I'm gonna ask you something maybe related more to leadership and culture. Now, Peadar, in the age of ai, and when I say age of ai, I'm not here talking about, you know, the coding part of the things more about, you know, the need for agility. Things can change very fast. So from your experience and you've scaled teams, uh, before, and you, you know, you're doing it.

Uh, so how do you maintain, you know, the. Cultural behaviors. Right. And what kind of cultural behaviors, uh, do you try to encode early? Because, you know, um, that they would compound later, and I'm asking this maybe how you did it at AudioStack and also maybe for [00:24:00] someone you know, who, who might benefit from, because the culture today is, is very, I would call it dynamic because things are moving in, in, in a very fast pace.

So what's your take on this? 

Peadar: Um, so I guess my, the question is, um, what's my take on. You know, cultural behaviors in the age of AI when things are changing, I think, I think like we, we often try to lean into curiosity. Like, so I think curiosity is a value or, you know, behaviors around that and interview process is quite important.

Um, I think there's a need to, you know, show an element of, um, you know. Self motivation, you know, the ability to handle, you know, things that might be that, that there is no playbook for. Right. I think, and I think fundamentally, you know, in a, a fast scaling company, you know, the playbooks of old are kind of going out the way.

And I think that's one of the [00:25:00] more exciting things about, um, you know, go to market and, you know, coding and various other things. Um, you know, and you know, there's a possibility that. Programmers, workflows in five years time will be sufficiently different than programmers, workflows. Now that you know, like, you know that, you know, things that people can learn in university is, you know.

Could become almost irrelevant. And you, and, and there's a challenge of, you know, of how quickly can universities adapt to these kind of things. Like, like a number of things we do internally that I think help you culturally is like, we have like lunch and learns you like, where people like, share what they're doing.

People put, um, like we have an ai, uh, slack channel where people are saying, well, this is what we did, you know, with ai, people are sharing, you know. Things like agents, MD files across teams where they can see, oh, this is how we used, um, you know, Claude code or whatever. Um, there we, we, you know, [00:26:00] we, we give, do some experimentation with like prompts, you know, and the sharing of these things.

We try to give like, um. You know, like, you know, have kind of internal wikis where people can see, oh, this is what you use for this, this problem. And, and, and that applies not just in engineering, but also kind of marketing you. We had like really great success with, you know, custom GPTs and kind of for like marketing content generation.

So I think like. You know, just, just being open-minded and like, you know, and just like, you know, you know, trying things and like, you know, getting your hands dirty and, you know, being able to experiment that I think like opens up like a range of kind of, um, kind of possible behaviors. 

Mehmet: Great. And, you know, as a leader, Peadar, like, and I'm asking maybe this applies to you and maybe I'm not sure if, if it's the same across the board as a CTO.

Where nowadays in this age of ai, you, you [00:27:00] feel like you need to spend most of your time, uh, on like, is it, um, architecture, like, is it like people, is it strategy, how AI have changed? You know, not the core role, of course, of the CTO, but I mean, the way A CTO would spend his or her day, you know, usually. 

Peadar: I think, I think one of the more, more interesting things about what AI means for, like, what leaders I, I'm making this more general than just CTOs, I think this is engineering managers as well, is the ability to sort of like still be, you know, using kind of like AI agents to be delivering code.

You know? So I'm amazed at like. One of the things I, I spent quite a bit of time on is actually, it's really quite tactical remark here, which is like, you know, around like things like security fixes, like, so can you use AI agents to automate security fixes? Security fixes are a kind of a long [00:28:00] tail of, you know, let's be honest, non often not prioritized work.

You know, they're very, very hard to get out the roadmap, but with the kind of. Ability for like, you know, bug reports to be generated and these kind of things to be generated, you know, in a kind of like semi-automated manner. You know, that opens up the ability to sort of, you know, like, you know, be much more productive around things like, like that.

I'm also very, I spend quite a lot of time on strategy and spending, like kinda looking at what, like new tools there are and kind, you know, the ability to, you know. You know, also like allow more generalists, right? So I think like, you know, I think one of the exciting things I've, I've seen is how, you know, designers, product owners can, rather than just writing, you know, like a PRD or you know, product review document can build kind of a prototype to actually show, you know, oh no, this is what I want.

Which also allows like, kind of like code [00:29:00] artifacts to be shared. You know, much more like. Like liberally, you know, and, you know, allow people to have a bit of a faster iteration period, um, or even more experimentation period as well. So I think, you know, just, you know, what does AI mean for productivity more generally is kind of like the kind of met up question.

Um, and like that also impacts things like observability and reliability of these systems as well, because I think we're still figuring out a lot of that. 

Mehmet: Just a follow up quickly on, on, on what you just mentioned, and this is something I didn't prepare for honestly. Um, do you see a, I I, I've seen, you know, in some organizations, kind of, they have unified the role of product and um, uh, the technology like CTO and CPO.

So I've seen C TPOs or C PTOs, whatever you want to call them, do you think. 'cause just of what you mentioned, we're gonna see more of this like where [00:30:00] Product mes with, with, with, with the CTO, uh, more than just having it or because of the nature of how things are done in software. Usually still we're gonna, in majority see two, you know, different people handling, uh, each one of these.

Peadar: I think it's still healthy to have two different people, but I do think there's like a, a broadening of some of these roles as such. Um, and even that just helps in terms of empathy of like, you know, understanding, you know, this is what, you know, this is how hard this would be to do, or having a bit more understanding of kind of the friction, I think like.

Technology often becomes a very, uh, kind of delivery focused role and products more about like what is possible. And I think it's quite difficult to have those, you know, some people do, but I think it's quite difficult to have those things all in, in one person's head. Um, but I do think there's like the, there's like a [00:31:00] widening or kind of, or change in the shape of those roles that I think is happening.

So I think maybe in a couple of years we might be using some different acronyms. 

Mehmet: Great. I gotta go back quickly because, you know, um, maybe we didn't talk about this to, to the AI and audio in general. Um, do you think, or let me ask you this way, um, AI is becoming mainstream, right? So what, what would be the main difference between a great audio experience from a generic one?

In the future.

Peadar: Okay. Um, I think, I think there's still, so like the question is like, what's everything a great audio experience in the future? I think, I think to go back to the creativity aspect, I think like the details and the kind of nuances I think are [00:32:00] always, um, uh, very, very important. Um, I think, you know. There were, there were questions about like, kind of like, uh, sym synthetic voice, uh, um.

Like how, how good that was. And I think that feels, I sort of hit a, a limit of acceptability, but I think there's still like levels of detail underneath that. And particularly around things like, you know, different languages and stuff like this, that these are always kind of like, you know, lagging factors.

Um, you know, in terms of like the kind of march of technology. Um, yeah, that's kind of like the best I can answer to that off the top of my head. Um, yeah. 

Mehmet: Yeah. That, that, that's, that's great actually. Now for builders, founders, CTOs, uh, in that space, um, what, what do you think they should be cautious about, or let's say [00:33:00] they would.

Be spending a lot of time on in the audio space, I would say, or anything creative by the way. It might be, you know, it might be video, might be art. I don't know. Where do you think, you know, the, the, the main focus for founders CTOs would be, um. Uh, w with all what's happening currently with ai? Yeah, 

Peadar: I think like, so the question is like, what should be the focus of like, founders, um, building and these kind of like, I, I'll just call it like creative media.

Yes. Yes. If you want, if you're, if you're looking for a kind of, uh, naming convention, I think it's quite easy. To like hack together prototypes. I think it's quite difficult to build like robust, reliable systems. And I think like some of that's around questions like ux, you know, like what is the, you know, what is the design UX experience that, that a customer expects to be able to do something, you know, at various levels of granularity.

[00:34:00] Right. So that I think is a, a, you know, a challenge in itself, both in terms of latency, both in terms of like. You know, like even like how you represent various things, you know, and. And, you know, and you know, like, and that's like in the same way that, you know, building a text editor is still non-trivial.

Even though we have a lot of like, kind of the functionality out there, um, there's kind of questions of like, you know, you know, how do you integrate into other systems and add more value to end? And then there's like, you know, there's just questions of scale. So how do you do it when you're like, constantly people are using it, it's growing.

You have like multiple, uh, users in multiple, uh, countries around the world. You, how do you support them? How do you handle kind of like, uh, you know, and how do you like, like get to the levels of like, um, you know, of management and configuration that they expect to be able to add, um, kind of, um. To add value inside of a large [00:35:00] organization.

So I think that's kind of like the things that we, you know, like that I think are the hard, hard problems. And you know, and I think fundamentally, you know, building prototypes is easy, but it's like the last 10% of, of any software development is still hard, superbly hard. 

Mehmet: Uh, that's very valuable. I, I, I can't agree more on this because I think it's only not only in software, I think in everything, you know, I mean, because people, and this is my theory, you know, we humans, we like to take shortcuts and we think like, but don't get me wrong, like, AI is great getting you, you know, the, the, the first.

Few steps as you mentioned, but you know, you need to put your human perfection. I think at the end, and this will be the secret sauce in my humble opinion, when it comes even to software products and, you know, anything that comes from from AI where we add the human touch. Uh, on top of it now as [00:36:00] we are coming almost to an end, uh, Peadar, for, for today.

And really, I enjoyed the conversation. Any final, uh, things you want to share, any final thoughts and where people can, uh, get in touch and learn more? 

Peadar: Um, you can check out our website@ww.AudioStack.ai, and I am Pat Coil. I'm, I'm findable on LinkedIn. Um, in terms of like final thoughts, um. I am very excited about, you know, how the kind of progress and like, um.

And like large language models in general and the effect us having on software development. You know, I've been spending a lot of time playing with or experiment a cursor and Claude code and encourage, if you're not enabling your teams to use them, you, you should definitely be, be doing that and, and sort of be building your own codes of practice internally.

'cause I think that's probably the level on top of that's, [00:37:00] that's harder than just the more kind of vibe coding, uh, uh, uh, area. 

Mehmet: Sure. You know, this is, this is, uh, uh, great. I would say now, uh, Peadar, like for the links you mentioned, I will make the lives easy for my audience. They will find in, in the show notes, uh, if they are listening on their favorite podcasting app.

And if you're watching this on YouTube, you can find in the description. Peadar, I can't thank you enough. For your time. I know how busy it can be. Uh, you said like, uh, you have your, uh, weekly Friday meetings, so I would not, uh, I would not, uh, take much time from you. Again, thank you very much for being here and this is how I had my episodes.

This is for the audience. If you just find. Us by luck. Thank you for passing by. Passing by. Uh, if you enjoyed, give me a favor, subscribe and share to dear friends and colleagues, and if you are one of the people who keeps coming again and again, thank you very much for your support, for loyalty, for making the CTO show in Mead even in 2026, similar to 2025.

Like [00:38:00] we are still ranking on the top 200 apple po, uh, podcast charts in different countries. Thank you very much for this. This cannot happen without, uh, you know, your support. And thank you for tuning in. As I say, always stay tuned for your episode very soon. Thank you. Bye-bye. 

Peadar: Okay, bye.