Interview by Jonathan Kersting
Can AI help you build software faster without leaving your business swinging from a technological tree branch?
Adam Wynne, founder of Pittsburgh-based Wynne Technologies, joins TechVibe to explain how his APES methodology, short for AI Project Engineering with Specs, takes AI-powered development beyond flashy demos and turns promising ideas into secure, scalable, production-ready products.
Wynne details why vibe coding can produce something that looks appealing but may not be ready to support hundreds or thousands of customers. His approach applies specialized AI agents across the entire product-development jungle, from customer discovery and technical specifications to coding, testing, security reviews and ongoing operations.
Hear how APES can build software up to 10 times faster, continually learn from emerging research and deploy teams of adversarial agents to inspect products from the perspectives of developers, users and security architects. Wynne also explores how the technology is helping companies improve sustainability, manage energy data and identify AI projects that deliver measurable business value.
So grab a banana, meet Tom the ape and tune in for a conversation about building software that is more than just a pretty bunch of code. This episode is packed with practical insights for founders, entrepreneurs and business leaders ready to turn their next big idea into something truly scale-ape-able.
The Pittsburgh Technology Council produces TechVibe to explore Pittsburgh's technology and innovation ecosystem.
Trannscript:
[00:00:00] The challenge is though, if you d- if you focus on vibe coding, you end up with a really awesome looking demo. You can get pretty far. But not all the way. You can't get all the way. And a lot of what becomes challenging at that part is, okay, now how do I, how do I make this production ready? How do I secure it?
How do I actually onboard my first 500 or 1,000 customers- Yeah ... and know that it's not gonna crash?
It's Jonathan Kersting here, bringing you another edition of Tech Vibe. As you can tell with our new mascot we have here right now, we're gonna have a ton of fun today. I am hanging out with Adam Wynne of Wynne Technologies, and there's a reason that we have an ape hanging from our TV set today, and we're gonna go all into that.
It's all about a platform that Adam has developed that I think is really cool when it comes to coding with AI. It's a thing. So many Vibe coders out there, Adam, and we're gonna really just explore how your platform is taking things to the next level. Right. And I think this is a very exciting iteration for that.
I see this technology [00:01:00] moving in. Yeah. And you've got some great experience. So when I was looking on your website, looking at your background and stuff, talked with you over the past few weeks, I'm like, "Man, you know your stuff." So welcome to the show today. Thank you very much. So glad to have you here.
Dare I say we're gonna peel a banana back on what's gonna be- ... a great show. It'll be appealing, that's for sure. Appealing, yeah. So no more puns from me, but you can put as many puns in as you want because you have the ape platform. So thanks for being here, man. Thank you very much. Yeah. So first off, your background, I mentioned before you've got some really cool experience working for some great companies and great projects.
Give us a little taste of what Adam is all about. Yeah, so I have about 25 years of experience- ... doing software engineering. Worked, started working in the internet field. I was work- worked for Amazon back in the early days during the whole internet boom and bust. Oh, cool. That's kinda fun being in Amazon in those like seedy, heady days, right?
Yeah, I mean, and it feels a lot like actually it feels now where- I was gonna say, right ... where the, you know the technology is burgeoning and it's like everybody knows it's gonna be big, but we don't know exactly what it's gonna be. So you know [00:02:00] it's a great time to be an entrepreneur.
It's a great time to be starting a business. It's a great time to, basically take a look at all of those software projects you always wanted to implement- There you go ... but thought it was gonna be too expensive. Exact- or take too much time- Or take too much time ... and all that kind of fun stuff.
And so what inspired you to say that I have this methodology, I have this system, I have this experience to be like I'm hanging my shingle for myself to do this so I can actually help companies? When I was working for the Department of Energy- we built high-performance architectures for processing- ... large amounts of data, analyzing it, visualizing data in new ways and things like that. Okay. After that I went and worked for Bosch, where I worked on a lot of edge technologies, built doing video analytics on the edge taking some of their hardware products and turning them into SaaS application software as a service applications that are able to read in data in real time, present customers- Wow
with dashboards- Okay ... and slice and dice their data in different ways. And then once the whole kind of AI coding revolution came a- came along, mid last [00:03:00] year when it seemed, it, it looked like it was really gonna take off and be the way of the future, I said, "Okay, now's the time to kinda take advantage of this-" Love that
And apply all my knowledge and then basically use all my software engineering experience and basically build an entire product engineering team out of software agents. That's really cool. I love that. And what I love most about it is you're, like, guiding the future of this now. You're actually able to shape how we can actually code with AI.
And what I love about it is 'cause part of your platform, Apes, of course, is AI project engineering with specs. Right. Right? And the specs is the important part, 'cause you can actually see what the code is. I guess for people that are kinda new to vibe coding, maybe I'll give a little bit of a thing, and you can correct me as I go along.
Okay. But the idea here is literally people are coding just by telling, a chatbot what they want out of a program, and it comes back with the code to do that. But you don't really see what's under the hood on that, and so you get something that you think is working, but [00:04:00] you don't really know how it's working.
Where you're able to do stuff where you can actually see how it's working, and you can document- Right ... inspect it, right? Yeah. So I mean, the tool, the tooling's at a place where, professional software engineers can be super, super productive. They can generate code, look under the hood and see all the code that was generated, and have an iterative approach to be able to feed back to the agent what they should change.
But also non-coders can also start and prompt their way to really good-looking websites. Yeah. Yeah. And really good-looking applications. The challenge is, though, if you do- if you focus on vibe coding you end up with a really awesome-looking demo. You can get pretty far. But not all the way. You can't get all the way.
And a lot of what becomes challenging at that part is, okay, now how do I r- how do I make this production ready? How do I secure it? How do I actually onboard my first 500 or 1,000 customers- Yeah ... and know that it's not gonna crash? Exactly. Tell me how your system does this, 'cause this, I think, is just...
As you were explaining this to me, I'm like, this is that missing piece now. Yeah. Where everyone talks about, "Oh, people are just gonna code stuff. They're not gonna..." It's no, we always are gonna have... I'm [00:05:00] not gonna say the human in the loop. I'm gonna feel like the human is the loop in this, to a certain degree, 'cause they're controlling all of this.
Yeah, the human's leading the loop, I think you could say. Ex- leading the loop. Yeah. I like that, too. Very cool. Yeah. So explain to me about how this platform works, 'cause this is game-changing to me. Yeah. It really is. And where most teams are starting is just by applying the coding agents to the actual coding process.
Okay. But if you look at the whole entire value chain for developing a new software product- Yeah ... what you find is that actually maybe- 20% of the whole process is actually writing code. There is working with customers, there's developing requirements- Okay ... there's writing those technical specs, there's testing it in detail, there is operating it.
So all that, is part of the, is part of the whole what I call the value chain of product engineering. Okay. And so if you optimize the heck out of the coding step you've actually probably might not even move the needle on how much faster it is to build a product. That makes sense. If it's just 20% and you make it so much faster, then okay you made a difference, but it's not like that big of a difference, right?
Maybe you save 10 or 15%, but you know, [00:06:00] your colleagues might wonder, well, for all that effort, what did this- Yeah ... really get us? But what I'm doing with my methodology that's called APES, is I'm actually looking at the whole product engineering cycle. So I will sit with a customer, we'll do the whole discovery together over a few days- get transcriptions of all of our discussions, get their raw documentation, any of their analytic calculations they've done ahead of time. I feed that into whole process. I have different agents that are set up for different stages throughout- Yeah ... the whole process.
Very cool. So I take this from raw discovery material to rough requirements to more technical specs and then once the specs get very technical, then you can hand it off to the coding agents. Exactly. That is... See, that is really cool., I've never seen it this way, which is why I'm glad we're talking about, 'cause I want everybody to know that you're- Yeah
here in Pittsburgh doing this. Yeah. That to me is really cool. Yeah. So I mean, so and I, honestly I think I am on the leading edge here because I see lots of people applying these techniques in their slice of the of this- Okay ... value chain. Surely. But you know, I'd like, I like to take a full perspective of the whole thing and see it through because what happens [00:07:00] is one of the problems that can happen with project product engineering is that, when you have what the user wants, that's defined upfront, but then you hand a technical spec over to the team- and the product manager is hoping that it comes out right. And so what you have a lot of drift of the user's intent and what the user needs. Yeah. When there's hope you're gonna get drift. Yeah. Right. That's how I see it. And it's just, it's just working with people. It's challenging.
The stuff gets, we all know the telephone game. Stuff gets dropped between the lines. And with more remote work, people aren't sitting around the wa- water cooler working these things out all the time. Exactly. Right. So, what these specifications allow you to do is basically the product manager and the coding agents can agree on, okay this is the shape of this.
This is what it's gonna look like. And then when you hand this off to the coding agents, they're actually referring back to, oh, this is what the user wants. This is the type of users. All the personas are built in. Okay. The coding agents can refer back to that. And so it really carries that intent through to the product completion.
So I'm assuming as you're working with clients, they gotta be pretty mesmerized when you can come back with a [00:08:00] finished project ready to go in just a fraction of the time. Yeah. And- Almost like they're like, "Wait a second. Let me double check your work here." And yeah yeah, the it is amazing. People are starting to expect it right now, to be honest with you.
Okay. Okay. Yeah. They're, "Can't you just throw that AI?" Exactly. Exactly. It's "Well, yes, but no." Yeah. So tell me about GridPulse- ... and what's going on with that. This is an offshoot of what you're doing with Apes and everything like that. I'm just curious- Yeah ... 'cause- So I work on a l- a lot of business-to-business applications with clients- Yeah
that I can't show- Okay ... all the time, and so I built GridPulse as a reference implementation- That's what it is ... that used- Okay. Yeah, so I used Apes to build GridPulse really fast. And that shows how you can use Apes to build- Yeah ... something fast, and then, how you have as a tool- And I collected data-
use with your clients ... on how long it took me to do it. Gotcha. And so- Okay, that's where that all plugs in. Yeah. Now it's all- Yeah ... it's all coming together. Yeah. It's all coming together. So I mean, I showed that you can build, 10 times faster using this kind of a framework- Right ... versus the the previous way of doing it, circa 2024.
Circa 2024. I mean, when you say circa 2024, [00:09:00] it's hey, that wasn't that long ago. Right. But in AI time, it was an eon ago. Right. So, I mean, I'm just curious, like, how do you see this moving forward? I mean, it's not like AI is "Okay, we're done evolving now." It... Like our primates here- Yeah ... they'll be evolving, but at a much faster rate.
Right. So how do you keep up with that? It's a halftime job, I would say, to keep up with all of the research- Totally ... and all of the stuff that's coming out. But what I've been able to do with my system though is I actually make Apes continually learn, so- Well, that makes sense.
Yeah. Okay, neat. Tell me how you're doing this. Yeah. This is interesting. So every month, every week there's new ideas coming out. Somebody develops some new fancy open source way of using Claude code. And so what I do is I actually have Apes go and read the code, the code- ... their code base.
Apes can read. And- I love it. Apes can read. And so n- and then we can have a discussion, okay what's different about this compared to what I'm, we're already doing? What techniques can we take? What architecture ideas? Yeah. And that kind of thing. And also can look at research papers and things like that.
So it's always learning. Yeah, so- And then it's bringing that [00:10:00] back. Now as it learns, how do you know what it's gonna change? Can, does it bring recommendations and you say, "We're gonna implement that"? Like, how does that next step work for you? Yeah, me as the architect I'm really in the loop on this one.
100%. Right. So, so an example is there's kind of, a technique that came out recently that's called adversarial reviews. And so what this means in, is that, typically you know- If you're talking about all people, an all people project, you have a junior, mid- mid-level developer those kinds of folks write features, and then more senior people will do the reviews.
And so you typically want at least two people to do reviews, and their job is to poke holes in it. So th- that's where the word adversarial comes from. Yeah. They wanna- But what you can do with- They wanna pen test it, right? Yeah. Right. But what you can do with agents is reviewers are now cheap. And so now I can write an agent or four agents that go o- one that takes the perspective of a developer, one that takes the perspective of a user- I love it
one that takes the perspective of a security architect. Okay. And they try to poke holes in it. That is amazing. And so I took this idea from from a framework and from some research papers. And so every time I do a [00:11:00] review, every time I finish a feature, I- I say review. I say, "Apes, review," and it- Apes, review.
They're like and it just spins up, it spins up four parallel agents and pokes holes in it. Amazing. And sometimes the agent's "Wait a second. We documented this pretty well. I don't know if we need to do the review this time." But every time I do it, it always comes up with something great.
That's amazing. So, it's just an example of, going out, finding a cutting edge technique, bringing it into my framework right away so the- Of course ... apes doesn't stay static. E- Exactly. These apes are on the move, man. Yeah. Simple as that. They are swinging from the trees. For sure. And what I think is really interesting about what you're doing is, I mean, obviously you're always learning and bringing back, and that just gets put right into your product, and then it goes right to your clients as well too.
Right. Yeah. 'Cause they're getting the tip of the iceberg pretty much at all times. Right. So I'm really curious, like, how you then would work with, how you work with your customers then. I'm assuming there are times when you're brought in 'cause it's a, maybe a new project, and you're like, "We're starting from square one," which is probably ideal, so you can get things done right from the beginning.
Yeah. But I'm assuming a lot of folks might be in the middle of something and [00:12:00] things aren't going right, or they thought they could vibe code their way into success- Yeah ... and then they're realizing "Wait a second." Yeah. I'm kinda there, but this definitely isn't ready for prime time," so. Yeah, so I'm working with a lot of kind of analytic-type applications- Yeah
where, maybe there's sensors in the field that are pulling data back. The data needs to be integrated with other data sources using APIs. And so a lot of times the customers will have developed- ... algorithms and, they have things worked out in spreadsheets- Yeah ... or maybe in, in quick scripts.
But they haven't operationalized it yet. So that, that, that's me catching them at the beginning to be like, "Okay, now Based on the fact that you've done all this legwork, now let's talk about how we can turn it into a a real, supportable project." 100%. Okay. That's awesome.
And so in that case, I can work with them, I can do the discovery, generate all those rough documentations, and go through that whole process I laid out. You can, man. Yeah. Yeah. The o- the other type of a situation is a lot of founders are actually starting to do the vibe coding on their own these days.
Yes. I'm thinking right now whole companies are being built 'cause "Wait, I think I might be able to code an app for this- I mean, yeah ... and that'll be my company." It's maybe. [00:13:00] Yeah. Yeah. Maybe. Yeah, so, so a lo- a lot of, founders, CEOs will actually start to vibe code their own product because they're like, "Wait a second, I got this huge quote from a traditional software company.
Let me try this vibe coding thing to see how far I can get on my own." They can build really nice looking products. But then oftentimes they'll get stuck at the point where it needs to then be deployed into production. Yeah. And so when I can come in at that point is I could be like, "Hey, I got architecture scaffolding that will help us make this ultra secure, make this scalable make it very cost effective as well."
I also have agents that are continually monitoring cloud usage to make sure that you stay within the budget that makes sense for you. Ah, I love that. And so it really helps you scale down to what you need at the moment, but then helps you scale up as soon- when customers grow. Gotcha. And I think the thing I keep kinda coming back to in my mind that I think is just so important is you mentioned the spec side of things, and that the idea that when this is done, like you know what's in the code and you can access the code.
Yeah. Like you own the code. Yeah. It's not just this thing is running, we don't know where it's running and how it's running- Right ... but it appears to [00:14:00] be working, and that's not how you wanna build a business, right. You wanna make sure that if you're building a business on something, you own it all.
Right. And you know how to tweak it and- improve it- Yeah ... and troubleshoot it if you need to, right? Yeah. Yeah, so all of the code, all the code and all the documentation and all the specifications, tho- that ends up in the same code repository, Gotcha. ... That a developer would work from. So, I can build something and hand it off to an existing development team.
I can hand it off to another developer. I can stay on and help- ... maintain and operate the system for them. How cool is that? You can provide that kind of flexibility. That's pretty amazing. Yeah. 'Cause someone might like, "I want you to hold my hand the entire time," right? Yeah. Or they could be like, "No, get it to this point, QB it over, and send it over to my- Right
to my team, and they can finish the rest." So that gives people lots of flexibility in how they wanna kinda build their project. Right. It's also good for founders too that they're looking for a technical partner- Yes ... but they're not in a position where they can give away more equity. Exactly. So I can kind of come and be like, "Hey, I've done this before.
I know what you're trying to do." Right. "And so I can scope it exactly to this, this level of price, and it's gonna take us this amount of [00:15:00] time," and they can get a professional product really quickly. So you get to have fun in a bunch of different ways. Yeah. It all day every day is fun. It's learning, it's building stuff and it's using new tools to do it.
Absolutely. And I, what I can see is this is your passion. Yeah. I can... The way you talk about it this is your jam. Yeah. And I think that's a, that's gotta be a lot of fun to be able to be working on the things you wanna work on and you have passion for. I know one of the things you've got a lot of passion for is sustainability and the green side of things, which I think is really cool.
Yeah. Tell me about your passion behind that and how AI's being applied to some projects maybe you're working on. I think this is a fascinating little green side- Yeah ... of what's going on with with Adam Wynne here. Yeah, so a lot of the a lot of the clean technologies that, that are coming around, so for example, you can have a, an a- a, an analytic system that deals with decarbonization for office space and stuff like that.
And so I'm working on a on a project with a customer like that right now, and they're able to basically dial in the efficiency of a commercial building- Really? Awesome ... so that they can save money at a predictable level for, the entire year and in, [00:16:00] in the future years.
That's, that's- And so as they're making- That's cool ... right, so as they're making green upgrades to their building- ... they can basically measure that those upgrades are being realized in practice. That's amazing, so you can actually... Yeah, you can now see in real time this is actually working or this isn't, and they can then adjust.
Yeah. And they have this powerful technology to do that. And it's also cool looking at the sustainability space. So corporate sustainability folks are typically like a one-person operation- Yeah ... but inside a corporation, and they're just constantly pulling data from spreadsheets- Right ... and they have to go get data from procurement, and they have to go get data from another place and manually pull stuff together.
AI's really good at, pulling data together, analyzing documents- Yeah ... analyzing policies and making sure that, we're meeting our goals and that kind of thing as well. You're basically giving whoever's managing sustainability, like a whole team to go out there- Right ... and pull all the good stuff so they can really make, they can show what's going on and see where it's lacking or where it's exceeding and- and take it from there. Right. So that's gotta feel really good to be like, "Yeah, one of my passions is this, and I'm applying my other passion to it." Right. So that's gotta be a long- Yeah ... it's a good way to live. Yeah. It is [00:17:00] great and there's a lot of talk in the region these days about energy with all the data- 100%
center stuff coming online. Yes. And that fits into the whole GridPulse angle as well. Of course. In Pittsburgh, I mean, we're kind of home to sustainability here. A lot- ... of sustainability that goes around the world has its roots that ties into Pittsburgh and what we have done- That's true, yeah
over the decades, cleaning up- Right ... our industrial legacy here- That's true, yeah ... which I think is pretty amazing stuff. That's true. So man, you're hitting it on so many different directions. Yeah. So what's next? I mean, obviously more customers is something that you're always looking for, but you're always evolving your technology.
Where do you see yourself moving over the next few months? So the, the obvious thing for me is to g- do product builds. Yeah. But I think also small, medium businesses need- to build out an AI capability as well. Ev- ev- every CEO I talk to- Yes ... they know they need AI.
They're not getting clear signals on what's working and what's not. And so I've done a lot of analysis in terms of what large scale consulting firms have fielded over the past year or so. Okay. And so then I can kind of turn that into a, I have another offering, which is kind of like a right-sized innovation program- Gotcha
around AI. Surely. That I can kinda come in and I can lead [00:18:00] a team of innovators in the company through a process of figuring out what, what use cases they wanna implement. What are the best ones? What are the ones that are gonna return value? And what are the ones that are gonna be the easiest to implement?
And so that we can show quick wins because, the problem that a lot of - companies are having is how do we prove the return on investment- Yeah ... that we're getting from AI? Well, that's why I'm really excited about the evolution that we're seeing now in the place. I mean, I, he, like eons ago, like last year people are still like, "We need to be doing AI."
And they don't even know what it means. But now it's like there's an actual business case. You can prove the business value. Right. So it, now I think it's really maturing to a level where it just becomes part of a function as to how a business that wants to be competitive does business at this point.
Right. Which is great for you because they need people to build these tools and make sure that they're built right. I can only imagine back in some of the earlier days, like '23, '24, that- Yeah ... people were building things, but they were wonky. You're seeing the potential but not getting the full potential.
Yeah. Or, or- And I feel like these are- ... or it would take- Yeah ... yeah, or it would take a year and a half to find out, [00:19:00] is this idea gonna fly or not? Exactly. So, one of the big things that's, a- this AI coding revolution has done is it's really shortened the amount of time that you can prove do we even have an idea or not?
You can figure that out in a- That's interesting. Yeah. That's a great point. Talk about a powerful thing for a startup entrepreneur. Right. You think you got this great idea, you think you're solving this problem that's out there, but now you can find out much faster. Right. And- Very cool stuff
and so it allows you to fail earlier- ... and keep iterating and keep coming up with new ideas that are gonna hit in the market. Absolutely. So my last question is, what is your favorite AI tool? So what I use, I base my product around is Cloud Code. Okay. You're a Cloud guy through and through? I'm Cloud Code on the command line on Linux every day.
There you go. Every day. Yeah. That's awesome. No, this is such great stuff. I'm very excited for what you're doing. I think this solves a lot of problems for people out there. I just love seeing people innovating with very powerful technology, and when they have a cool little mascot like our ape friend over here.
Yeah. This is Tom. Tom. Yeah. Well, Tom, thank you for being with us here today. You were very quiet back there. I guess we gave him his extra bananas ahead [00:20:00] of time- Yeah ... so it kinda kept him busy, but we're all good. But once again, a- thank you for hanging out with me today. Yeah. What a fantastic story.
Yeah. I'm glad you're doing what you're doing here in Pittsburgh, man. For having me. It's been a lot of fun. Absolutely,. I will see everybody here on the next one. And in case you forgot, this is Jonathan Kersting with the Pittsburgh Tech Council, and of course Tech Five.
I just love telling stories like this, and my new friend Tom.
[00:21:00]