Multi-LLM Skills Cold Open

SPEAKER_01 0:00

We have skills that we've created. So in Claude, we can actually call all of the other LLM providers. So I'm using Claude Code, and it's actually going through and creating images over an OpenAI for me. And then it quality checks them to determine hey, is this actually a good image or is this a cruddy image? And hey, go open AI fix this. So we get to use multiple platforms together. And I think that's hugely powerful.

Joshua Schmidt 0:27

We

Meet The Panel And Guest

Joshua Schmidt 0:28

are live. You are listening to the audit presented by IT Audit Labs. My name is Joshua Schmidt, your co-host and producer. Today we're joined by the usual suspects, Eric Brown, managing director of IT Audit Labs, and Nick Mellum. And our guest today is Alex Bratton. He uh has been on the show before. Alex Hales from Lextech. And he also has a super cool um, is it weekly, Alex, uh, AIY, your seminar that you hold, um, bringing everyone up to speed on AI topics. So Nick and I are fans and we join as frequently as we can.

SPEAKER_01 0:58

Yep. And that's something we tackle twice a month to just share and help folks, uh especially mid-market folks, figure out. So what do we do with this AI thing and how do we make it real?

Joshua Schmidt 1:06

So for folks that haven't seen the previous episode, maybe could just give us a quick little bring us up to speed about who you are and what what you do and what you like to do.

SPEAKER_01 1:13

Uh absolutely. So I love conversations like this because I am an applied tech geek. Uh, my title is CEO and chief geek for a reason, because tech is supposed to be helping all of us thrive at work. So, how do we do that? Uh, LexTech does that primarily in the enterprise markets. And then we've got this new division, uh AIY, because all AI really needs a Y to point it. And that's focused on how do we get real business value from AI, uh helping enable folks so they can do things themselves. And um just rolled out my new book recently, Capability Equity, really focused on how do we point AI at the right things and build value for the company by giving folks superpowers.

Eric Brown 1:52

And Alex, it's a little dangerous to um have you and Nick together. So I think it was maybe a year ago. Nick comes in and he's got the what's the Apple um the Vision Pro. The Vision Pro. He's got the Vision Pro on. He almost convinced us to get like 10 of them for the company. We were gonna do all Vision stuff.

Nick Mellem 2:12

We're still working on it.

Eric Brown 2:14

Then last week he comes in and he's like this new iPhone situation that they got, like the two iPhones glued together. The duo. Yep. Gotta get the duo. It got so bad, Alex, that he even convinced me I need the duo. So we both uh we're both getting the duo. I feel so bad for you.

Nick Mellem 2:36

I'm good at spending other people's money.

SPEAKER_01 2:39

If it's replacing the iPad mini that's walking with me anyway, it's perfect.

Nick Mellem 2:43

Yeah, it's perfect.

Eric Brown 2:44

I kind of go back to what was the first iPhone? Was it the two? The iPhone, was it the two or the three? What was the very first one? Uh it was the iPhone. Was it just the iPhone?

SPEAKER_01 2:55

But then it went from the iPhone to the three, then, didn't it? Well, because it was a 3G was it was making the uh switching the cell signals, yeah.

Eric Brown 3:02

And then they had the 3GS and so the original iPhone, which kind of blew the doors off of everything and started this whole thing that we're on. Um he I don't know, Nick, if you had that first one. I missed the first one. The first one I had was the 3G, and then it just took off. And I think the duo um might be the same thing. I know others have played around with two phones together, but I think Apple will actually do it properly, and it could be just awesome. So I'm I'm looking forward to it now.

Halloween Costumes And Tech Banter

Joshua Schmidt 3:34

Well, I'd like to get more into the duo. I'm sure you guys would too, but um, before we he head into more uh tech territory, uh let's do a little bit of icebreaker here. So if you're dressing up this year, what what are you dressing up as? And if not, what was your favorite costume that you wore? I think I already know Alex probably has a Star Wars answer, but uh let's start with Nick. He he seems raring to go today.

Nick Mellem 3:55

Yeah. Oh well, you know, I've got two young daughters, so I don't have a choice in this matter. Um and I was just found out maybe three hours ago that I've been volunteered to run a booth, a Halloween booth at my daughter's school for a Halloween party. So I'll be 3D printing some items for this table and doing some fun things there, but I think my daughter has told me I have to be Beast. So I mean, maybe it's fitting. So she'll be Belle, I'll be Beast. Our littlest daughter is gonna be our youngest daughter is gonna be the little teacup, I think.

SPEAKER_01 4:26

I'm sorry. Nick, you said that I was thinking, wait, you're gonna paint yourself blue? Other beast.

Nick Mellem 4:30

Other beast, other beast, yeah. And then uh because we were uh last year was Monsters Inc., so we were all Monsters Inc. uh characters. Nice. How about you, Eric?

Eric Brown 4:42

Uh this year I don't think I am. Last year we we went to uh a party that a friend had, and we just I think we went as skeletons. Why? Because it was easy to go to Amazon, order the thing, zip up, done.

Nick Mellem 4:59

What was the shirt that Eric wore to the ugly Christmas party? Oh, the biker shirt.

Joshua Schmidt 5:05

The biker with the chest hair sticking out. Yeah. It's fitting. How about you, Alex? What what's your all-time best costume for Halloween?

SPEAKER_01 5:16

Uh well, the the one I like more recently is a little more low-key with just the glowing chest piece and a sport coat and you know, Tony Stark. But the uh the favorite of all time was definitely uh Darth Maul. Uh and that was with the the full head with the spikes and the face paint and the robe and everything. Um you will not be seeing pictures. Um but the fun part was we were hosting a Halloween party and we were low on a bunch of stuff. So I went to the grocery store as Darth Maul, and it was awesome.

Joshua Schmidt 5:50

I'm sure everybody loved that. Yeah. Well, my grandmother was a seamstress, so we had custom-made um at growing up, we had custom made costumes pretty much every year from you know, birth till, you know, maybe maybe middle school. But I think my favorite one was uh Teenage Mutant Ninja Turtles. You know, we had the N95s painted green, you know. We had like the turtle shell knee pads, you know, and that was during the 90s when we were just all about teenage mutant ninja turtles. So it was super awesome.

Eric Brown 6:20

But uh is that how you got into all the furries stuff?

Joshua Schmidt 6:23

Yeah, that's that was my entry into my furry career. You can find me on furry onlyfurries.com. Um yeah, no. Um yeah, but I I'm this year, I think I'm gonna be going as bluey. I think we're doing bluey this year. Yeah, but I'm I'm all about the zip-ups now. Uh you know, I got a lot going on. I can be there. We're not doing the uh custom made uh the full regalia. We're going with the zip-ups, keeping it simple. And it inevitably rains on Halloween here about half the time, you know. So it uh more snows to the challenge. More snows. More snow or snow. Yeah. Yeah.

Instinct As A Digital Assistant

Joshua Schmidt 6:59

Well, we're getting into some tech stuff today, guys. Uh uh, one of the things that uh we brought up Alex when we were talking about topics was um this instinct app. I didn't know I know Eric and Nick have gone all in on instinct. And um, I just wanted to get some security uh reflections on that. I wanted to see, you know, kind of bounce that off of Alex and see what he thought about that as well. And and and you know, talk about Siri and some of these other kind of seamless AI integrations that are starting to appear and kind of reflect on those. And uh, and yeah, so so a couple, a couple days ago, I guess a week or two ago, uh, Nick hit me up with the Instinct app uh text message and said, you know, sign up. And I got in there and I got a little skittish. Um, I started thinking with my cybersecurity hat, and I'm like, this is a little too seamless, right? Like um, all of a sudden I'm signed up and I don't know where my login is and I don't know where my preferences are. And Nick straightened me out. But um, but I know Eric, you've been an early adopter and and you've been a big fan. So maybe you could share with us what what you're using it for and why you like it.

Eric Brown 8:00

Yeah, it's um it is a it has taken the role of a digital assistant, and you and you can kind of recreate the same thing um in any of the major AI platforms, right? There's ways to build digital assistants using skills and and hooking up with MCP servers into different applications. And this this is kind of a uh a simpler way to do it where you don't have to really you don't have to do much but sign up for the platform and then you can connect it to the other platforms of your choosing. So I think a couple of weeks ago we were playing around with it and we hooked it up to order lunch for the office. So like, you know, it it was interacting with us, took the lunch order, and then um you can give it your credit card if you want to. Uh you can give it it it stores passwords in a it has its own password manager. So I think it's you know kind of up to individual interpretation of how secure that password manager is. I I haven't found any documentation on how they're securing passwords. So I I wouldn't put anything uber sensitive in it, but like if you want to just have some fun and and play around and you know, maybe throw a credit card on it or a credit card that's got um, you know, it's like uh where you can limit the spend um or a virtual credit card and you know, see see what the art of the possible is, I I think that's cool, but I I would I would be cautious before doing a lot with it yet.

Joshua Schmidt 9:40

Yeah, Nick brought up that, you know, a good point after we were having that conversation. It's like it's kind of our job as cybersecurity professionals to see what there's out there and test some of these things and kind of take them for a spin because if we just stay too secure and we don't try these things out, we won't know how they work or really what their limitations are. So I I know Alex and Nick could probably bat this around a little bit too.

Free Range Agents And Data Risks

Joshua Schmidt 10:04

What were your impressions, Alex?

SPEAKER_01 10:05

Uh it's very similar to other iMessage-based tools that I've seen. Uh, Lindy is a good example of one that I'd used for a while. Um there's there's a lot of platforms out there. I in general, I call this category free range agents. They're out there running somewhere, doing something, but we have zero transparency into what's actually happening behind the scenes. Uh, totally agree with Eric. It's something awesome to experiment with. This is the future. It's terrifying to see people using this for actual work today. That's the same problem with OpenClaw. We're still not there. Um, and in this particular case, the problem is this is a well-funded startup, great valuation, and zero information about what they are doing with the data that goes in there, except some of the things that they've already been caught in. So early on, they had statements they were training on everything. They took that out. And now it's kind of we we only train on the stuff you let us and stuff that it generates. It's really fuzzy. Um, so from a privacy perspective, um, I put this right up there with Muse in terms of, oh hell no.

Nick Mellem 11:15

How about you, Nick? I was gonna ask you about Muse. Yeah, no, I I've been having fun with it. Um, kind of what Josh said, you know, as cybersecurity professionals, you want to get in and we want to test everything. You know, being safe is you know, give it instructions up front. But I really am hooked to it, to how you interact with it, right? We're all everybody's running their life on their phone, right? So if I can set it up, if I can every morning, if I can get with my cup of coffee, I can see what the traffic looked like going to my daughter's school, see what meetings I have coming up, and then get a briefing on the day, right? So like it's those little things that make life easier. Now, this isn't just for instinct to do, right? But it's coming to my text. I'm maybe already in there, fielding some messages, right? It just it seems really cohesive to me. So I think that's what draws me to it. You can do this with other AIs, but it was one of the reasons I like Instinct.

Eric Brown 12:08

I'm gonna see if I can get my instinct to socially engineer Nick to open up something that he shouldn't.

Nick Mellem 12:16

I'm gonna tell my instinct to block Eric.

SPEAKER_01 12:19

Make sure you don't friend him. Uh, but the but the the usage model you're describing, Nick, that's exactly why so many people originally got pulled into OpenClaw. Is that the exact same thing of uh being able to have that iMessage or Telegram in the early days where you're just talking to it, I think we'll see that interaction model appear in Claude, OpenAI, Google, and others. It's uh because I agree with you, that's a great personal model, but I want to know what it's really doing with my information. And um even found this in the using OpenClaw and Beyond. You train it how to do something, but everything that you're training it with is it's kind of sort of learning and it's building memory. It's not building processes. It's not writing processes, it's not writing skills. So it's kind of a next time you do this, please try to do it this way. That's really fuzzy. I don't love that.

Joshua Schmidt 13:13

Yeah, one of the things we've been talking about around the office too is exactly what you're you're touching on there, Alex, like owning the process and not renting it out. I think that was one of uh one of your topics. And instead of yeah, exactly. I want to hear more about that because that's something that really resonated with me. It's like you can pay people to do things, you know, you can find third-party folks to do things. And sometimes that's really valuable, especially like Nick and I are doing that for some tools right now. It's really going to fast track our learning, right? Um, but if we're going to create a process internally and have to own it at some point, why then outsource it? Why not keep track of it, agentify it, um, then improve upon it, right? And and that's kind of what you're you're you were touching on there. Could could you expand on that? I think that's a really cool concept.

SPEAKER_01 13:56

Uh absolutely. And again, the we can chat with these things, the same way we could chat with just ChatGPT as we were getting started, we can go back and forth and we're building a much bigger session memory. So it has some context, it has some idea of what we're trying to do, and we ask it to do it again, it'll go back and look at what we had done before. But what it doesn't have is, hey, every time Alex wants a flight booked, make it look like this. Add it to the calendar in this way. Um if he's looking for this kind of information about a person he's about to meet with, here's the details he cares about. You could kind of do that with uh custom GPTs previously, so or reusable prompts previously. But in this case, since we're training it by just talking to it, there isn't an artifact there that's the rules or the process that you're talking about. Um I call that a capability. If you're going to have a capability, and again, if it's unique to you, if it's hey, help book me a flight, I don't need to own that process. If it's hey, tell me about this person I'm meeting with tomorrow, I'm very particular about what I want to know. That process I want to define and make sure it's being repeated, that's a capability I want to own. So that that's the kind of thing where AI skills specifically, because we can put the prompt-like language and the most important piece and some deterministic, a thin slice of code, but I don't yet know if we can actually load a skill that has Python in it into Instinct so that one plus one always equals two if it needs to. So that we get that deterministic behavior. That to me is the power of skills, is we have a prompt plus deterministic software all in one bundle. I don't know yet if the free range agents can all support that.

Nick Mellem 15:34

Well, we'll have to test it with instinct.

SPEAKER_01 15:36

Are you uploading now?

Nick Mellem 15:39

No, well, we will be. We'll be I think Eric might be hard at work on something. Yeah. He's doing something fishy right now. And I don't know.

Eric Brown 15:46

I'm not doing anything fishy. No. Yeah.

Driverless Cars As Process Lesson

Eric Brown 15:51

The Alex, you're from the Chicago area. Are you in Chicago now? Uh currently Florida. You're in Florida. So in the Chicago area, and and maybe compare contrast to Florida, how are you seeing the adoption of the driverless vehicles? Is that in either area yet?

SPEAKER_01 16:10

Because I I know in Texas it's coming. Oh, in Texas, all over, California, all over. Um in Florida, I haven't seen one. Actually, I think I might have seen one gold cyber taxi.

Nick Mellem 16:23

I thought they were in Miami. I think they're in Miami right now.

SPEAKER_01 16:26

Haven't been in Miami, but I yeah, in areas I've been, I haven't seen them. And I haven't seen them in the Chicago area either. So it's super early, I think, in those areas. We're gonna see them. Yeah. And uh I will say for Florida, we really need them. There's a bunch of folks that shouldn't drive.

Joshua Schmidt 16:40

Minneapolis City Council is currently trying to keep that from happening here, of course. Minnesota.

Eric Brown 16:46

Yeah, yeah. The uh it's interesting it because you talk about process, Alex, right? And and the and the defined things that the AI can do. And and I I think it was in maybe the release of Tesla's, I I don't know what they're calling their the their AI driver. I think they they they rewrote the code, I think it was three years ago now, where it it was just the the code was getting into two it was just like the logic was getting out of control with it of just a bunch of if then, you know, if this then do that. And it was it was not really good. It it was pretty bad self-driverless experience. And then they took thousands and tens of thousands of hours of camera footage from Teslas of real people that were driving, and they scored those people and they said, okay, you know, if somebody's a a four or a five, like they're they're a you know, a prudent driver, let's just say, right? That's how they they they they ranked them. Um, and then they trained the AI on the fours and fives of like what would a good human driver do in these scenarios. I think the the thing that that happened for me that I was like, wow, this, you know, AI is here and it like what it can do is so powerful. Is it's coming up to I had the the you know the the driverless stuff going and it came up to a traffic circle, it stopped, and this is a traffic circle with four entry points to it, it merged in, it went around, it got off on you know where it was supposed to go, and it was it was flawless. And you know, I've done that now hundreds of times since. But it and to me, just kind of going back to you know what you're talking about, but uh on the process side, they they've somehow it's teaching itself through repeated actions um to to improve. And I I I wonder where that crossover will be with the agents that we bring into our lives as they're observing what it is that we do, what is our routine, how do we do these things? You know, Josh's routine is probably different than than Nick's, and then can they learn from that and get better over time in a custom way? And I I don't think we're there yet. I think it's all just kind of mass learning and then applied through prompting to the the individual level. Um I uh they're they're not rewriting themselves, I don't think, yet at the individual level.

SPEAKER_01 19:35

No, it's really been more the uh the the memory files. Um, and I'll using Claud as an example, kind of that Claude MD or agent MD, just here's my core instructions and daily memories and short-term memories, but it's it's memory-based, which again is still best effort. It's not codifying that into the rules, exactly what you were talking about. That's that's where these agents are gonna go. And that's why in a year year or two, it'll totally be normal. But it'll be okay, here are here's Alex's best practices expectations. Here's how you work with Alex, here's how you work with Nick. We'll totally see that. Um, but I I think these agents are still way early. Um again, OpenClaw changed the world, and we're you know, 10 months in. Um it the next step of that's gonna be exactly what you're talking about.

Joshua Schmidt 20:21

How

Apple Siri And On Device Privacy

Joshua Schmidt 20:22

do you see the future of AI kind of coming into like Claude and maybe Apple stepping finally into the space a little more solidly with some of those seamless, interfaceless AI uh you know interactions?

SPEAKER_01 20:36

Uh I one of the things I think we're gonna see with Claude, and actually, I mean ChatGBT is already there, the the voice interface, the interaction there, that to me is actually even more natural than the iMessage texting interface. So having that voice interface where you've got the full powered brain behind it is just amazing to be able to do that again, having it have access to your skills, your background, your memories, those kind of things. I think that has been really amazing. Um, the one thing I'm really excited about with Apple and Siri is um they have rightfully recognized the LLM brain is a commodity, right? It's not the thing that matters. They've focused on specialized, much smaller language model types of things to do specialized tasks, and whether that's voice to text as an example, whether that's picture enhancement. Um, and then they partnered with um actually most of the major providers to do other interesting things. Um one of the things that I like about Apple's focus is it's actually security first, unlike pretty much every one of these companies we're talking about. And that that I find that really interesting that Apple, the consumer company, is suddenly now the privacy champion that even in the enterprise we need to be paying attention to. Running this stuff on device is hugely interesting from a security perspective, and it's got access to my data in a way that I don't want to give that data to any other app. So that's really interesting. Then being able to fall back into the private cloud compute. That's where that customized Gemini model is running up in Apple's data centers. So if it's something small enough that Siri can answer on the phone, it does. If it needs a bigger brain, it can poke out into that cloud. And part of what I, again, I really respect that they've done that is a verifiable security process that they're following, that none of that data ever resides on the hard drives. Some of the challenges we've heard from other LLM providers logging things, they're not. That's really impressive. So Siri is having that privacy piece, huge.

Joshua Schmidt 22:35

Yeah, you guys are a little bit more bullish on Siri, right? Because it, I mean, I Eric and I agree. Like uh we've talked about this before, like how we're not impressed by the talk to text on like the Apple Watch, for example. I'm find myself editing more than if I would have just, you know, typed it out on my phone.

SPEAKER_01 22:50

Oh, it's 27 or 26?

Joshua Schmidt 22:52

This one's ancient and I have no idea.

SPEAKER_01 22:54

This is like one of the 2027 changes everything.

Nick Mellem 22:58

I still haven't found the dictation on 27 to be like it is on Gemini or like on a Claude, right? Or using Whisperflow. Because I still use Whisperflow on my iPhone. Because I yeah, I found that I've been having a great experience with the new Siri, Alex. It's been fantastic. I've been asking it questions just like you were saying, but I've still found the dictation to still trip itself.

SPEAKER_01 23:21

Interesting. And on the on the phone, it's well, on the phone itself, I used to use Whisperflow for some of the similar kind of things. On 27, I'm actually purely using just the microphone for dictation. I've been pr very, very happy with it. I like that Whisperflow does some formatting, but I'm also finding it actually removes some of my words when I didn't want it to. And even when you set it to the don't change any process, don't process anything, it still does some of that. So I'm liking that if I want an actual raw transcript, the Apple part does that better, I think.

Nick Mellem 23:53

Yeah, and I I think I saw that the one thing downside to Siri right now is she doesn't creating memory from what I'm what I'm seeing. There's no option to save memory or update memory from what you've been teaching it. So I saw a lot of people or have been reporting to you know, turn it to never delete your conversation. So when that comes, you will have all these conversations that you've vlogged to learn from. So I guess that if there is a downside, that would be one.

SPEAKER_01 24:20

I I would agree. And I think that's an opportunity for Apple in the same way that you know our data is on the device. I can see, and again, Apple is never first to market, but they usually try to do it better than anybody else. I would expect in the next year, the same way now Siri has your chat sessions, that didn't exist before. Exactly that next thing of, okay, what's the memories? How do we manage that and maintain the facts that are outside of our apps that we want to have available? That's going to be interesting. Because that personal many memory and corporate memory, that's kind of that next nascent, hey, this has got to get figured out. I would agree with you. It's not there yet. I think that's an opportunity we'll see from them.

Nick Mellem 25:00

I'm going to say I saw the there's been rumors for like the past maybe year of AirPods with little cameras on them. So you can be walking around and be like, hey, Siri, what color is this Yeti water bottle or whatever it is? You know, what's that car that just drove?

SPEAKER_01 25:15

Exactly. As a sensor, not a recording device. That's going to be the difference. And did you just trigger Betty in your office there?

Nick Mellem 25:22

We read the Siri to pipe down.

SPEAKER_01 25:25

Actually, I don't have it listen. I always press the button. Yep.

Nick Mellem 25:28

Yeah. I've got a home too, a home pod back there that's listening.

Eric Brown 25:32

The record in 27, it's always recording what you're saying on the phone. And then can you go back and say on the watch?

SPEAKER_01 25:41

It's specifically that's a part of the watch.

Eric Brown 25:42

Just the watch. Yep. Can you do that? That's 15 seconds, but do you have is there something where you could go back like an hour ago? Because this would be great around the house. Um, where I mean I I hear from time to time, it's like, well, I already told you that. It's like, well, no, you didn't. Now I knew that's weird on the thing. Oh, you're looking for proof now, huh? Yeah. I mean proof. Maybe that too.

SPEAKER_01 26:03

Maybe that that's the feature that still hasn't been released yet, is the summarize my day. And I'm very interested what it what it includes. Is it a, hey, you were on a podcast today with these folks and you had lunch with Jim? And I don't yet know what level of the summary it is.

Joshua Schmidt 26:18

Yeah, and Alex, you c every time I talk, I get excited about having this Apple ecosystem containing my data and kind of migrating a lot of these tools and processes over to an Apple device or or just using them within Apple. Like how far are we away from being able to build these processes and kind of own them and live in them with the Apple device? I mean, we've kind of started to get there, what you were just mentioning with Siri improvements there. But uh, how far are we away?

SPEAKER_01 26:45

Uh well, that's a really interesting parallel because Siri is never going to be what Instinct is or what Crockbot or what Muse is. Apple's really approaching this 90 degrees from the market in terms of, hey, this is my interface to my information. So some of the capabilities that we're talking about would likely be Siri in front of apps that are doing some of that work. So Siri is the coordinator. That's where we essentially get app intents on the iPhone, where an app can expose its data and its functionality to Siri. So I fully expect we will see personal assistant type apps come to life that have some of those capabilities and things that then expose them to Siri, and we can do things in that way. I would lean more into some of those kind of things. Uh like Eric, I don't have Alexa, and I I think that that's one of the biggest security holes around because them, Muse, and other organizations, those companies make their business model on our data. That's not okay. We just have to recognize that it's getting siphoned off. Um the reason I think the Apple ecosystem, number one, having been inside it, I can see yes, there's a lot more security here in terms of the way they're setting it up. Um but the uh the potential future of building out our agent and capabilities, uh, it's not there today because it's different, but I think we're going to see the same way. Again, OpenClay Claw came out of nowhere. We'll see something on the Apple platform. All right.

Joshua Schmidt 28:17

Yeah,

Capability Equity And Owning Workflows

Joshua Schmidt 28:18

I I can't wait because uh yeah, I've been just a little anxious about that and and also wanting some of that functionality. But we've been building a lot of processes here and um kind of want to switch gears into your book. Um, we've been building a lot of processes here, you know, on different, you know, I'm using like five different AI agents or five different AI platforms, right? And then, you know, doing pictures in Chat GPT and then bringing them into a Claude project to do, you know, a co-work or to do, you know, to do some kind of Claude code to create a white white paper. And then maybe I'm researching something on Google or in perplexity. So um as we kind of like build these processes up, you know, why would we want to then give them to someone else and rent that that intellectual property out there, right? So maybe you could talk a little bit about like how do we get from having all these disparate sources of information and processes and like bringing them together into one spot so they can make them way more useful and and way more easy to use, right?

SPEAKER_01 29:23

Well, so it it really starts if we look at um and what I map out is really kind of six phases of how we get to these capabilities. And where we are today, pre-agents and pre-AI, is if you wanted to use software, it's big software. It's HubSpot, it's Salesforce, it's you know, we're we're renting big SaaS systems. And what that means is number one, our data's there. And number two, we have to wrap our business around those systems. So what that did is as a huge symptom, lots of spreadsheets came into being to hold the data between system A and system B. So the way we actually work, we had to create our own tools and trackers and dashboards and stuff that those systems don't support. Yes, in the old days it was the right thing to do because it was really expensive to build that stuff. Fast forward to this year, now that we have these AI skills that we can put together and we can move around between systems, we can actually rewind to instead of having had wrapped our businesses around somebody else's software, in those skills, you can do exactly what you were talking about, which is here's how I want to work. Here's how we as a company work and do things and build the skills and use chat as the first interface for nearly free. That's the amazing part. So instead of spending a million bucks building something or hundreds of thousands of dollars, the expert in the process can put together and build a runnable set of skills, runnable IP, for hey, here's how we create a proposal, here's how we create graphics for marketing pieces, and we've got those skills, and then they can call each other and we can get value. And then eventually we can actually make micro apps and other things with that. But to your question on the data, those skills don't imply that we dump the big systems as a first step. We don't have to. Those skills should be the front to all of those systems. So as a user, I no longer care where the data is. And then as we roll forward, use the skills, build micro apps. That's my word for what's the 20% of stuff that's I'm doing for my whole day that is painful to do in chat. Looking something up in chat two, three times, awesome. 50 times, nah, all right, let's build a little micro web app for that. But the combination of those two things, that's now where we start to live. And the interesting thing is then we can go look at those backend systems and see does anybody ever log into them anymore? So we now have an opportunity to start sunsetting some of those things. Maybe we're standing up our own database for the data that we actually wanted, which was probably only five or 10% of what the big system held. So there's a migration path here that's not a lift and shift. We live the new process. Somebody says, hey, we really need to track these two pieces of data about a customer, and you didn't have to go reconfigure Salesforce. That that's I think the huge change. So we're gonna be able to own the user experience front-end piece and the processes. Uh we'll start to own more of our data. And you know what? If we're still using a back-end system because it does stuff, great. Who cares? Lease the commodity. Don't build everything, lease the commodity. But if there's something unique to you, how you want to handle the data or otherwise, like the images you were talking about, we have skills that we've created. So in Claude, we can actually call all of the other LLM providers. So I'm using Claude code, and it's actually going through and creating images over an open AI for me. And then it quality checks them to determine, hey, is this actually a good image or is this a cruddy image? And hey, go open AI fix this. So we get to use multiple platforms together, and I think that's hugely powerful.

Eric Brown 33:01

So Alex, you know, you're talking uh about like the the back-end piece example of Salesforce, right? Um, and and Salesforce over the years has gotten really bloated and and quite expensive for organizations. And I I kind of postulate and wonder where's the future value of a Salesforce where you know you can go out and do a lot of this on your own. Why am I paying $300,000, $400,000, $500,000 a year for something that I could probably hire a dev team agency to do, have it done in two months and have spent a quarter or a fifth of that?

SPEAKER_01 33:42

And that's the that's the question. If I'm only using five percent of Salesforce, exactly, what's the 5% I need? But that's my recommendation. You start with building the skills first, live in the work. So you you essentially rather than writing a spec, which you maybe get 80% right if you're lucky, and then having to do a lot of expensive software maintenance, living the new processes skills, then we start to build some of those pieces. The SaaS companies are gonna have to transform their value prop. Their value prop is no longer with the user interface. Maybe they can be a great data store. Great. What other services are you gonna provide us around that data? Are you gonna add, do some value add around lead scoring or research or those kind of things? If if all you are is the user interface to my data, you're dead. Yeah. Those companies won't exist in two years. There's nothing there. There has to be some process or some value add there. So I agree with you. It's gonna change.

Eric Brown 34:36

So, Nick, it sounds like uh shorting, shorting Salesforce is the way to go.

Joshua Schmidt 34:41

Short that's like ask your agent to do that on Robinhood for you in there, Nick.

Nick Mellem 34:46

Um, I'll do that.

Joshua Schmidt 34:48

Yeah. So is that what you're getting into in the book, Alex? I I think that's kind of is that the general thought thrust, or what else can we take away? Because I I think we're all waiting for a copy. Eric's got a book coming out, so we can do a little book club. We can do a little small, but uh very cool.

SPEAKER_01 35:03

Well, and it it is the key of rather than having to build the big stuff that we had to do before, we can now use essentially we're gonna use AI as the software. We're gonna chat with it, we're gonna build the skills. And that right there, we don't ever have to log into a Salesforce or HubSpot again. And then we can make it easier for ourselves and do the next thing and the next thing. And that's the way we then own what's unique to us. If we're doing something that, yes, every other company does this the same way, let's not custom build that. But realistically, every organization has dozens and dozens of things that make them different, that they're better at, or that's just unique to them. That's the stuff that we build as those skills. That's the white glove treatment working with a customer. How do we do this thing? So again, leasing the ordinary, but owning the stuff that makes us who we are, that's the difference. And we can now do that this year because of these skill things.

Eric Brown 35:58

Alex, you and I have been in coach, I don't know, maybe three years, something like that. Um, we're talking about strategic coach, Dan Sullivan, strategic coach. And I I think I was it was one of my first 10x sessions that I went to. And I don't remember if you were you and I were in the same 10x group at that point in time. Uh, but somebody asked the question um, how many people in this room have written a book? And it's boom, everybody's hands are shooting on me. And I'm looking around, I'm like, shit, everybody in here's got a book. Oh, and that was about two years ago. I was like, I have no idea what I'm gonna write about. Well, yeah, I I started the process about six months ago. Um, and I know you've been working on yours. You've kind of been ruminating on this for a while, right? Like 20 years.

SPEAKER_01 36:48

Well, actually, so it has evolved over time, but in terms of the decision to write the book, so that was actually a capability I wanted to own was to be able to go from idea to book in people's hands in two months. Oh. Because publishers can't do that. We can't traditionally do that. So I built a whole bunch of skills and capabilities to help me do that. That's awesome.

Joshua Schmidt 37:08

So you're living, you're practicing what you preach, in other words.

SPEAKER_01 37:12

That's the first chapter of the book.

Joshua Schmidt 37:14

Well, how I wrote the book. How do we get a copy uh of your book, Alex?

SPEAKER_01 37:19

Uh it is uh Capability Equity. It's it's up on Amazon now, uh, so it's available for Tyndall and uh on paperback as well. And yes, you guys will be getting copies.

Joshua Schmidt 37:28

All right. And and yeah, we'll definitely get you uh on the waiting list for Eric's book too. He's nice. Uh we're all we're all waiting for that as well.

Eric Brown 37:35

Um still waiting on the title too, Josh, because you bothered me every day.

Nick Mellem 37:39

I wasn't gonna throw you under the bus. You said it, not me. He called Josh out, not me. That was a win for the game. I love it. That was a win.

SPEAKER_01 37:47

So you haven't asked Instinct to help you with the title yet?

Eric Brown 37:51

I haven't. I've asked Claude and Astra, but I have not asked Instinct.

Why Public AI Skills Are Risky

Joshua Schmidt 37:58

Speaking of skills, uh Eric and Nick had sent me over uh skills.sh. Do you know that repository for for AI skills, Alex? Or do you have one that you like to go to for or do you just build all your own and don't mess around online? No, there's security issues there as well, probably.

SPEAKER_01 38:15

There absolutely are. I will look at external skills for inspiration. I will never ever run them.

Nick Mellem 38:21

Yeah.

SPEAKER_01 38:22

So no, I don't I don't have a favorite um skill there because I think in general, this is again an example. If somebody else wrote it, it's not necessarily mine. And if it's my business process or something, again, for inspiration, sure, I'd love to know. Oh, cool, what did they do to prep for your day and some of those pieces? Absolutely. And then in 10 minutes, I have my own version that I know is safe.

Nick Mellem 38:45

I had found that skills page, Josh, that I sent to you guys the other night by I was talking to Claude after we had a Builders Guild meeting yesterday session with Alex and AIY. We were building skills, and I was talking to Claude late last night. I was like, you know, what are some of the most popular skills you see people building? And we were going back and forth, and it said, Oh, you should check out this website, and that's how that came about.

SPEAKER_01 39:07

Well, there was an interesting uh study, and I actually talk about it in the book, of um security researchers who actually hit probably the top five repositories, I think it was. They went through thousands of skills and plugins. Um, 30% of them had security vulnerabilities, and 80 of them had malicious code in them. Not cool. So not cool. So, no, we cannot trust, you can't trust anything on the internet. You really can't trust runnable skills on the internet.

Joshua Schmidt 39:33

Even if you're using a Mac, yeah.

SPEAKER_01 39:35

Even if you're using a Mac, it's something that's gonna because it'll have access to your cloud, co-work, cloud, chat GPT, whatever it is. So yes, that that's a that's just a big no-no.

Nick Mellem 39:44

Great for inspiration.

Joshua Schmidt 39:46

Right. Yeah. So in some ways, this is more nefarious than than just like uh phishing email, right? Because you're you're potentially just inviting it right into the middle of your ecosystem. It really doesn't have to break down any walls. It's it's just being let in the front door, so to speak.

SPEAKER_01 40:01

Absolutely. And again, that is the exact that's the exact same concern I have with Instinct Muse or the next one. It it's the same attack vector right there. It's got access to everything.

Always Listening Wearables And Meta

Nick Mellem 40:12

Yeah, everybody's been talking about Muse because uh Meta had their uh event, was it two weeks ago now? They also released new glasses to compete with the Vision Pro that are coming out, I think, later this year, beginning of next. Uh they did look cool. Um they're so much lighter and smaller than the Vision Pro. So I think Apple will be coming out with a much lighter weight version. But they also talked about the Muse pin that is gonna be coming or like a keychain. Did you see that, Alex?

SPEAKER_01 40:40

I did not. Is that the one that I've heard about that uh is always listening to you?

Nick Mellem 40:44

Yeah, it's gonna be like a little pin or a keychain that you can wear and it's got the a screen on it that's gonna have their little Muse character, you know, the little Yeah. That that's the scary.

SPEAKER_01 40:55

Well, and and and again, because of the company behind it, we all know Facebook's business model. We can't just know. Personally, experimenting with it, sure. But it's anything you ever ask it, you know it's ending up in those databases. I'm not cool with that. Uh, but back to the the pins or the other wearables, that's what I think I'm really interested to see Apple's take on the always recording on the watch and how they scrub the information as opposed to every other wearable, where the entire conversation is going to end up in a database that's not yours.

Nick Mellem 41:28

It'll be interesting to see if Apple comes out with like a Fitbit competitor without a screen, and that will be an implementation of a listening that will you'll use your phone as an interface then.

SPEAKER_01 41:40

You can use the AirPods for that.

Nick Mellem 41:41

You can.

SPEAKER_01 41:43

I think the AirPods will do step counts. Yeah.

Joshua Schmidt 41:47

Wow. It's a spooky season, y'all. We got a lot of uh cybersecurity uh implications here to think about with all these devices, but it's always fun to explore latest and greatest tech with Alex here. And uh, you know, Nick and I have been students of yours for a while now, but uh yeah, it's been fun to sit in on these uh AIY classes. And I didn't get to join the Builders Guild yesterday, but I know Nick had a great time. And um, yeah, we're looking forward to doing more of that in the future. So for folks out there that don't know, um check out Alex's new book, Capability Equity, and you can find that on Amazon. Also, you you can check out AIY, right? Um you can find that and Alex Bratton at LexTech on LinkedIn and get connected to the community. It's a really awesome community. So um I know we're kind of running short on time, but do we have time for an Eric Eric Reacts video?

Eric Reacts To Cliffside Runway

Joshua Schmidt 42:36

Does Eric want an Eric's React video?

Nick Mellem 42:39

I think he does. He does.

Eric Brown 42:41

I think he does. Gives people what they want. I never know what's coming, Alex. I never know what's coming.

SPEAKER_01 42:47

So that's fun. We gotta do that then. We gotta have fun.

Joshua Schmidt 42:50

It's really not that crazy. Don't don't get don't get worried on me.

Eric Brown 42:55

No big deal. Uh Alex, while they're pulling it up, um favorite pizza Chicago. What are you Giordano's?

SPEAKER_01 43:03

Uh Giordano's is a top five, but actually it's Uncle Pete's. Uh it's a local, regional. It's again thin crust, lots of cheese. So similar to Giordano's, but it's there's a couple of them around. They're awesome.

Nick Mellem 43:16

Alex, weren't you saying yesterday in the Builders Guild, do you have an app on? Is it burger rating or pizza ratings?

SPEAKER_01 43:21

Uh burger. Yeah, I'm a plain burger guy. Uh burgeralex.com is burgerology, the science of the study of burgers. Yes.

Nick Mellem 43:31

I'm gonna check it out. That's so cool.

SPEAKER_01 43:33

Did you build an app for it or or the website for it? Uh it's actually it's hosted on Tumblr. Um, it's been up there for 10, 15 years, and I just occasionally throw the reviews out there. You'll find stuff all over the country, all over the world, actually. I think there's some from France, Italy, and a lot of warnings of places not to order burgers at. Oh. Like McDonald's. Do you do you do fish places?

Eric Brown 43:57

All sorts of like Nick's vegan. So do you do these new vegan burgers they have?

SPEAKER_01 44:04

No, no. My my focus is it's a plain burger with ketchup in the bun. None of the cheese, none of the bacon egg avocado weirdness. Just, you know, I just plain bur plain burger. Oh, that's just wrong.

Joshua Schmidt 44:14

Yeah, plain burger, man. A little update for the longtime listeners of the podcast. I did get to go up into the sky with Eric last week. It was a great time. We I uh got some videos coming out um, yeah, with Eric and I talking, talking biz up in the sky. It was a really great time. Um I came down just, you know, feeling giddy and feeling I did the adrenaline rush and um feeling like, oh, now after I did that, everything else is gonna be a little boring today. Um in a good way though, like, you know, uh kind of inspirational. So thanks again, Eric. It was super fun. But uh ever since then I've been watching more of these videos and uh Eric Eric's instilled in me a sense of peace about some of these common flying things. But have you heard of Luka Airport, Eric? No. Okay, we're gonna we're gonna we're gonna check it out today. If you don't know Luca Airport, I think it's in Nepal. So this is one of the most common airports that they fly into to get to base camp for the Everest hikers, right? It's a little sketchy, and there's a ton of videos online. Um, but I I wanted to get your your opinion on this one, Eric. Okay. Because uh this looks just a little crazy to me, right? Okay. Can you guys see this? Yep, see it. Okay, so we're we're taking off right into right off the edge of a cliff here. And uh and then you're probably what? You're probably already at altitude when you take off. So isn't the air a little thinner up there? Don't you have to get going faster in order to get off the ground?

Eric Brown 45:39

Or lighter, yeah.

Joshua Schmidt 45:41

Or lighter. Is that why we're flying smaller airplanes here?

Eric Brown 45:46

Yeah well, the the airplanes will have a maximum um runway length that they have to use, right? Like so uh jets and things like that probably couldn't get in there. Right. Okay, there you go.

Joshua Schmidt 46:00

So would this make you nervous as a airline as a pilot? You know, you kind of cut it looks like you're flying straight off into a mountain, which probably have to get altitude pretty quickly, and the wind's probably whipping around in there, huh?

Eric Brown 46:11

Yeah, well, I mean, you can see him like so he get what's what he's doing here is he's turning around, he's getting the maximum use of the runway here by turning around right on the numbers, and then you can see him, he's holding the brakes here and he's doing a run-up. You see the propellers are um are spinning up there, and as he moves forward, you're gonna see him move the tail rudder there. So he's moving the tail rudder a little bit. Looks like there might be a slight crosswind, maybe, not really. And uh now he's just you know he did a uh max power takeoff. Um, so where the where the engines were already spooled up, so to speak, and he could get, you know, or she a little bit of um better performance out of it. But I I wouldn't be apprehensive um about that takeoff. There's a a similar um runway just outside of Albuquerque. It's probably my favorite one. Um it's just north of Albuquerque, and it's in between, it's in a valley, it's up on a plateau like this one. Um, I can't remember the name off it of it off the top of my head, but there's only one way in and one way out similar to this, where it doesn't matter where the wind's coming from. You have to, you can only land and take off one way. Um, and it's a really cool airport because you're kind of up on a plateau like this one, but on on the sides is a valley. And we were flying in there about was it last summer, not this current summer, but last summer we went down and um Kelly and I flew in, landed in Albuquerque, fueled up. Um, it was pretty windy, and it was so windy that we couldn't actually land at that airport because the it was like maybe a 18-not crosswind. So you'd have winds blowing across the runway. It was kind of like a quartering crosswind, and it just wouldn't wouldn't have been safe to uh to force it in. So Kelly didn't get to hit it uh that time, but hopefully the next time we're we're down there we can do it. She's probably thankful. Well, it it was one of those where it it was we were getting down there later in the day. So what happens is the sun heats the ground, and then you've got the air rising at different rates, like if it's air rising over blacktop or over water or you know, over over um, you know, dirt fields, what have you, like the air is rising at different rates, and you have these pockets of air and this turbulence. So it was it was definitely a shake and bake moment, even getting lined up. And um, as we're getting lined up, it was kind of like that last, the the last video that we saw last week or whatever, where you really had to come in at a at a crab, but like we were maybe dropping 10 feet and then back up and dropping, and it's like, man, I cannot get this thing lined up. Kelly, I don't think we can make it, right? I don't think we can get it in there safely. Um so I I tried, but it's just I had to abort, go around, and just said like heck with it. So what do you do? Hit the gas and find another airport? Yeah, we we just went back to Albuquerque, um, which was like maybe 10 minutes south uh of there. Um yeah. But sometimes, I mean, you just gotta make the right decision, you know.

Joshua Schmidt 49:24

Agreed. And in cybersecurity and in aviation and adopting new AI tech. Yeah.

Final Takeaways And Subscribe

Joshua Schmidt 49:31

Well, thanks again, Alex, for joining us today. It's been great to have you on. Uh we've always loved you as a guest. Yeah, let's let's do it again sometime in the near future. And um, if like I said, uh, you can find the book. Thanks again for watching. You can like, share, and subscribe and source us wherever you get your podcast. We have video up now on Apple and Spotify. You've been listening to the audit presented by IT Audit Labs. I'm your co-host and producer, Joshua Schmidt. We have the usual suspects, Eric Brown and Nick Mellum, and today our guest was Alex Bratton of LexTech and AIY. We'll see you in the next one.

Eric Brown 50:01

You have been listening to the audit presented by IT Audit Labs. We are experts at assessing risk and compliance while providing administrative and technical controls to improve our clients' data security. Our threat assessments find the soft spots before the bad guys do, identifying likelihood and impact, or all our security control assessments rank the level of maturity relative to the size of your organization. Thanks to our devoted listeners and followers, as well as our producer, Joshua J. Schmidt, and our audio video editor, Cameron Hill. You can stay up to date on the latest cybersecurity topics by giving us a like and a follow on our socials, and subscribing to this podcast on Apple, Spotify, or wherever you source your security content.