Signal&Noise Live at AI Con: Lucas Longacre Talks with Ken Johnston, Founder of AiGovOps Foundation

Watch on Youtube

Listen on Spotify

What happens when companies move so fast to adopt AI that they forget everything they already learned about building technology safely?

In this special edition of Signal & Noise Live at AI Con, Signal & Noise Executive Voice contributor Lucas Longacre, Head of Product at Inlightened, sits down with Ken Johnston, co-founder of the AI Governance Operations Foundation (AI GovOps), for a candid conversation about what it really takes to deploy and scale AI inside an organization.

Ken argues that the rush to embrace AI has created a massive case of enterprise FOMO. Companies feel enormous pressure to demonstrate that they're "doing AI," but in the process, many are abandoning fundamentals that took decades of software engineering to establish: observability, testing, CI/CD, security, rollback capabilities, cost controls, and disciplined product development.

Lucas and Ken dig into the rise of what Ken calls "demo theater" — where an impressive AI prototype can be created in hours and appear 90% finished, even though it may represent less than 10% of the work required to turn it into a secure, scalable production system.

They also explore:

• Why enterprises need to bring DevOps, DevSecOps and FinOps discipline into AI

• How AI can dramatically increase the "blast radius" of software failures

• Why observability may be one of the most important — and overlooked — components of enterprise AI

• The difference between an impressive AI demo and a production-ready product

• Why companies should build AI projects around learning loops, not just deployment

• How natural-language interfaces could finally replace dashboards with direct answers to business questions

• Why AI-generated code is creating entirely new challenges for software development and code review

• The opportunities — and dangers — created by vibe coding and the democratization of software development

• How Lucas is using AI inside product development while controlling security, token usage and access to enterprise data

• Why human experience, judgment and taste remain so important when working with increasingly capable AI systems

It's a fascinating, funny and highly practical conversation about moving fast with AI — without forgetting everything we learned before AI arrived.

Read the full transcript below.

---

Lucas Longacre (00:00.878)

Okay, so hi I'm Lucas with Signal and Noise. Can you introduce yourself please? Hi.

Ken Johnston (00:05.614)

I'm Ken Johnston. I'm co-founder of the AI Governance Operations Foundation, what we call AI GovOps.

Lucas Longacre (00:12.14)

AI GovOps, that's a mouthful. one of the things that I'm came here to research for myself more than anything was this idea of scaling AI properly. Because right now I work at a company, I'm the head of product for you know a 16-person team, we like three engineers. So we are coming up with innovations and you know experimentations constantly that I get to either do with my internal teams or like, you know, a little more vigorously put into the product that takes a little longer, but at the same time, we're owned by a

Parent company that is, you know, thousands of employees that you know they're on workday, is a lot some of their stuff they use for invoicing and for so I've noticed that the rollout for them, of course, is going to be take much longer, and I see I feel like there's less of a strategy behind just they throw these tools at people who have very little experience with them and they're just kind of seeing what sticks. So I was just curious from you know an enterprise background, you know, what are you seeing?

see that is like what's working with companies that are rolling out you know ai in their you know to update their their stack.

Ken Johnston (01:19.5)

Yeah, so well that's that's a great open ended question. We'll we'll break it down and take a couple of pieces out of it at a time. But you know, one of the things that I've been noticing when I talk to companies

Is this big rush because everybody hears about, you know, AI and they think that they're falling behind. So you've got this FOMO activity happening across companies, the fear of missing out on the big AI, and it's like we've got to get some of the AI out there at all costs. And so the challenge I'm seeing from my perspective, and and I'm actually intentionally taking a different point of view around AI right now, and that I'm actually focusing more on the AI film.

failures and the risks because I like to look at patterns that are causing issues and so instead of maybe telling you here's the pattern of success I'll give you a couple of anti-patterns.

One of the big anti-patterns we're seeing is what I call a failure to adopt lean AI, or basically DevOps and AI. Because everybody is now treating AI like one of two things. Like it's low risk, get it out there. And so you're seeing manual deployments, you're seeing lack of investments and instrumentation and observability and

so even when companies get AI out there, they don't have the controls in place that you would normally have. Nobody would stand up a new mobile app or stand up a cloud service without having basic observability or rollback and some basic tests in their CI C D pipeline. Yeah. But when they're doing AI, they're like, it's a model, it's special, we're gonna deploy it manually and we're not gonna have a test environment or a dev environment. We're just gonna put it in production and nothing's gonna go wrong. Yeah. And of course, everything is happening.

Ken Johnston (03:12.048)

happening and going wrong. I mean one example that I like to point at, point at is Grok.

Everybody remembers a few months back Grok got in trouble because they changed the system prompt and now suddenly people could do things with images that you're not legally allowed to do. And we're not going to share the details on that right now. But the problem they had with that is that they didn't have the ability to do control feature suppression. They couldn't automatically roll it back. And so they had to leave that in production for a long period of time. And so that's one of the things we're talking about is task radius and AI. Go ahead.

Lucas Longacre (03:45.058)

The thing

I'll just say this wild, just to the I I w I never saw from that perspective. I knew that they were having issues, but I never thought that yeah, like why don't they just roll it back? That's basics.

Ken Johnston (03:55.512)

I know.

And they're they're out of they're XAI, they're part of X. And they didn't have an automated ability to roll back a mistake like that. Okay. And they left exposed. And so AI, because people are cutting corners, is increasing the blast radius. We talk about that in cloud development all the time. But people are forgetting the fundamentals when it comes to AI and they're leaving them exposed. And the thing, if you don't mind me, I'll drift a little bit onto my soapbox for a moment. But what people don't realize is this. Because we've been cutting corners

AI, the number of lawsuits and the number of settlements have been spiking. And now people don't realize this, but the NAIC has the AI

framework that they're pushing now, so that's the National Insurance Organization. And they're separating AI liability from the rest of corporate liability. And so now when you're cutting those kinds of corners, when you're creating and the reason they're having to do it is like I just said the blast radius is so big. So people are afraid of missing out. Yeah. They're cutting corners because everybody says you've got to get AI going, but they're creating these risks in their system and

The bill is coming due because the insurance industry is tired of paying for sloppiness. And so those are the anti-patterns I'm seeing. Now, of course, that leads you to the exact perfect pattern, which is lean AI fundamentals. Do not forget your basics. When you do, when you treat AI development like you do any other development, when you've got product market fit, when you know the business impact you're trying to do.

Ken Johnston (05:27.756)

The other mistake or anti-pattern I see is people doing demo theater. Somebody vibe codes something cool and everybody's like, ooh, that's AI. Let's launch it now. But you know, that vibe coded demo is looks 90% done, but it's less than 10% of the real work. And is it actually aligned to the business impact you wanted, or was it just a cool demo someday and you needed to get some AI out to make the board happy? Which one are you doing?

Lucas Longacre (05:52.3)

Yeah, it's so f it' interesting. So I created a a really quick demo for the sales team to get something out the door just so they could do sales against it. And then of course it ha I had to then build the damn thing. Yes. And which was in a way a fun challenge. But the interesting part was sales caused the company to be like, Well, what's the timeline on this? I'm like, Well, it's gonna take a long time to develop. Like that that that demo I had up at Rodding in like a day is amazing you can do that. But actually make it work is like back to the basics of software development about like

Iterating on it and like literally I had to fine-tune this whole program thing over and over again to get results that are usable and actually workable. But I but I think you're you're touching on something that's really interesting that I've kind of been seeing about AI to begin with is like eventually it's just gonna be infrastructure that we build on top of, right? And it's not gonna be sexy or interesting at all. It'll just be another tool we use in our toolkit for software development now.

Ken Johnston (06:48.052)

Absolutely 100% agree there, but we're treating it as a bolt-on, we're not including it in our normal practices. You know, we've all, if you're in this space, a lot of us have been reading that, you know, the volume of code is increased so much that the number of actual humans reading the PRs is plummeting. Yeah. So you can't keep up with the volume of code. So now you've got to start getting well-trained agents to help you just slog through all the code that you're now accountable for. It's amazing how AI is changing the fundamental.

Of software development in ways we should have seen. I can't believe I didn't anticipate that code volume would kick my rear end in terms of doing code reviews, but it makes total sense.

Lucas Longacre (07:29.408)

It does, yeah. No, I never didn't even think of it from that perspective too. Granted, I think in some ways wanting to do make stuff lean is, you know, the goal. Like, you know, sure the demos like this rickety thing you throw together, and then like how do you simplify and make it, you know, usable. what one of the things I've noticed that is actually kind of hilarious is what you're describing is people who are vibe coding, which I think that d democratization is fascinating because you have people with who total different perspectives and not maybe the tech background that are creating things we maybe never would.

Have thought of, but the flip side being they don't have any education or discipline on how to build things safely. And so, for example, a friend of mine, I saw he was building this kind of like a an app where he can buy and sell stuff and trade stuff. And like that for me would be a red flag instantly. I'm like, wait, is there a monetary transaction happening? Uh-huh. And I've I think I added something benign on purpose not to like spook him, but I was So, what are you using for like your database? And he's like, I don't have to worry about that stuff. And I was like, my god. I was like, Are you sharing? Is it all in the browser? But like just those fundamental.

Ken Johnston (08:25.833)

Lucas Longacre (08:29.252)

my gosh, yes. I that's the thing is like we're by allowing people to then build whatever they want, there is a huge danger that I think gets opened up. Granted, aga the opportunity for people with perspectives and ideas that never would be coding or or building products, I think is fascinating. I just wish there was some kind of like education on it or protocol. You s maybe that's the thing I should build is like here's your AI safety you know bot that'll just go in there and just like audit your thing to to

Ken Johnston (08:58.594)

Well, and that's great. And you know, really back to your thing about in the enterprise and where you're at with the larger company, companies that are doing this right. So I like to make this comment. It's like you wouldn't change change your IDE for your developers without figuring out how to implement your tools in the new IDE and providing training. So why are you throwing out claud code, co pilot, and saying developer fin for yourself? Why don't you have standard system prompts for all of your developers with all of your rules, how security

is supposed to work, but no. It's like everybody's got Claud code, go for it. Yeah. And so you're letting it, you're, you're not treating it like a change in your development environment. And it's very possible to do many of those things you just described. Yeah. Put them into the prompt library, put them into the CI C D so that what comes out is consistent with your development standards. And we can do a lot of those checks too, but you know we're just like, get them the tools because we've got to get my God.

Lucas Longacre (09:54.966)

This is like reinvention of the wheel though. So funny, you're like, there's a reason why we have DevOps, there's a reason why this stuff was built to begin with. So why are we removing them? And that in some ways they're even more important because people with less experience are, you know, modeling around in it. I don't want to be too negative on it, but I do think the pressure, and I I don't necessarily blame like the heads of companies or CEOs, whatever, who were from what I heard across the board and met in almost every industry, was like these mandates to use AI, the fear of missing out. Yeah. But at the same time, I do think that was like

Anybody who pumped the brakes and said, like, okay, before I deploy any these things, I'm just gonna wait and see or like see where the business opportunity is. And I definitely leaned into that. Like I my I came from the startup world, so we were encouraged to break things, right? yeah, yeah. And so but when I actually took on a full-time position, I was like, even though we were told like you really need to start implementing AI and have like put it into the product and figure it out, I was like, I'm in no rush. Like I don't understand the job I took on well enough yet to start doing that. And it took me about a

year and a half at the company while everybody else is building things with it just observing and then finally I got to a point where I saw like a a a glaring opportunity. And even then it's like I I put it through the same process of of feet of shipping a feature. And you know it took way longer than I would have hoped. But it's still like I feel like there's a reason it's safe and secure and also doesn't burn a lot of tokens. Like I made sure it was very token conscious.

Ken Johnston (11:22.638)

All of the basic bivecoding that people are doing and the tokens that get burned. Do you see the article in Fast Company? One company got their anthropic bill at the end of the hundred million dollars, their entire year's budget in a single month because they didn't manage token usage. So, you know, we've seen FinOps drop on the floor, you know, that came in and saved us because we're all overspending on cloud, you know.

Lucas Longacre (11:33.51)

but it was like

Ken Johnston (11:47.822)

We see DevOps and DevSecOps slipping as we say, AI is different. I've just got to get it out there. And I'm like, we're developers. Yeah. We work with CIOs and CISOs and it's like, why are you letting AI around the edges?

Lucas Longacre (12:02.83)

Real, yeah. So so I'll say though the good the so one of the things I was made sure to implement was controlling how the different teams I interact with it use AI in our product but also in our workflow. So I audited a lot of their their work and I helped them develop tools for them, but I didn't just say go use AI. I actually like watched how they worked and built you know, with the help of AI built like automated systems that then you know that was burning tokens to build that for sure. Like it definitely was a token cost. But the final

use of it would make sure it was only used where necessary and then that was like written scripts and written you know you know python scripts or code that we make that would then run without burning tokens so it's like I you know I think there's a way to do it smartly and I feel like we're all gonna get there eventually and I think this is the hard this is the hard cost was like those people that dove in head first might have broken their necks on the

Ken Johnston (12:56.31)

You might have. I the pole. You you know, there there was some people had to go first. Some and the patterns that you know we're seeing, would I have called those net anti patterns out a year ago? They weren't visible to me. That's true, yeah. And so, you know, it's hard to say. I c I do this th I I have this phrase I talk about with a lot of folks, but one of the things to remember with AI or in my opinion, is focus on the learning loop.

AI is so new. When you do your AI project, I do want you to have it aligned to business objectives, but what is your learning objective with it? Because we need to learn what we're doing with AI. And one of the things that bugs me the most when people ship AI, the corner that I least like them cutting is they're cutting observability.

How are you going to learn from your AI project if you're not monitoring it, if you haven't implemented drift detection, if you aren't actually logging the transactions and the things that it's doing to understand what it's doing? You put it out there, and so cutting observability, like I'm actually okay if you cut automated deployment because you needed to get a project and try it. And maybe it was low risk. But to put it out there without any observability and instrumentation and not learn from it is unacceptable.

Lucas Longacre (14:09.634)

Yeah, you're right. You that's gold though. The logs that you have is like, you know, anyway, that that's a great point. And w it's funny, we are we have all the logs of the stuff that we're doing in our system, partly because one of the things I built was a a natural language query that would essentially allow all of our staff to be able to talk to the data, just be like, I I want this data from across departments, because I was running into the problem where I'd build dashboards for each individual department and they never worked right. Like they would always be like, Well, what about this scenario?

I'm like, then I would literally be doing data pulls constantly, like pulling data from the you know from the database, the SQL code. And so I just did a translator that takes what they asked for, turns it into SQL code, put pulls it. So there's essentially a firewall between you know the data and the large language model. Where so it's just translator, it only has the schema to go off of. And by doing that, essentially I just I said have at it, like pull whatever take it'll join whatever tables you want, and then you can like make sense of the data you need and interpret it. And but getting that.

That to work right took a lot of you know on our finessing on our end to like get that to the translator to work properly with our data. very secure, and it's you think about it if it's just writing SQL code, that's so little token usage. yeah. and and then, but then you know creating a table that the you the user can then use. But I'll say even that was like, yeah, we're we were logging every query somebody wrote in anytime something didn't work right, anytime something didn't, you know, didn't return exactly what they wanted, we could go back and see the thought process of

Ken Johnston (15:43.838)

Become fuel to help train the model to be in the future. And so, yeah, absolutely. One of the I hate to burst your bubble a little here, but I used to run a data science team at Microsoft, and I had this rule, and I called it Priya's Law, because that was the person that ex that first experienced or crystallized it for me. But

Usually when an executive asks you for a dashboard, they don't want a dashboard, which is like you said, yeah, dashboards that people didn't really use. They want an answer to a question. And that's the mistake a lot of us make. And honestly, even with bybecoding, I see more dashboards showing up. And that's the wrong answer. Investing in getting your data layer correct so that people can ask natural language questions and get answers. Because they don't want a dashboard, they want an answer to a question.

But yes. Yeah, so you were getting that and that's why you said I'm doing this layer. So you experience dashboard.

Lucas Longacre (16:44.386)

Literally it was the theater. It would no because I what I I kept recognizing that I would give them what they wanted and I'm like, God, it still doesn't work for them. So then I would give them, I'm like, wait, this is what they think they want, but this is what they really want. Didn't matter. And but yeah, the the natural language thing is what do you really want? What are you really looking for? What is yeah, what is in the data that you want to grab?

Ken Johnston (16:56.45)

Didn't matter. It just leads to another question.

Ken Johnston (17:05.622)

That's the thing is when you when you get a question answered, you think of another question. So dashboards only answer questions that you thought of in advance. Yes. That is why they will always ultimately fail.

Lucas Longacre (17:19.182)

Which is why I'm excited that we're moving away from that. I mean, one of the to me, one of the huge advantages of AI is the natural language, is it the humanization of this technology where we can interact with like with it like we would like another human being. And to me, that is so valuable for so many companies. Like especially because I think, you know, I come from originally the film and television world, so I didn't come from tech originally. And so what I had to do was manage teams of people on set or in a documentary fashion, you know, like I was constantly

working with departments as well as a flow of a project and I'll say that technology is almost no different right it's just that the language is different the actual technology is different but the the being able to communicate clearly and know what you want and have goals and aim for them and also recognize that it's never going to be where where you start and where you end up two very different things. The storyboard you draw the script you write and what you end up with the end software development is like no different

Ken Johnston (18:14.99)

Yeah, it absolutely is. I should share one story with you, and I I'm glad that you talked about the storyboard, because I might be a technologist, but I'm also a a frustrated novelist. One day I'll I'll have an amazing novel out. But I very much believe in the importance of stories. So one time, and I did this literally so that I could have this story to tell that I'm about. I'm in a meeting.

Lucas Longacre (18:35.19)

Sure.

Ken Johnston (18:38.094)

And I'm not really paying attention, but I have a note taker and I'm gonna wait for it to summarize at the end. And I'm bored and they mention my name, so I answer a question, and then I go back to not paying attention. And then I decided to completely stop paying attention and I went to my note taker and I took the transcript of the notes. And I think we were just well, we were discussing a potential product.

idea that we had. So I gave it to one of my AIs and I said, What are they talking about and what are the requirements? And this goes, this is the product and these are the requirements. So then I said, can you turn that into a PRD for me, a whole product spec? And said, okay, here's your full 20 page PRD. So then I took that and I went over to Lovable and I said, Can you make me a prototype with this PRD? And it makes a whole website. And then I jumped back into the meeting at the end and I said, Well I got bored and I made this prototype. What do you guys think? And everybody's like, my gosh, can we ship it? Can we ship it? I'm like, no

But that's how easy it is to vibe code and to get into prototype theater. Yeah. Anybody can do it and all you do is take the transcripts of the note taker and feed it through and you got a flashy something. And I did it just so I could tell that story. We did not ship that thing.

Lucas Longacre (19:49.132)

No, but that's I mean but it again, there's like a cool opportunity there that how quickly you can create, but then there's the flashing warning danger signs that should be going off when absolutely see how quickly those things can you know come to life. I don't know, that's I that's how I feel about the situation we're in currently. I'm like AI has come along so far so fast, like it's why I'm almost like scared to to talk too much smack about where we are currently 'cause I'm like there was things that I was complaining about a year ago that have been fixed or better. I know. Like and it's like I d

Did not see that like each time I'm shocked with the speed of of growth, whatever, but it's like I still am not convinced that this is all just gonna end up being like some really cool infrastructure that the next thing that is really way more interesting and better is gonna be built on top of is the my gut feeling in many ways.

Ken Johnston (20:36.354)

Definitely it's evolving incredibly fast. I had a weird experience last night. I was trying to shot down

Lucas Longacre (20:44.43)

Take up to

Ken Johnston (20:48.28)

But yeah, last night I was I was trying to edit some slides, and so I took the deck and I gave it to to Claude and I said, Hey, I think I should add a slide or two here. Can you analyze it and tell me what you think? And then I went into thinking mode. And doesn't it drive you crazy when your AI starts thinking and it takes so long to get back to you? And I'm like, so I decided to go to bed. I went to bed, I came back the next morning, and then I discovered what it was doing. It decided to rewrite the entire PowerPoint deck for me. no. And I'm like

What did you do that for? I just wanted two or three bullets and some ideas for some slides, but it created a whole deck for me. It's like I've got to start remembering to not tell it to over function. Like I've asked questions and I get a whole Word doc or I get a whole PRD when I'm like, help me research this, and it thinks I want a Google Doc or something.

Lucas Longacre (21:33.63)

Yeah, I hate it when it goes on meth for a bit and just goes like crazy. It's like you're like, stop it, put it down, put the tools down. But again like from a novel writing thing, I will say I'm about a hundred and twenty pages into probably about a two hundred

Plus page novel. This will be my first finished one. And I use Claude as my writing assistant, and I'll say it's a fabulous writing assistant. Because for me I would constantly run into my biggest roadblocks I realized was writing was what what's the name of this character or what's the name this place? Or or you know, I feel like I don't like I'm not being authentic with this scene because I can't visualize it or like what what's the time period, what's this? Like so I use it as my researcher and my writing assistant, and it

Allows me to plow through what would typically be like a stop in my writing and and then I put it down. And the funny thing is I've tested it with actually writing scenes. Like I've I've loaded in like the first I think I had like 65 pages. I loaded in and was like, let's experiment, see about it, writing a chapter. So I gave it the outline and I said write it in the voice to match this. It was awful. really? Awful. Well, because it's very good at re doing what you've done previously, but it's not if there's something still like it's it's a mediocre machine, you know what I mean? Like it never

Ken Johnston (22:40.67)

really?

Lucas Longacre (22:50.734)

does anything that's too like with humans we have inspiration of oddness and weirdness that I'm sure they might be able to get there someday but like I I I'm don't think anybody should be paranoid about AI novelists taking

Ken Johnston (23:02.638)

I think it's helpful for coming up with creative ideas for character names. Super And usually I don't use the exact one it did, but it's like, here's a first name and a last name and okay, give me three more variations on that.

Lucas Longacre (23:13.206)

It's also taste. So I'll say give me five names every time it's g it' if the one it recommends is terrible. But it gave like like you said, or there's a one that if I mess around with that that one and honestly this is how I work with writing partners. So when I have a writing partner, which I love, is I'll work with them, we'll like bounce ideas off of each other. So I throw something out, they correct it or come back with a different idea, and through all that process you kinda get something greater than where you started. And I feel like that's the value of it if and I'd recommend if you're if you're

Like I'm gonna do finish this novel. Yeah, using it as a a researcher, as a you know, just a partner and a road a roadblock stompers like

Ken Johnston (23:50.532)

no, I totally agree. So like the book I wrote that's coming out from Pearson that's gonna be called the Lean AI handbook. Hope you don't mind me plugging my book. It's supposed to come out in September. And originally Pearson didn't want me to use AI to help in any of it because, you know, they're an actual publisher. Good, yeah. And you that's it. The the rules on it have changed since I started six months ago. But I insisted on one thing with them. I built a rag model to help me do the research. And I called the rag model and I gave it a little bit of a personality and a character. And

Lucas Longacre (23:57.87)

Please do.

Ken Johnston (24:20.162)

Then and I also told it that because I was writing a tech book and I'm old, I'm like, well, be a Gen Z voice for me. So it started to be a Gen Z voice and it would say these weird things. So I actually started, so I I'm not calling it a co-author, but I did give it, I gave it some some poll quotes in the book. So the AI generated these pull quotes and I gave it credit in the book as as like, hey, this is what a Gen Z person might say about what we're doing in lean AI.

Lucas Longacre (24:48.718)

But think that is a cool use. So again, like that's a collaborator with you. Play the keys over. It's not there yet. Like that's why I think any nobody should worry about it being like, the AI novelists are gonna take over the publishing industry. It's like, no, the human taste and the and you using that is clever in its own way. That I like again, it's the collaboration between man and machine, I think it's

Ken Johnston (24:52.684)

It was totally a

Ken Johnston (25:10.574)

And that's why it's so fun. And back to earlier we were saying the learning, it's also in our own uses of AI. Sometimes like that little thing I did to try to see if I could vibe code from some meeting notes. Yeah. That's a learning test. I have to do things if I'm gonna be in the AI space to try to stay ahead of people. And the amount of weird stuff that I do just to experiment is is vital to keeping me where I'm at because people ask me questions all the time. And if I just did a traditional way of learning, I probably start to fall I'm or let's put it this way.

I am falling behind because AI is moving so fast, but I'd fall further behind if I wasn't willing to go try stuff.

Lucas Longacre (25:46.91)

So one of the things that I've noticed in a lot of the discussion I've been listening to podcasts and just you know watching things is that those of us who are kind of higher up in our careers and have a lot more real life experience of building things and failure and all that, when we were given these tools can create like wide as well as tall, right? So like you have a lot of expertise in many different ways. Whereas if you're just starting out, it's almost becomes a hindrance because you didn't have to like earn a lot of it. And I feel a lot of

Listen, I don't want to judge young people starting out. I feel terrible. Like I actually felt very lucky to be from analog to digital, like I was on that. You know, so like for a young person now growing up, granted, I feel like they're gonna bring kind of inspiration and ideas we never because we're too polluted with our background. So like they're seeing these tools in the hands of people who have no restrictions with it is gonna be awesome. But I do think like unless you really understand how to build things properly and it you and you have that kind of

like rugged expertise for years and years of that, it's like it these tools can be even more difficult to navigate.

Ken Johnston (26:53.582)

Yeah, that's a fascinating point of view. I haven't thought enough about it. I actually also run a community for our AI governance group. And next month we're doing a a our topic is gonna be AI governance and education because there's some research already being done by researchers about what is the impact on AI in terms of youth learning. Yeah, okay. It's like how are they gonna develop because we are we're using AI often

of this this decades of experience.

Lucas Longacre (27:26.988)

Also like how many books have I read in my life that when I go and ask a question to like g give me this, I can just like cut corners, but I also have like done a l so much of that.

Ken Johnston (27:36.088)

On your ability to detect a hallucination? Yeah. yeah. my gosh. It's like, wait a second, that's not real. Or even in like guiding, because sometimes AI will get stuck and you're like, no, the topic is broader than that. Trust me. I've read books on this. Yes. I need you to go in this direction with your research. And you nudge it and it's like, wow, otherwise I would have been stuck because it focused in on a little narrow niche.

Lucas Longacre (27:58.892)

Yeah, I mean honestly that's the this stuff's too new to for any of us to know, but I think having those community group discussions, have bringing in experts in it, is essential. anyway, I would definitely love what is the name of your book coming out? The Lean. I'm gonna definitely keep my eye out for that. And how can people find you if they wanna like find

Ken Johnston (28:12.366)

Green AI handbug.

Ken Johnston (28:18.014)

There you go. And so you can find me that way. All right. Bet. Thank you so much.

Lucas Longacre (28:26.328)

Thanks. This has been an amazing conversation. I appreciate all the talk. All right, thanks.