Episode 4: The Verification Tax with Hunter Hillegas

Brian Milner (00:21) Welcome in everyone. We are back for another episode of the People Over Prompts Podcast. This is the place where we're talking about the future of teamwork in this age of AI. I'm your host here almost always, Brian Milner. And today I've got a good friend of mine here, Mr. Hunter Hilligus, that's with us. Hunter, thanks for coming on. Hunter (00:46) Hey Brian, thanks for having me. I I like the intro music there. Brian Milner (00:50) Thanks, thanks, yeah. you know, gotta gotta keep my rockin' roots, you know, evident there in some way, shape, or form. yeah, so we're we're we're here today to kind of continue this dialogue and continue this discussion about how AI is really changing, you know, how we work really and and how teams kind of work together. And I wanted to have Hunter on because there's this particular kind of Hunter (00:55) I love it. I love it. Brian Milner (01:19) Annoyance. I I don't know if you call it annoyance or a tax or whatever that I think that we we all are kind of encountering when we use AI to help with coding and and other tasks like that, which I'm just calling the verification tax here. basically, you know, we we've all been there where you you've you've been working through something and something spits out code or text or whatever it is that you're trying to get from an AI and it just is it it's not that it's completely wrong, it's that it's slightly wrong. And it's that verification of having to to pick through every last bit of it and and make those small tweaks that starts to feel a little bit like attacks. I saw this stat, Hunter, that came from a 2025 Stack Overflow kind of survey that they did. it said eighty-four percent of developers Hunter (02:14) Uh-huh. Brian Milner (02:18) are saying that as of today they use AI tools in some way, shape, or form. I mean that's huge. Eighty-four percent are using AI tools in some way, shape, or form. But the same survey said that only half of them say that they trust the output. So I just I I'll I'll just start with that. Like what's your take on that? Do you identify with that? Do you do you think that's has that been your experience or is that that kind of more the setup? Hunter (02:24) Yeah. Think I I think I filled out that survey. so I can't remember what I wrote down, but I do remember it. yeah, you know, it's it's interesting. And it also I should probably preface everything that I'm gonna say on this episode with we're recording this halfway through 2026, and you know, this whole this stuff, this AI, the AI industry and the tools that they're providing. Especially for people making software, it's changing faster than anything I can remember in my entire career. So, you know, whatever we say may turn out to be a little bit different down the road. We'll see. So I'll preface it with that. I do think you know, especially I I really started in earnest with coding AI coding tools kind of right after Chat GPT came out and the original three point five open AI. API was released and it was like day one. Wow, this is amazing. This is I don't even know how this is possible. This is just a total mind-blowing situation. And then day two, it's like, look at that code. That's not so good. I mean, it kind of works. I the the button is purple and in the left corner, but yikes. so you know, it's of course that's real. I mean, I guess I would say that that's true with Brian Milner (04:00) Yeah. Hunter (04:07) you know, lots of software, right? If you don't have tests and we talk about t automated tests and testing and a lot of the work that we that we do. So I know that's familiar. We can talk more about that stuff. But it, you know, of you need to be able to verify that the work you're producing is correct. And with the AI tools, you know, that you you need some rigor. And we can talk about specifically what that looks like and the kind of experiences that I've had. I'll also say that as the tools have gotten better. my level of trust has gone up over time, not down. So, you know, I guess not maybe not maybe predictable, but definitely that's been my experience. Brian Milner (04:40) Hm. Well, I so I I wanna press in on that for a second 'cause I I agree. I I think my mine is the same way. Like I my level of trust has has grown. However, I I'm curious what your take would be. How much of that do you think you'd attribute to just the models getting better at what they do, and how much of that do you attribute to your knowledge that's grown about how to really interact successfully with it, how to give it the right context? Hunter (05:11) Yeah. Yeah, that's a good question. of course the models have gotten better and also not just the models. I mean, especially, you know, usually I'm thinking in in terms of the big front US frontier lab models like OpenAI, anthropic, you know, maybe to a lesser degree Gemini from Google. and less so some of the open source models and other stuff that I've played with, but don't use as much on a day to day basis. They've definitely gotten better. especially in terms of OpenAI and Anthropic, with Codex and Cloud Code respectively, they've also built these harness tools, which are a really important part of the improvement situation with with the outputs. That the the harnesses really know how to get the most out of the models and those things really do go hand in hand. and so I think that's been a pretty important part of the leap that we've seen in HNTIC software engineering, especially in the last six months or so through the very end of 2025 up to now. But you ask a good question because yes, I think, you know, myself and others that are using these tools often, we have learned quite a bit about where they're good, where they're bad, how to prompt them. I and maybe you and I even had this conversation. I can't remember. But you know, back in the in 2023, 24, as the stuff was really exploding, the idea that we start to see job listings for like prompt engineering. And I think, you know, there were a lot of people in software that kind of chuckled at that. Like, that's not a real job. What are you talking about? and I get that, you know, it's good for a laugh. But the one thing that I've learned probably over everything else in working with these tools is that like what maintaining the right amount of context, the right level of context, context at the right Brian Milner (06:44) Yeah. Hunter (07:05) level of specificity is really key and maybe the most important thing. So I think it's a good question. It's my I guess my to long way around to to answer your original question. I think it's probably a combination of both, right? Of course the tools are getting better. They seem smarter. and you know that's a that's a benefit. But also we're getting better at using them. Brian Milner (07:27) Yeah. th that verification process of it, I mean I I I don't I I'll I'll tell, you know, the story to everyone who's who's listening. But like when I first started to do that kind of coding and I was going back and forth between the different you know, I I had subscriptions to a couple of services and I wanted to make you you know, 'cause you'd run out of tokens on one and you just hey I I still got some over here, so I'll use those tokens. And it was kind of that going back and forth and it was you know, very quickly apparent to me that like there's there's a better way of doing this. Like I'm sitting here between these agents. Hunter (07:49) Yeah. Brian Milner (08:04) going back and forth and you know basically what we would call an orchestrator now just kind of taking the the output and then going back to it and saying, we'll do this now next. and you know now there's enough of of the harnesses on top of that that you can have, you know, AI orchestrators that can pass things back and forth. And the the thing I always struggle with is the when you build into that workflow testing You know, there's there's the question of what's the practical amount that's needed. I think there's there's the verification's important, you know, the whole the whole thing that we've borrowed in the AI industry from the the whole Reagan years of trust but verify. I I think that's that's very, you know, ingrained in what we're trying to do, but how much verify? You know, like do I just have the core testing? Does it work? Does it come out okay? I run through all my regression tests and it comes out okay. Do I then have also a security engineer kind of tester that has to check for flaws and and holes? Do I do I have I don't know. There may be a legal kind of testing to make sure I'm in compliance with various laws. And that's what I struggle with sometimes is how much testing is enough testing, how much verification do you really need? Hunter (09:19) Yeah. Yeah, I mean I think, you know, like so many of these things, it depends, right? There's a sliding scale depending on the risk that you take on by not testing, right? So if you're writing software to create legal briefs and the judge is going to hold you in contempt if you cite make made up cases because you didn't test your software enough, you know, that's probably a higher bar than, you know, you built an app to manage your kids' basketball schedule or something, right? It's like, well, I didn't show up to practice on time. That is bad, but it's not as bad as, you know Brian Milner (09:53) Right. Hunter (09:58) G getting fined by a judge. So there's a sliding scale. Brian Milner (10:00) Right, right. Yeah. Yeah, a and and and that that brings up a good point too, 'cause like i i it it makes know part of this podcast is kind of trying to re examine at a much higher level whether we're even appro approaching these questions from the right standpoint and trust is a good example of that. Like, you know, saying trust but verify What is that look what is trust with AI? Like sh is that are we even using the wrong metaphor there, the wrong word to say trust with an AI system? Hunter (10:32) Yeah. Yeah, that's I mean, this is we're getting into the philosophical here, I guess, somewhat, but it is a it is a good question, right? And I know that there are some people that get very offended when you say things like, you know, the model is smarter or dumber, because they're not humans, right? But we like to anthropomorphize these things. It's just part of what we do. but they aren't humans, right? So it's like they don't have intent in the way that a s a software engineer on your team has intent and all these different motivations that Brian Milner (10:38) Right. Hunter (11:01) you know, contribute into the stuff they do on a day to day basis. So trust I I is that the right word? I that's a good question. I don't know if I have the answer to that one, but it is I do think all of the vocabulary that we use around those things. Well, we basically just take in a team member and apply all the same kinds of, you know, nouns and verbs that we would to their work product that we do these models. And that may not always be appropriate. Brian Milner (11:24) Yeah. Well what what's been your experience 'cause I know you've worked you know, in your career, you've worked with human developers, you've worked with AI developers, you work with combinations of these two things. So I'm I'm kind of curious w how that has shifted the workflow process for you. what what's different about working with some combination of AI agents or AI agents with smaller teams of of people? Hunter (11:53) Yeah. I mean, not to sound like too much of a shill for big AI, but I I'm getting more work done than I ever have in my entire career in terms of output. I don't mean like line to code output, though that might be true. I don't know, I don't measure that, but in terms of features shipped, users delighted, you know, the the kinds of metrics that I actually do care about. and that's 'Cause I am able to use these tools to, you know, leverage what they're good at, which is not everything. And there's all kinds of different tips and tricks that we can talk about if you want to, to getting some better results. And again, they what they are varies quite a bit over time. But I I have found especially well, it also does vary by the type of project in terms of how how effective they can be. They're they are better at certain things than others, and then there are certain models that are better at certain things than others as well. different focuses that the training regimes have put into place for some of these different models. So there are some specific details there. But generally speaking, I have found them to be a tremendous prof boost in terms of my my output, especially in the last six months, since I guess November of twenty five, a little bit longer than that, I guess then. But When we saw was it now I'm losing track of my version numbers. I think it was like Opus four point five and GPC five point three, maybe or five point two. Anyway, which were both which, you know, it's kind of funny too how these big labs so closely track in terms of their big releases and and their capabilities, right? It's like they both came out within weeks of each other, both were a big step forward. and you know, they've continued since then. But Brian Milner (13:23) Yeah. Hunter (13:48) It's it's been really a transformative thing for me. I enjoy it. I and maybe there might be some sort of it might be growing some bad habits too. I have definitely found myself because I'm these days I'm using codec more than cloud code. I do use both. and I think what I'm gonna say applies to both. but I have found that I we're going out to dinner or something, and I'll be like, wait, talk to my wife for like, I just need five more minutes. It's because I have to get my prompt in. before I leave so it's working while I'm out doing something else, right? Is that healthy? Question mark. Also, both of these tools, you can access them on your phone, right? I I'm and so it's like, well, you know, maybe I'll just shoot off a quick little prompt to like get this thing working so that it's done when I get back. Which, again, you know, debatable if that's an entirely positive thing. I mean, I'm there is more work being done, but there's also something to be said for just, you know, touching grass and, you know, trying to put all this other stuff out of your mind. So Brian Milner (14:17) Yeah. Yeah. Yeah. Yeah. Hunter (14:46) You know, it's a double edged sword, I guess, in some ways, but I am having a ton of fun. I can say that. Brian Milner (14:50) Yeah, I I I agree and you know, like I I've gotten caught up in that same thing before where it's like you know, I I'm I'm sure there's gonna be people people listen to this and go, yeah, I identify with this. But like I I I've I've had those five hour token refresh windows where like you're pushing towards the edge of it and you're like, I just I I want it to kick o over one more time and then I'll send something in big and then I can go to bed. You know, like that kind of thing. so yeah, I've been right there with you as far as that's concerned. I think that the it it kind of has this weird kind of new pressure. You're not doing it all the time, but you have these little windows that you you need to be there to keep it rolling, you know. Hunter (15:28) Yeah. Yeah. And I've also found that the thing that I've been doing more recently too is sort of building loops on top of loops. So basically like having threads in my agents that manage other threads and so that they can kind of take action on the little decision making questions that I don't have to supervise them quite as much. which is also, you know, we're getting very meta here. I was like next time I have loops on tops of loops on the Brian Milner (16:00) Yeah. Sure. Hunter (16:06) loops and it's like how far down does it go? But it is it is pretty interesting and amazing and I it will the thing that I continue to be impressed by trying new stuff that I just would never have guessed would work and it does. And I'm you know that's still kind of as someone that's been doing all this software stuff for a long time it kind of blows my mind. You know, my first computer had 64k of RAM. That's K kids for those not for those of you that Brian Milner (16:07) Yeah. Ha ha ha. Hunter (16:35) You know, we're born after the year two thousand. and now, you know, we're being able to just talk to a computer and it can go off and do all this amazing stuff. And of course it does dumb things sometimes, but it's pretty amazing. Brian Milner (16:37) Yeah. Sure. Yeah, I mean a and and I think getting, you know, circling back to that verification kind of gap, I I I'll I'll give kind of example that you I mean, as I said, I I I always struggle with how much is enough. And I don't know if you've had this scenario happen to you, but I know there've been times when I've had I've designed some kind of testing agent that's testing things and the the way the system's set up, it can reject something and send it back to the other developer and the developer has to fix whatever it is and it you know there could be several rounds of back and forth with this tester. And I I know I've look I've been following along the dialogue to see what's going on and there have been times when, you know, the tester will kick it back for a reason that I'll look at as a human and say, gosh, that's really picky. You know, like it y technically yes, it's a fail, but I mean, come on, we could live with that, you know? Hunter (17:35) Yeah. Right. Yeah. Brian Milner (17:41) And that Hunter (17:41) Yep. Brian Milner (17:42) that's kind of the line that I wonder how do we how do we solve that problem? How do we solve for, you know, the a yeah, yeah, the the having that little bit of judgment to be able to say, that's really not that big of a deal. That's okay. Hunter (17:48) Human taste. Yeah, it's a good question too. I I mean I I part of it I think is we just need some better models, bigger models. But also I I am constantly you know t for most of these agentic tools, there's some kind of file or sets of files that it can read to get rules and information about your project, whether it's like agents.md or claude.md. I'm constantly updating those with The basically like I actually have a codex task that's like go through everything that we worked on this week and make suggestions to update these files based on the things I told you to try to have it get smarter, which does help sometimes and sometimes it doesn't. And you know, there's yesterday I think it was, I was like, I have a reference page and I need to re I need to like port this over to another system. It should look identical. And you can look at both, you know, like make them look the same. And it's like, great, I'm done. And I look at it, I'm like, Brian Milner (18:28) Yeah, smart. Hunter (18:47) This looks the same to you, like there's buttons over here and this one's like cra clipped and I'm like, What's happening here? So they definitely can make mistakes on both sides of the of the I got it finished, go check it out and also you know, the very the minutiae that a human tester would be like, We gotta ship this, this tiny little thing doesn't matter. Brian Milner (18:53) Yeah. Well, I think we're all kind of at that frustrated point a little bit that you you you see how much it can do and it's the limitations that you're like, man, it really can't you know, it can't go that a little extra I mean the whole it we're we're on that frontier moment of of kinda like the you know, Carpathy and the second brain and l you know, him saying I'm a loop designer now and not, you know, like you were talking about. and and I think it's it's sort of at that point where people who are doing this Hunter (19:27) Yeah, yeah. Brian Milner (19:35) Our brushing up against that rough edge and saying, can I almost hack it? You know, like I know it doesn't have general intelligence. I know it it can't really remember. I know it really can't reason and think. But it are there ways that I can trick it or approximate it to to do enough of that that it, you know, really helps and solves, you know, maybe that kind of problem that you know, it it that kind of thing, you know, that kind of error is okay. You can let that through. I had one the other day where it it this isn't this was not a tester thing, but it was the the agent was expecting me to approve something and it it the f it it was looking for an exact phrase with a period at the end. Hunter (20:18) Mm-hmm. Brian Milner (20:19) And I I put the phrase in but without the period. And then it kicked back to me and said, Well, I can't s I can't go forward 'cause you didn't put the period on the end. And it's that kind of thing that you just pull your hair out and you think, I just lost what, ten thousand tokens or five thousand tokens from that exchange of verifying it twice and everything else and you know, it's that kind of thing that we're like, why why can't we just give it that l little edge of intelligence to to know what's what's Hunter (20:47) Yeah. Brian Milner (20:49) You know, practical. Hunter (20:51) And it's it is also funny to me that in however you know, however many years into this being like generally available tools, we still have to say stuff like, Try really hard. Don't don't don't give up. Like it's like, Okay, buddy, like good work. You know, it's like it seems kind of silly that that's a thing, but it does sometimes help. Brian Milner (21:02) Right. Right. Yeah. I s I actually I don't have the stat here on me, but I saw a an article this has been a while actually that I saw this, but it was basically talking about how much it cost open AI for people to say thank you. And Hunter (21:22) Hmm. Yes, I think I read that someplace too, which is like amazing, right? Which makes sense. Yeah. Brian Milner (21:29) Yeah, I mean it it's a it's tokens and you know the and we're we're trained to be polite as humans and it it's reflecting kind of a human personality and so we feel like I don't wanna hurt their feelings, I wanna tell they did a good job, you know? Hunter (21:39) Right. Right. And you know, I'm like I'm like one percent Skynet might happen, but like if Skynet does happen, I'm wanted to be noticed like one of the nice people. Brian Milner (21:46) What? Right. Right. I wasn't one of the one that said, you know, F off to you every day when you didn't do what I Right. Yeah, no, I I I completely get that. i that's a whole nother episode because it like I I think we could get into like the personality of, you know, do you treat your AI you know, like a I don't know, almost like, you know, the an indentured servitude, you know, to you Hunter (21:53) Yeah, exactly. Exactly. Ha ha ha. I know people that do that, yeah. Yeah. Brian Milner (22:17) yeah, I mean it's not human, so it's you're not really gonna hurt anyone's feelings and b you know if it I don't know. yeah, I I mean it I think I think it's an interesting time and you know, I d I don't have any solutions to to this verification tax. I I think it's I mean, my only solution is kind of the thing you said, which is periodically almost like, you know Hunter (22:22) Right. Brian Milner (22:42) a retrospective, have moments where you can look back and say, All right, well what did we learn from this? You know, like is there anything that I could do differently? Is there anything you could do differently so that we can build out into the skills? Is there anything we can take away from this so that next time we don't go through this back and forth of, hey, there wasn't a period there. You know? Hunter (23:00) Yeah, totally. I think that trying to 'cause there's all kinds of little knowledge bits in between in these conversations that we're having outside of the product they generate the code or whatever they're making for you. There is all this little knowledge that sometimes gets sucked up and applied, but sometimes doesn't unless you have a process to go do that. And it's funny, I didn't really think of it as being sort of akin to a a retrospective, but it kinda is, right? So it's like applying these Brian Milner (23:25) Yeah. Hunter (23:27) These techniques and tools that have been used on, you know, human software teams for you know, why not apply as many of those to these as you can, right? You might need to tweak some, but it makes sense. Brian Milner (23:37) Yeah, and it's entirely a different animal because, you know, you're not having to deal with safety and trust issues and that kind of stuff, right? I mean it's it's just all data. but but yeah, I think that the concept of, you know, periodically looking backwards and saying, Hey, what do we learn from this last little bit and how can we do it better? I think still applies and and can help in this these situations and give you that little hint of almost intelligence, you know, in there. Hunter (24:04) Yeah, and this sort of opportunity for self improvement. Again, self. I know people will get mad at me for using self. This is not a thing. It's not a person. It's a thing. But still, you understand what I mean. Brian Milner (24:10) Yeah. Yeah. Yeah, no, I completely understand. so i if if someone came to you and said, Hey, I'm doing this stuff and you know, in my experience the the verification part of this is it just takes too long and it's it's taken all my tokens, like I I'd rather spend that on building new things, wha what would you w where would you be inclined to look first to try to solve that problem for them and help them to to get the amount of verification that they need, but not waste waste cycles on that kind of stuff. Hunter (24:51) Yeah, the the tokens are not infinite part of this probably changes my answer somewhat. because that's a it's a constraint and it's a real constraint and you know, who knows what's gonna happen in the future as far as what tokens cost and whatnot. I think, you know, we we're living in an era of at least partial token subsidy and we'll see what happens over time. Brian Milner (24:59) Yeah. Hunter (25:16) Well, I mean, I guess there's probably a bunch of different ways you could attack this, right? So first of course is to just sort of ask what level of testing is needed for what you're trying to do, right? So don't over test just because you can. I mean, if you have the tokens for it and you want to, then I guess that's fine. But given that you do have constraints, just make sure that you're trying to attack this at the appropriate level. I mean I have been the benefit of many years of doing software, so written many, many, many, many, many, many, many tests by hand. Brian Milner (25:26) Yeah. Mm-hmm. Hunter (25:45) And have some sense of what I think needs to be tested and at what level. Not always right, but at least a starting point. And there's a lot of tools to tell you, you know, our code has you know X percent of coverage or whatever. These are just sort of classic software development tools that you can point at who depend no matter who wrote the tests, whether it's the AI or humans, to get a sense of okay, this is you know. X percent covered, and that's what we deem acceptable. And this core module where the really important business logic is is 100% covered, and then you know the about screen is not, but we can live with that. I mean, you can make some sort of strategic decisions about what the most important thing is. And I will say this might be heretical, but you know, you could still write your own tests, human being. You don't have to have the AI do them, right? And I know some people that do that, that are due a sort of traditional red-green kind of style where they are. Brian Milner (26:29) Yeah. Hunter (26:38) They are having them, the humans, write the tests, and then having the AI system implement the features, knowing that they have put a fair amount of rigor into the tests themselves. They know that the tests are what they want them to be, and you know, specify the requirements in the way that is appropriate. So there's a there's a few different ways you could attack this. Or I guess you could just go work for one of the foundation labs where you get unlimited tokens. You don't have to worry about it. Brian Milner (27:01) Yeah, right. W that that must be nice. well we're we're almost out of time. I d I I wanna close with one big question for you though, and so follow along with me on this one. It might take a little bit of explanation, but If you had a high school buddy, an old high school buddy or family member or somebody come to you who is, you know, fairly technical, maybe not a programmer, but you know, in and around software for, you know, part of their live lives. And you know, they they're they're interested in this kind of stuff. They'd like to get their hand hands dirty on this. What what's one lesson that you feel like you've learned over even just like the last six months? that if if you were gonna give someone advice to start off and, you know, hey, whatever you do, make sure of this, what what would what would be your your biggest piece of advice for someone like that? Hunter (27:53) Hm, that's a good question. Gosh, I'm have this long indeterminate pause here as I try and think of a good answer. no, no, it's it's fine. It's a good question. I'm just trying to think of which way I wanna take this 'cause I mean there's the sort of very pat meta answer, which is like this is supposed to be fun. So at least in the in the scenario I'm imagining here, this person isn't like trying to, you know, feed their family whatever they're creating here. They're just trying to explore and, you know, like a hobby. So, you know, just try to remember the stuff is supposed to be fun. I guess Brian Milner (28:01) Yeah, don't mean to put you on the spot. I just Yeah. Ha ha ha. Hunter (28:27) I mean this is a very sort of a random one, but it works better than I think. So I'll just throw this one out there. What I've found, which surprises me, because I'm a very much a I'm gonna write a prompt, I'm gonna like actually make sure that the the punctuation is correct and that that's grammatically if I don't spelling mistakes, like I'll you know, and all the the stuff that I would do when I would be writing for other humans, I've actually found that you can literally turn on the voice dictation mode and kind of ramble and it will figure it out. It's really good at that. So you can even with your ums and the ahs, and like, go back. Actually, I don't think it should be green. Maybe it should be green. I don't know. Green's good, right? Okay, let's just do green. I mean this total rambling, incoherent mess. And it does it. And that's kind of amazing to me. So for someone that isn't steeped in the sort of software development lifecycle world, that they just want to play with this stuff, that might, you know, just try it. Try just talking to it and see what you get out of it. You might be surprised at how well it does. I certainly Brian Milner (29:01) Ha ha Yeah. Hunter (29:22) was 'cause my like I said, my inclination is to go the opposite and be very s very specific. Brian Milner (29:27) No, I think that's a great tip and and I think you're absolutely right that some people don't even know that there is a voice mode that you can activate it can talk back to you too if you really would rather have kind of that conversational style. if if you r if it really bothers you that it you know you'd have the ums and ahs and stuff, you know, typed out, you could use something like Whisper as well that would just translate and pull all that stuff out for you. I I I I I had to get used to like I I I prefer typing. I I just I prefer typing than to talking and I I had to get used to the fact that Hunter (29:47) Yeah. Same. Brian Milner (30:04) I don't have to go back and fix all the spelling errors or anything else. Like anything that like I double tap a key on accident or whatever. It knows what I meant to say. And you you kinda have to go back a little bit and say, well, what's my point here? I'm communicating with this thing. Does it understand what I meant? Yeah, it understands what I meant. Hunter (30:06) Yeah. Right. Brian Milner (30:22) So it didn't have to be spilled perfectly. And I I think that's kind of akin to this is is to say, Don't waste time, you know, going back and taking out the you know, the the double T that you put in on on accident. It it knows what you meant, it'll figure it out. Yeah. Hunter (30:25) Yeah. Mm-hmm. Yeah, it is it is I totally. And that's just, you know, that's just not the way that I usually work. So it's been something where I'm like, you know, I'm the kind of person that if I misspell something in a slack message after I hit enter, I'll go back and fix it, right? Like so Yeah, exactly. Brian Milner (30:47) yeah, yeah, me too. Me too, yeah. If I see it for sure that you know, I'll do go ahead change it. Yeah. Yeah, awesome. Well hey, Hunter, I really appreciate you coming on. this has been great conversation and thanks for sharing some of your experience and and just trials and and errors with us here as we kind of all explore this together. Hunter (31:07) Yeah, of course. Brian, thanks for having me and congrats on the new show. Brian Milner (31:10) Thanks. Brian Milner (31:11) Well that was just a great conversation and I really appreciate Hunter coming on. Hunter and I are colleagues and have been co-workers in the past and you know, we've worked together on building classes, we worked together at Mountain Goat and I've always really, really respected his Well, so much about him, j his his intelligence in this area and his skill, and his work ethic. So I just really appreciate all that he's done and and really thank him for coming on coming on and helping us to to get that little bit of a different perspective, maybe on on how this thing is is headed. So thanks again to Hunter for doing that. y we're we're still a new podcast. I mean this is only we're still in the the first five episodes here of this podcast. So what I'd ask from you, this is really important at this early stage is if you can like and subscribe to this and whatever podcasting platform you listen to this show in, really, really would appreciate appreciate appreciate that, excuse me. And if you would also you know, be so inclined. We'd love it if you would tell a a friend or a work colleague about it if you found this to be useful in your line of work and useful to the kind of work you do. we'd like this to grow into a community of people who are trying to figure this stuff out and you maybe maybe our collective brains together can you know kind of come up with with some ways to do this that that we might not have come up with otherwise. So yeah, we'd love to invite more people into this growing community. here for this show. that this show also does appear on our website. it's at agilityevolved.com slash podcast. You can find all our past episodes there. The show notes for our podcast are going to be there. Any other links or anything that you you might need to find about the show we'll we'll keep them there on that that website. So agilityevolved.com slash podcast. And as always I would really Really love to hear from you. I I really appreciate after the first few episodes here, we've gotten some good feedback from you, you guys, the listeners, and that that means so much to me that I hear from you, and it means so much just to to know what's working and what's not working, and what you'd like to hear next, who you'd like to hear from next. All those things are really vital, I think, to this being the kind of show you want to stay subscribed to and listen to over and over again. So feel Feel free to send me an email, Brian, at agilityevolved.com. That's my website, agilityvolved.com. And speaking of that company, agilityevolved.com, I'll just make a little shameless plug here for we don't have any advertisers or anything on this show. Really now, Agility Evolved is our sponsor, you know, my my own company. And what I just say is if you're interested in taking any kind of a scrum class, a certified scrum master or a product owner, or advanced versions of those classes, I Would just say, you know, check out our website, agilityvolve.com, look at the upcoming classes that we have there. Really appreciate you doing that. And you know, we can spend a couple days together talking through some of this stuff. Every class I teach is injected now with with AI and how to use that practically in our work here. So I'd I'd love to meet you in person or virtually over a Zoom call, and you know, we can talk through some of this stuff then as well. So that's about all we got for you for this week. So thanks for joining us, and we'll talk. Talk to you next time on another episode of the People Over Prompts Podcast.