LaunchPod AI - Derek Pharr
===

Derek: [00:00:00] We're a resource constricted team. And so just one afternoon I hopped into Rept in tandem with Cloud Code and said, look, look, just build me a prototype. And it wasn't good. I was so excited 'cause I was like, man, in an hour I did this from idea to proof of [00:00:15] concept and ruled it out and I didn't waste anybody's time including my own.

That feels really good. Welcome to Launch Pod ai, the show from Log Rocket where we sit down with top product and digital leaders to talk real practical ways. They're using AI on their teams [00:00:30] to move faster and be smarter. Today we're talking with Derek Farr, CPO at S Sparkle, who's leading AI adoption across both product and content teams.

In this episode, we'll discuss the three biggest barriers at any company to AI adoption and how Derek overcame them. How [00:00:45] Sparkle built a content intern with Claude Projects, it significantly improved user engagement and how Derek prototyped tested and rejected a new doomed product idea in under an hour.

Avoiding weeks. A wasted engineering time. So here's our episode with Derek [00:01:00] Farr.

Jeff: All right, Derek. Good to see you again man. How you doing? It's been a little while.

Derek: I'm great. It's good to see you too. I can't wait to get you back in Seattle and we can hang a little bit.

Jeff: It was fun hanging out. So for context, Derek and I met at uh, one of the dinners we did in Seattle. You and Neha know each other, who was on the show a [00:01:15] little while back too, right?

Derek: She's great. I mean, she's, if you haven't listened to her appearance on the pod, just put it on pause, come back. But go listen to her first 'cause she's fantastic. I've had a chance to get to meet her a couple of times here in Seattle and [00:01:30] she's just, she's smarter than anybody has any right to be.

She's great.

Jeff: Yeah, no she's smart people. The whole like building an AI intern in their slack and this was wild. So yeah, definitely check that episode out.

Derek: after listening to that I wrote her and I was like, you gotta send [00:01:45] me the prompt you used for that. And it was so good. Yeah.

Jeff: I was stoked on that one. We're actually trying to do something similar now. But good news is we have a whole different thing to talk about here. Like the whole theme of this show has been predicated on we're [00:02:00] all, as product and go to market, people trying to figure out how do you actually use.

AI on a, you know, day to day, week to week basis. You, your team, how do you enable people? Not just building it into the product, which I think we're all also trying to do. Everyone has the board going like, sprinkle a little AI magic [00:02:15] on there, but how do we actually use it to be, faster and smarter and better and all that kind stuff.

So, you have a bit of a different kind of set of things. You didn't build an AI intern not yet. At least it sounds like 

Derek: I, I want to, yeah. 

Jeff: You will a couple, shortly it sounds like, [00:02:30] but we're gonna talk about a bit around how do you actually get a team to move and

be excited about ai and I think you broke down like three kind of hesitancies that you've seen people have around it.

And then how'd you get past it? [00:02:45] How'd you help teams? Like really uplevel here. So maybe lay that piece out first before we go into it.

Derek: Yeah, absolutely, Jeff. It's an interesting thing, ? Like you get in tech, especially like Seattle's a very technical town

and you talk to people and yeah, AI is sprinkled everywhere. It's just the pixie dust you see as well. [00:03:00] It should be , like everything changes really

quickly. It's just ridiculous. And Sparkle is a content company, like we deal in trivia but we're also a tech company. And so you'd think that it would be this just like perfect marriage to sprinkle AI into everything we're doing. But [00:03:15] what I've found is that there isn't resistance, but there's hesitancy and when you think about it, it actually makes a lot of sense. So on one front we have like this amazing content team. And an amazing content community. And the content folks [00:03:30] are very hesitant around ai. They're like, this is coming from my job, or I don't want AI trivia, or, what even is this? And why is it, you know, why is, why are we here?

Like, I just don't wanna be here.

Why are we here? And I, I definitely [00:03:45] understand that like it's everything's changing quickly and it's a strange new world. Um. But it's hard to kind of allay that because I, I like the job market's hard and it feels like AI is coming for jobs everywhere. And so when people are concerned [00:04:00] about their job, I take that very seriously.

Like, I, I hear that in my soul, ? But for me, AI and for I think most of your listeners, AI is just, is this incredible tool and so. I think there's two parts where you kind of hear people's [00:04:15] concerns, but then you also outline like how good it can be a tool. And so one way I hear people's concerns and really let them know that I get where they're coming from is that writing and writing in general, like writing trivia is, it's a craft, ?

Like it's an art form. [00:04:30] And I can elevate trivia because I work in it and people might roll their eyes, but we take trivial things very seriously here at Sparkle. And AI is, at least in this moment in our history, not very good at writing trivia.

Jeff: Or anything else from scratch, really?

Derek: yeah, [00:04:45] exactly. 

Jeff: it's 

Still, it's still really, you can tell that beyond the M dash and all the kinda little tells, it still has that uncanny valley of it. It lacks, I dunno. I used to say it lacks taste, but now that's become such a, a trope to say [00:05:00] is just like, that's kinda the easy answer to say, but it does, it, it misses the mark.

Still, it's not there maybe one day, 

Derek: yeah, it's like there's a nuance to it, ? Like good trivia takes nuance. Anybody can ask the question like, what's the capital of New Mexico? ? Like, that's not good trivia. That's just a [00:05:15] question people ask. That's just stating facts. Good trivia is more nuanced. And so a better question, even though it's not a great question, is what's the capital of New Mexico spelling counts? And what that does is it layers this extra thing in. And it's a trick. It's kind of a mean trick because [00:05:30] you get people trying to remember like how many Qs and how many letters when in reality Santa Fe is the capital of New Mexico. But you know, that's a better trivia question. Then AI is even capable of at this point.

And when I start to have those conversations with people on the team and then start to illustrate ways [00:05:45] that it can actually make their job easier and free them up to be more creative people and do more creative things, then you start to see some of those light bulb moments where people come around a little bit more.

Jeff: Yeah, I mean that, right? You start to get into what makes good trivia and [00:06:00] there's maybe if you had a giant, giant repository of great questions that were all rated by. Response and how people felt about it and all those kinds of things, you could probably start to get there. But I dunno. I, I've, I've entered [00:06:15] huge amounts of context data into things, uh, into, you know, uh, GPTs and still in the end, what comes out needs a human touch at the end.

Derek: Yeah, it really does. And. That's actually a good segue if we wanted to go there into like, something that I've done [00:06:30] around our content that did lead to one of those light bulb moments.

Jeff: Let's let's hear it.

Derek: Okay. So you ever use Claude Projects?

Jeff: I am starting, dip my toe into Claude. 

Derek: One thing that I found that it's really good for is around content.

So I'll go into Claude, create a [00:06:45] project, and then just feed it knowledge, ?

Like you can give it this body of project knowledge and then you give it parameters. And like any good thing like this, you use AI to build ai. So I'll go into. Chat, GPT and say like, Hey, help me with the Claude prompt. And then [00:07:00] build that back over in Claude, test it out, see how it goes. Give that feedback either to Claude or back to chat and tweak the two working together until I get to a point where I feel like. It understands what I'm trying to do and we can work [00:07:15] forward, because for me, using any of these tools is a conversation. It's not a, it's not a set and forget. It's this back and forth and it's iterative and I think that's what most people are finding, that if you really wanna use these tools successfully, that's how you do it.

Jeff: So how are you using content though from that [00:07:30] standpoint? So you got cloud projects and , you put content into it. I've seen all over Wellington now, people saying, Claude is so much more of the code tool.

It's fantastic for content. I use it to edit this. I use it to do this, but I've never seen the actual, what does that look like? Or what do you do with it? [00:07:45] Just more I've seen people say, I use it for that and it's great, and it puts this magic.

Derek: So to get a little more specific, what I did was I took every geography trivia question we asked at

one of our pub trivia shows across the country for the last year, and dumped it out into [00:08:00] CSV file. And I uploaded that as part of like the knowledge base that this project has.

Uh, and then I put some of our style guides.

I put some of the like the output that we do from shows. I put just brain dump of thoughts from my own experience and all that kind of created this like, knowledge base. [00:08:15] And then the project instructions are, okay, Claude, here's what you're doing. You're gonna help me analyze trivia and related metadata. I want you to be able to summarize. I want you to be able to identify signals like duplicates, question length,

readability, missing fields, potential [00:08:30] bias. Look for patterns. All these kind of different

things, right? So it has this, um, kinda like Neha was talking about in hers, like she has a product intern.

This becomes like a content intern. And so with that and through back and forth with this, this product knowledge [00:08:45] repository, you take all of these geography trivia questions in there, and then you can start to ask it interesting questions like uh, give me every question we have about Michigan.

And so it does that. Okay. Tell me what questions I ask too much of tell me [00:09:00] where I might have gaps in our trivia

and what you find. I bring up Michigan because a lot of our content folks are in Michigan. We get a lot of Michigan questions now. It used to be way worse over the years, but historically, yeah, they ask a lot about like the Pistons and Motown [00:09:15] and the Red Wings and so when you can use a tool like that to just help you with some of your blind spots, in addition to being a sparring partner it's one of those moments when you show it to people, they're like, oh, another great actual use case of this is, if you're writing questions, [00:09:30] we have questions in a lot of different places.

? we've been doing this long enough that it's not all as consolidated as it should be.

And so when you start to consolidate it in this world, it really helps you with, okay, I'm writing a trivia question. Gosh, have I asked this before? Have I asked this recently? Have I asked it this [00:09:45] way? You pop that into your little project and it comes back and be like, oh, you asked this exact same question on February 4th at this show, and we have get rates.

People

got like 64% on it, and you can be like, okay, I still want to ask this question. So then [00:10:00] you can massage and go back and forth on how to ask the question differently.

Or maybe you ask for suggestions. Like, okay, can I ask this question that I asked about Michigan? But is there, can you give me ideas based on other content I've already fed you about Washington or about Boston, or about any other [00:10:15] number of places?

It's pretty cool once you put these things in here, it's just, it's getting people past a little bit of that hesitancy.

Jeff: Right. So, so you're actually recording not just the questions, but because I assume it's digital interface where you guys are doing triva. It's, the answer [00:10:30] is the correct rate, the wrong rate all that kinda stuff. 

Derek: All sorts of stuff, like what night it was asked who asked it, like which host did it, which part of the country. Right. We do trivia in like 30, 35 different

states, so it, we get regional [00:10:45] information and so we could tell. You know who's getting that question right? More in one part of the country versus others. 

We have A and B sets where we ask the same question, but we'll ask it maybe a little differently depending

on the audience. And so all of that goes in there. And then you can just, you can have a conversation [00:11:00] about the past to help you with what you're doing currently.

Jeff: I gotta ask, do you guys have a target correct rate, or how do you kinda look at that and go that's a good question. I assume you don't want a hundred percent people getting it right. 'cause it's probably too easy. A hundred percent. Like I'm a big Jeopardy fan, my wife and I watch Jeopardy every night [00:11:15] and we play against each other 'cause we're cool like that.

But. you know, It's definitely episodes where you watch and you go, I'm no Ken Jennings here, but I can usually hold my own and when I go through a round and get maybe two questions right. I'm like, well that probably wasn't that [00:11:30] great for episode, but you also start to look at the competitors and realize, they're coming out of the show and it was just a very low scoring game.

'cause they all barely got anything. Right. It's just, it's a boring show. Like, it's terrible to watch. How do you guys kinda look at that piece and does this help you?

Derek: So that's a great question. , it depends on the [00:11:45] question, right? So

when you go to a pub trivia night of ours, we, and this is kind of the, the dirty little secret, maybe I shouldn't tell, but I'm going to, is the first question of the night should be maybe the easiest

question. Not a hundred percent get right.

We don't want everybody to get everything all the time. It should be easy and it's game theory, [00:12:00] right? You just wanna bring people in slowly to the experience so you don't

just slap 'em. And so, like your first question, maybe even your first couple questions are gonna be easier so that people can feel smart

because they're smart, they just don't know it yet, right? Then as the game goes along and so like, that's gonna be like [00:12:15] 80 plus percent people getting it. You know,

The events folks when they listen to this might slack me later and be like, Derek, you got your percentages wrong, but just go with it.

Jeff: Directionally accurate, 

Derek: Yes, exactly 

And I'm gonna, I'm gonna use that term now, uh, for the rest of my life.

That's fantastic. But then as the [00:12:30] game goes on, you know, you want to dip into that like 50 to 70

range kind of depending, and then as the game is close to ending. You want it to go back up so that people kind of finish on a high note

and then we have a final question. And the final question should be hard, but not too [00:12:45] hard, right?

If 25% of the people are are getting the final question, that's not a great final question because that means only one in four teams, and it's not just individuals, it's teams. Went to pub trivia and got it right, and 

that's a little on the low side, but then conversely,

if it's too much the other way, [00:13:00] so there's, but to answer your second part of your question, yeah.

If we go back and analyze data. And we have a lot of things in Power bi and the more we move them into projects and other things like this to help us analyze that data and we see that, oh, our questions are trending too easy, especially in [00:13:15] one part of the country we, that's when some of those other kind of question sets come in where we go, okay, maybe these folks need to get something that's a little bit harder.

And maybe even the questions we're asking across the spectrum are too hard.

Jeff: And you could've, I mean, you could've done that before, right? But it, it [00:13:30] probably, the actual work to do it was exponentially harder and you needed a different skill set of person to do it than probably you had around. 

Derek: That's actually a really good point because like the person that you have doing it is more on the data scientist side of things. And they're great, like they're super smart people. But you want to [00:13:45] democratize the data science and put it in the hands of the people on the front lines, the writers, the hosts, the people running the events, and there's just this gap between the two.

But AI bridges that gap really nicely. So that's part of [00:14:00] the of. Of, of helping people just get past

even what they're afraid of.

Jeff: The content team starts to see it, not as it's coming to take our job, but it's helping me

create better experiences for people and , better questions and make sure my questions work, all over where they should. I mean, has you, [00:14:15] has that played out? Like, have you seen that play out from a standpoint of kind of the team adopting solutions or seeming more positive on insights from it?

Derek: there. Yeah, 

we're getting there. I mean, I think for both content and product, it speaks to the next level of hesitancy, [00:14:30] which is

just people are buried 

and so it's really hard for people to, to poke their head up and be like, oh wait, I gotta learn a new tool. I gotta figure out even what a language model is.

I

gotta, like, it's just, it can be a lot to ask. It's like people are just trying to [00:14:45] dog petal in the ocean and you throw them a puppy and so that's that's a lot.

So that, that can be really hard for people, which slows the adoption a little

bit. And so that, that goes back to a little bit of what I've been talking about, which is this just show, don't [00:15:00] tell.

Way of doing things. And I have the luxury of doing that. I'm in leadership. I get to be a little bit more divorced from some of the boots on the ground stuff. And so it gives me a little more space to learn these things, experiment, try different , projects, and then roll them out to people and get [00:15:15] people and involved and say like, okay, look. Look, I know you're busy. Just gimme an hour and I wanna show

you this thing. And so that helps. But it's a slow process just because people are buried and trying to get them to make the investment that will help them get unburied is hard when you're just trying to get through the [00:15:30] day.

I mean just to, just to free up that time for creative people to do creative things, I think is so important.

I mean, Ramon talked about this a while back, like earlier in one of your pods just on, everyone hates writing PRDs and everyone hates kind of gathering requirements. But if

you can use these tools to really help [00:15:45] make that process go easier and faster, it's great.

Now there's a little bit of a stigma that comes with it. I've had engineers when I've sent them something, I've used chat. They're just like, you just had a robot do this. I'm like, yeah, I did. Is it wrong? Is it more thorough than what I gave you before? And so that you have to kinda get [00:16:00] over, , even just the emotional resistance that comes with some of that.

Jeff: One of our things was you can't. Right from scratch with ai. You can't just have chat gt spin up, your post and send it to us and we couple push back that. Like that's, you know, that's not fair that you guys can use it and we can't.

And it [00:16:15] was like, well if you wanna use it to, to word something better or you want to use it to make it stronger, you can. But in the end it comes down to like, if I can't tell it's ai. That means you probably spent more time on it than just putting it into chat. And that's fine.

But the point is if we can just read it and tell like, no, [00:16:30] go, no, that's not how, because then everyone can tell, and that's not what people wanted to read there. But for internal docs, it works great.

Derek: I mean, internal docs, it's, it's amazing for like our help files. Like we used to have this

old, like clunky, terrible FAQ and we've moved that into just like a proper help desk [00:16:45] and just for like combing through with forums on the site of questions people are asking. Like you dump all that into chat and just be like, okay,

give me the top questions that people are asking. A lot of times people have already answered those questions for them. Cool. Let's just give [00:17:00] me something that's human readable and natural and simple, and then you pop that into your help guide. People don't mind so much that's AI generated because it's helpful. It's actually like,

oh, I want to know how to do this thing and it's just laid out for them nice and easy. and

so it's [00:17:15] really finding those use cases that work for people.

Jeff: Yeah, I think we kinda went about explaining the problem, the solution, slightly roundabout fashion, but Right. Like the idea is there, there's a cohort of people who just , they are factually really busy. And the funny part is they're probably the ones who would most benefit from [00:17:30] a lot of these things.

And so the answer there sometimes is. , either a, set a way that they can take some to go into it, or B, like someone who knows, who has been able to use it and can understand these tools, the ai, capabilities a little bit more. Help them get in there and help them solve one or two [00:17:45] things real quick, like get them some wins and, and don't tell them to go use ai.

Show them here's how you could approach this problem. But sometimes people are just so. Locked in and busy in their flow it's almost impossible to pick head up and take the time to do it. Which similarly, like if [00:18:00] you as a listener find yourself too heavy in flow to pick your head up and, and make sure you're getting, , this podcast every week and you're enjoying it if you're on YouTube right now, just take half a second.

Hit the subscribe button. Write us a review if you want. Give us a, like, if you're on Apple Podcast Spotify podcast. Subscribe to the [00:18:15] podcast, write us a review. It really, really, really helps get the word out to anyone who you think would benefit.

It helps us do the show more, bring more of this content forward to you. Uh, also the number one thing you can do, if I can ask for one thing, is tell a colleague, tell a friend. So if I ask one thing is tell [00:18:30] someone about the show before we move on, 'cause there's a third one and it's really good. But I do wanna go back to one other thing, which is you were talking about in Claude, I told you I was gonna have implementation questions in Claude.

So for the kind of trivia and answers and all that kind, are you just kind of uploading like a CSV after each show? Or, or how are you actually [00:18:45] getting that in there and kind of helping the AI understand,

Derek: So for our shows, they all go into Salesforce. , and then for deeper, like over time they get archived off. And so for that, I'm not an engineer. So for that I went to an engineer and said, [00:19:00] Hey look, I need all this data. And then the engineer goes, how do you want it? And I, so I, they gave it to me as a CSV.

But for other things, you can get it as JSON.

And those are the two formats I think that are the best for that.

And it doesn't have to be fancy. Just gimme a dumb [00:19:15] file and

let me point Claude at it. You could do more with Zapier and, and like true integration and all those kind of things and, and we'll probably get there over time. But for something like this where especially I would just want it to have a proof of concept to show people Yeah,

just give me a, just gimme a data dump and I'll [00:19:30] take it from there. And,

something like that. For the engineer, it took 'em like 10 minutes.

Jeff: yeah. As soon as the idea over time, you can just upload the kind of JSON from the most recent show and it will append or you're not that far yet into kinda like ongoing updates.

Derek: Yeah, exactly. Back to that whole data schema integrity thing. There's still more to do to get it more [00:19:45] consolidated. But yeah, I mean, we have in the past just Google Doc after Google Doc of trivia questions.

That's, it's just, it's too much. It's

Jeff: Yeah. I have found JSON to be a great unlock in a lot of kind of prompting stuff like that just gives you a little bit more structure. It's [00:20:00] simple enough that even non coders can understand it.

Derek: And non feel smart talking about it. Be like, oh

Yeah,

it just was a JSO blub we put in there. Yeah.

Jeff: exactly. But it gives you a lot more kind of control, especially I've been playing with VO three for video and that's really helped kind of

get some cool output.

So. [00:20:15] Alright, so the third one though it's a bit different, right? Because it's more around trepidation around things like environmental impact ethical, moral issues, the kind of it's a fuzzier issue.

So how did you go about that one? 'cause that's gonna be the toughest one, but I think every company's gonna have some [00:20:30] people who are gonna put pushing back on that piece.

Derek: Yeah, I mean it's every company, every family, everybody, like, there's a lot of meat in that soup and it's, it makes sense. Like none of this is great for the environment. I mean, just. It's not right

and [00:20:45] technology and nature are honestly, are often at odds with each other. And then there's just how I do my job.

Like, oh, my job is fine. I'm good at my job. Why do I need to change? Like I, this has worked before. Why do I need to do it again? Why [00:21:00] are you making me change? And so there's even that like ethical of like, how much do we. Try and get to the heart of what people are doing in the act of creation.

Like these are, these are hard, gnarly questions to try and just take on in a one-on-one or in a meeting with a [00:21:15] few people. And so again, a lot of this just comes down to like, acknowledging from the outset and hearing what people have to say and seeing, what can we do to alleviate concerns that isn't performative and actually helpful?

Right. So yeah, the environment's a big piece. And so one of the things we're looking [00:21:30] at Sparkle is, okay, can we get involved in plant more trees or can we do things to be more carbon neutral? Or, you know, what, what can we do to help offset the use of these tools as opposed to just running away from these tools because not using them isn't gonna work,[00:21:45] 

Right.

Like we're at the point now where. Not using AI is gonna be like showing up to work without your laptop charger. I mean, it actually might be without showing up without your laptop. Like

it's such a fundamental part of your job that you, you have to use it, but [00:22:00] it gets, it gets sticky when people have like very real objections to using it to put them in that position.

And so I think the more that we can do to. To realistically help people just feel better about using the tools and, [00:22:15] and put them at ease is the second part of that job. And then I think the third part of that is, is just continuing to experiment and try new things. Like each model is good for its own different reason, but each company behind it has their own [00:22:30] motivations and their own kind of. Set of core values, and so making sure that as a company the tools we're using are more closely aligned with maybe some of the core values of the people behind the tools. I think that can help a lot as well.

I try and also just swing the other [00:22:45] way. It's very real concerns and there are real actual big problems. There are,

but I we're at this really cool point in history. Like I've worked in tech for almost 30 years and I can't remember the last time I've [00:23:00] been this excited about. What we can do, like what we're capable of. And so I'm always talking about it. I'm always like trying to show off things that I've done, even things that haven't worked well. I'm like, look, this didn't work at all. I think that the more you can just [00:23:15] have honest conversations about it, acknowledge the bad parts, but also make space for the good parts.

There's real amazing world problems I think we're gonna be able to solve. History is littered with big, disruptive things. That

people have been [00:23:30] just terrified of for good reason, but then hopefully, kept in counterbalance with just amazing advancements and just incredible things that you can do. And I think that this is a big one, but it's honestly in some ways no different than some of the big changes we've been through [00:23:45] as a people before.

Jeff: The one thing I do want to jump back to though, 'cause we totally glossed over this, is on the element of kind of show don't tell. You also had a great kind of story around prototyping and rapid prototyping and Right. That's, I mean, [00:24:00] that's one of the huge use cases I've seen for this is it's lovable and bolt and V zero and a million tools kinda popping up every.

Every day it seems like to do AI coding. I know I've built a, I think I've talked in the show about, I've built a couple things that we use internally [00:24:15] now. They're never, ever going to be production ready for, anything more than five people here. But it gets us a lot of value and it's something an engineer never would've built.

But at the same time, you were also able to test out some new ideas and kinda quickly validate, you know what, this is a bad rabbit hole to go down. This is not gonna yield fruit. Let's [00:24:30] avoid it. It got you the answer a 

Derek: Yes. Yeah. I mean, so years ago, like a decade or more, we had like, we had this game idea of something we wanted to do on the site. And you know, for something like that, like I have to spec it, it goes through design, it goes through, [00:24:45] uh, engineering and qa. And we had something that went all the way to QA and we're testing it and we're like, oh, this kind of sucks.

Like, we don't like this. And so all of this like. Weeks and weeks of developments and work went into this thing that we scrapped

and we [00:25:00] just, it went out the window. So flash forward to, a couple months ago I had a different idea for some kind of like cool word play trivia thing, but I, you know, we're, we're resource constricted team.

I mean,

we we're, we're busy all [00:25:15] the time, as is everybody. And so just one afternoon I like popped into I think I was using Rept at the

time in tandem with cloud code and like I took like screenshots from an old book, uh, that had something that's kinda similar to what I wanted to do. And then I like [00:25:30] sketched some stuff out, like on my iPad.

And then I brain dumped all these things and said, look, look, just build me a prototype. And I didn't spend that much time, I mean, under an hour and went back and forth with it of just okay, now try this and what about that? And feeding in, you know, all this kind of [00:25:45] answers and then play tested it

and. It wasn't good, right? Like it wasn't a good experience. And I, Jeff, I was so excited 'cause I was

like, man, in an hour I did this from idea to proof of concept and ruled it [00:26:00] out. And I didn't waste anybody's time, including my own, because I

learned a lot through the whole process and we're not gonna do that thing and like that, that feels really good. The other thing I'll add is that like, through the course of just even building some of these things on your own, [00:26:15] whether they're good or not. It makes you a better product manager

because then you really, it puts you a little closer into the shoes of an engineer or a designer who has to build what's in your, in your head, and then you [00:26:30] start to realize like, oh, when I'm giving them requirements, I actually need to make sure I'm accounting for this piece that I didn't

really think of before, that they just do because they're good engineers or they know me, but you get more empathy for other people in those type of jobs. And you really started to [00:26:45] appreciate some of those

things, like different UI constructs, data schema, that you didn't just, you didn't have that perspective before because you weren't the one building it. You were just the one coming up with it. So there's so many tools out there to just prototype and build that, just jump in and [00:27:00] start building things.

Jeff: Yeah, and I think I've found, like when I first started out using these tools. I would, it took me days and days to build something extremely simple because

I kept hitting a point where I was like, oh, I messed this up, and I'd spend hours trying to fix [00:27:15] it. And then realized the only answer was to roll it back and kinda start again.

And I started to realize, oh, there's ways it does very well. There are paths you start to go down that are kind of hard to get off of once you're down to a certain level. And there's decisions you make, some of which have many branching [00:27:30] choices, and some of which have like only one or two.

And if you made the wrong choice earlier, you're gonna be in a bad spot. And. I'd like to think this is 'cause I've gotten better at doing this. I think probably what it is, is the tools have gotten a lot more advanced in what they're capable of doing. Plus maybe a little bit [00:27:45] of, I've gotten better, but also one big unlock was as much as, you know, everyone craps on, uh, writing PRDs, but I've started working with chat GPT to write it.

So I kind of give the description. I spent a lot of time writing what I want to do,

but I found a huge unlock has been asking it. Basically, once I [00:28:00] send it, say. Ask me questions one at a time until you're really confident in building this. And what it comes back with is usually either a things I hadn't thought about like, oh, that is something we need to figure out probably before we get [00:28:15] there.

Or b things. I just kind of assumed like, that of course is gonna go that way. And then it asks a question and you realize, oh, there's actually kinda three or four ways that could have gone. I really was obtuse on what that could have been. But doing that has been able to [00:28:30] really, I've been able to move a lot quicker on a lot of these tools now and build things that the team now uses a lot, lot more.

There's a lot more valuable because that was all forethought and I spend less time doing it, which is great.

Derek: Yeah. One prompt hack I like to do in that exact same use case [00:28:45] is, write out what you're gonna do, get there and go, okay, give this a spec review but be 10 x or a hundred x more thorough than I was.

And that kind of phrasing absolutely helps. The other thing in that vein, it goes back a little bit to [00:29:00] what we were talking about earlier. Like, I always have my AirPods in, like, I'm always just walking around and except for when I'm trying to talk to people in real life, I take 'em out.

But I have four dogs. And so I'm always walking dogs. Like, I don't recommend four dogs. Like, don't do that. But with four, you're [00:29:15] walking them all the time. And so I'll put my AirPods in and put like a note taking app on, or even just Apple notes and just brain dump. Just go and like, okay, here's this feature I was thinking about. I wanna do this, blah, blah, blah, blah, blah. I think we need to account for this.

Here's the use case. Then [00:29:30] I'll take that brain dump and then I'll run it through Claude and be like, here's all my raw thoughts. Help me clean this up so that it can get closer to something that resembles a product we wanna build, and then take it from there. So just kind of using those stolen moments of time to really to just get what's out [00:29:45] of your head into something more usable has been really helpful for me.

Jeff: Yeah. I've also, this is not an AI hack, this is a people hack a little bit, but I've found there's been a couple things that, there's the whole class of little micro tools that we [00:30:00] want internally for like the, go-to-market team that just is never, ever gonna be worth an engineer spending time on or is so boring or like the value.

Out of it is much less than the month of engineering time. It's gonna take them to make it or something like that. And now that's [00:30:15] kind of a couple hour project to, to do on our end because we don't need a perfect, we don't need perfect architecture. We don't need enterprise standards for safety and all those kind, you know, off and all those kinds of things.

But the other thing has been when we do have. [00:30:30] Some pieces that maybe we want in the product or, or that do require more, you know, real, uh, professional engineering. Um. It's a little bit like the theory, if you want information on Reddit, don't ask the question put the wrong [00:30:45] answer in Reddit and people come outta the woodwork to correct you.

I'll take a stab at it and get it partway with some of these tools and then show a couple of engineers, be like, look what I did. This is great. I mean, we're like halfway there already, right? And we've definitely had one or two where they're just like I have to fix this. And you know what?

[00:31:00] Uh, we had that thing much, much faster shortly after that. Did I get a little bit of trouble maybe from whoever was running the sprint? Maybe. Usually it was something we're happy we had faster, but, uh, it definitely kind of got people to, to move in a couple pieces faster than they would've [00:31:15] otherwise because it's much easier to kind of get some going also.

But I think there's the element of once they see it and they start, go, that's, it's actually pretty cool, but I could do this so much better. 

Derek: . Yeah. I mean, when you're building tools like that, it's really hard to go to an engineering team and say like, Hey, I want you to build this really cool thing, and it's for an [00:31:30] audience of two.

Like it's for two people, but then if you get far enough along and then you can even show them intent or you can show them like, okay, it's not that hard and it's not that hard in that, like you might think I'm asking for the world, but I'm actually [00:31:45] just asking for this little bit of a sliver. 

Then, yeah that, that helps you get a long way. 

Jeff: As much as I always have fun when I get to hang out with you, Derrick, and I'll be back in Seattle soon, and we'll have to, uh, you know, get together, you do have a, a actual job [00:32:00] over there. I do too. I gotta let you get back to it.

So, 

Derek: Yeah, I suppose. I mean, yeah, believe me, I think you and I could nerd out on this all day.

Jeff: Oh, I could keep going. But it was great having you on, man. Thank you so much for coming on. I really appreciate it. If people wanna reach out and ask more questions about kinda , how you guys [00:32:15] did some of the stuff around club projects or a bit of the implementation questions, if people have questions about there's a really cool use case too about how you guys organized your Sparkle Con.

Is LinkedIn the best Yeah.

Derek: LinkedIn is great. I love it when people wake me up on LinkedIn.

But I also write a weekly newsletter. It's [00:32:30] called Chief Rabbit, so chief rabbit.com.

And then finally, if you're in Seattle the last week of October is, uh, Seattle AI Week. Uh, and I'll be given a talk at that actually.

So, stay tuned for some details. Yeah, yeah, I just found out recently, so, and it's in kind of a similar vibe and [00:32:45] showing off some of the stuff, so that should be really cool.

Jeff: That'd be nice. All right, well there you go. You can catch Derek on LinkedIn or check out his newsletter or see him at Seattle AI week. And then, you know, if you're enjoying this, like I said you know, if you're on YouTube subscribe if you're on apple Podcast or a Spotify [00:33:00] podcast, subscribe, give us a, like, write us a review. It really, really helps. But, uh, above all else, tell a friend, but yeah. Derek, thanks for coming on man.