Dreams and Delivery
Real stories from tech pros who deliver at work and build their dreams alongside it. We cover the tech landscape in an age of AI with layoffs and opportunities that come from change. Techy creatives, this one’s for you!
Dreams and Delivery
Andrew Cunliffe: RavenMCP - What it Means to Be a Builder
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
This is a conversation about what it actually means to call yourself a builder in the age of AI agents, why judgment and taste turn out to be the scarce resource instead of production nowadays, and how the things you care about outside of work end up shaping what you choose to build.
At the moment in tech careers, tools are changing faster than job titles can keep up. Andrew Cunliffe has been living inside that moment for the better part of a year, building his own fleet of AI tools on nights and weekends while running design on Intuit’s AI+HI Expert Platform by day.
Andrew went from the 82nd Airborne to startup app development to 8 years at Intuit, where he now leads design for the AI expert platform inside QuickBooks and TurboTax. Outside of work, he’s built Raven MCP, an open-source design tool that runs a “taste engine” right inside your workflow instead of a separate canvas, plus HighLvl, a calorie-tracking app built for athletes.
Bonus: Andrew walks through a live demo of Raven MCP and explains why he made the deliberate choice to open-source it.
About Andrew Cunliffe
Andrew Cunliffe is a Senior Staff Product Designer at Intuit, leading design on the AI+HI Expert Platform inside QuickBooks and TurboTax. Outside of work, he builds and open-sources his own AI tools, including Raven MCP, a design system and “taste engine” for AI-assisted product work. Connect with him to see Raven MCP and his other projects in action.
About Christine Kahn
Christine Kahn is a Staff Technical Program Manager with 12+ years of experience across Intuit, OpenTable, 20th Century Fox, and Realtor.com, specializing in AI/ML platforms and program delivery. She’s also a published musician and sound bath meditation practitioner. Dreams and Delivery is where she brings both worlds together — real stories from people who deliver at work and dream beyond it.
© 2026 Dreams and Delivery LLC. All rights reserved.
Welcome to Dreams and Delivery, a podcast for tech professionals. Real stories from people who deliver at work and dream beyond it. I'm your host, Christine Kahn. By day, I'm a staff technical program manager in big tech. By night, I'm a musician and a sound bath meditation practitioner. This is a conversation about what it actually means to call yourself a builder in this age of AI agents, why judgment and tastes turn out to be the hard part instead of production, and how the things you care about outside of work end up shaping what you choose to build. Andrew Cunliffe is a senior staff product designer at Intuit, leading design on the AIHI expert platform inside QuickBooks and TurboTax. Before Intuit, he was in startups at this 82nd airborne, an interesting runway into a career spent turning complex requirements into interfaces opinionated people can align on. Outside of work, he's building his own fleet of AI tools, Raven MCP, high-level, a fully designed jujitsu brand, and a business platform and an AI-narrated daily show, treating nights and weekends like a second full-time job. Andrew, welcome. It's so nice to see you.
SPEAKER_01Yeah. Hey, thanks so much for having me. I'm stoked to be here.
SPEAKER_00So tell me, you went from working in the 82nd Airborne to startups and then eight years at Intuit. How did that path happen?
SPEAKER_01Yeah, I know. It's like kind of a weird path. So like most designers, they either went to art school or you know, some kind of media school or something like that. I'm born and raised in Rin County, just north of San Francisco. It's like where I live now. I actually live a mile from where I grew up.
SPEAKER_00Nice.
SPEAKER_01And I there's a lot of people in my family who had been in the military. My dad went to the Naval Academy. My half-brother spent 30 years in the army in special operations. And so when it was like time for me to get out of high school and stuff, like the only thing I ever really knew that I wanted to do was join the military. So I wanted to do that. And the the way to actually like pick and get a cool place, the army was like, yeah, if hey, if you want to be airborne and you want to be in the infantry, no problem. So I chose that. And I chose Fort Bragg, the 82nd Airborne, you know, storied division uh that jumped into all over Europe, Italy, Normandy, Ad-Day. Just really storied division. So it was really cool to be part of that. And while I was in, I did two tours to Iraq. I joined when I was 17 and like turned 18 right before I went off to basic training. Turned 19 in Iraq during the invasion. Did my four years, and I was like, yeah, I don't think this is what I want to do for the rest of my life, at least at this point in my life. I was 22 years old. You know, all my friends were like in college, kind of having fun, back home, living their life, et cetera, or graduating college. And so I got out and went to college, but I had no idea what, you know, and you have to pick a major up front. You can't just like find out. So I was like, well, my mom's in finance, my dad's an engineer. I don't really know if I want to do either of those, but I gotta pick something, I guess, along those lines. So I was really into MMA at the time, and I mean still am, but Chuck Liddell was the champion, uh, light heavyweight champion of the UFC and like a like my favorite fighter. And he had an accounting degree from Cal Poly. So I was just like, all right, I'm just gonna choose accounting because Chuck has accounting. Uh and I went to school for that. I didn't like accounting, so I switched to finance, got out, got done with school, did like a little internship in uh for a small hedge fund here. And I was like, man, I don't really like this either. And a friend of mine, he was working at Wells Fargo at the time was like, hey, my brother's like building apps for companies, and he's like really, he doesn't want to do the front end side, and he like hates talking to the clients and stuff. So why don't you go see if it's a match? And so then that's kind of how I got started. And we started building stuff for all these different companies across the United States, some in Japan and some in Australia. And then it was kind of like, you know, the thing about freelance is you're either like super busy and you're like, how do I stay on all these deadlines all the time? So I'm trying to deliver for three clients. And then it's like, I have no clients and no money coming in. Like, when's that gonna start? So it was like, okay, it's time for a full-time job. And I started on the engineering side, and I was like, yeah, again, I've like, I don't know how many times I've pivoted, but I've been, I was like, ah, that's not for me. So I went to design and that's where I've kind of settled in, but I've always like not fit the box for product designer because I like to get my hands dirty in the code too. And also, like, you know, I I went to school for business. I understand business metrics and like how products work and how they're marketed and all that kind of stuff. So I like to be involved in that as much as possible too. But really, like the the end all be all is design. If you don't have a well-designed product, you're not gonna have customers.
unknownYeah.
SPEAKER_01So that's mainly where I've you know had the biggest impact.
SPEAKER_00Yeah, really cool background. Just love hearing your evolution throughout your journey and in the workspace. Would you always consider yourself a builder, even prior to the whole AI builder craze that's going on now?
SPEAKER_01Yeah. I mean, like we can kind of talk about, you know, what I think about the word builder too. But one of the things I'm working on for my portfolio, I've been really into making like cinematic scrolling websites lately, especially ones showing timelines. Like I did one for the 9-11 Turbo, like its whole 50-year, like every model, you're just like walking through different period correct garages. Just something I did on the side because I was curious and want to see if I could do it and it would be fun. But like going back to even like middle school, I'm a 90s kid, middle school, early high school, like building GeoCity sites with HTML and like you know, inline styles. And then also, you know, I used to play a lot of games like Half-Life 2 and different mods for Half-Life 2, and I would make skins inside of Photoshop for like different uniforms and weapons and things like that, and then put them on the internet and the forums for like people to download. So I've always been doing stuff like that, you know, and like if it's not in a computer, which I've spent a ton of time, then it's yeah, with cars. Like I had a 1970 Dodge Challenger that was my first car, and I tore it apart and built all kinds of you know things. Like I can't, I just can't leave stuff alone. I always have to change it. I'm always trying to see how things work, why they work. I make them work better, at least what I think is better, which you know sometimes is very subjective.
SPEAKER_00Yeah, I love that. You're taking me back to GeoCities days, but also, you know, even myself, I tinkered with building my first computer back in the day. And it's really cool to have access to new tools now where it's enabling us to kind of dabble in new things and contribute in ways that we didn't typically think to. I'd love to hear your thoughts on what you think the AI builder is and what does it mean to you?
SPEAKER_01Yeah, totally. I mean, so we get that, you know, like you see that word kind of thrown around a lot. I think I mean and we probably talk about taste too. I see a lot of design leaders saying taste, which is something like I don't feel like we've talked about as much previously as as recently. It was always about like craft, and craft is there, and like taste is part of craft, I guess. But on the builder side of things, it's kind of interesting to me because there's some companies I've heard that literally gauge their employees' output on like how many tokens they've spent, right? On AI tools. But just because you're spending tokens doesn't mean you're building the right thing, right? So, you know, that's that's a piece of it. And then the other piece is like the mindset. Like I I've always had this mindset where I'm curious things can be different, and then I like to solve problems, which I think like designers are problem solvers in general. You know, I think a lot of people in the product space generally are problem solvers, they just kind of solve different problems in different ways. But to me, a builder is someone who is super curious. They want to push the limits and see what they can build. And then also when they see a problem, they're able to identify like, is this worth building a solution for or not? Because then, you know, if you're not building for a solution, are you actually building anything? Like, I guess you are, but maybe not really. That's why the term is like kind of funny to me. You know, it's like, did you just build something or did you build something worthwhile? You know, maybe that comes into like the taste and product sense piece of it as well. But there's a lot of people building stuff out there, and that's like I I know you read the post that I wrote on Substack recently. Yeah. But definitely like as designers too, I feel like we have this moral obligation to kind of like help people figure out what to build and how to build it. Because that's generally what a designer does. We're like, okay, first we have this business problem or product has come to us with this customer problem. We need to validate that problem first and then figure out how to solve it. And sometimes you might actually have a solution without a problem, but you also need to figure out if that solution does solve that problem well or what problems it even solves.
SPEAKER_00Yeah, you're making me think of a line that I saw on your Substack article about judgment and taste and how judgment without accountability is just output. And I'd love to hear more on your thoughts there.
SPEAKER_01Yeah, I'm it, you know, it kind of comes back to the same thing like just spending tokens, right? It's like, okay, I can just spin something up and put it out there. It's like, hey guys, look what I made, you know? But does it actually like solve a problem? I don't know. And then, you know, on the accountability piece, it's like, hey, I'm a designer. Now I can just spin up anything with Claude Code. But if I'm not, you know, using modern engineering best practices, like what's my stack? You know, am I using Next.js? Like, do I have proper Tailwind classes? Or am I just spinning up some random HTML and CSS, you know, to go out there and it's like not something that's extensible and doesn't have great components in it and things like that. You know, it's like who's accountable for that? Like it might be absolutely beautiful if a designer built it, but from an engineering standpoint, like that code was cheap, but the engineering is expensive. And it's the same thing where, you know, there's all these engineers potentially, or even people that aren't engineers that can spin up, you know, code with good modern frameworks. But what does it look like? Like, you know, how how do people click through it? And I talk about on there, if you if you extrapolate like how all the misclicks or, you know, when people don't understand what the page is about, and you think about all the time we spend on our devices, like how many days of our lives are spent with poor UX throughout, you know, our lives, right? So we we have way more time now, or we should, in order to work on the details and like, you know, really execution has always been the bottleneck. So strategy was compressed, and then the finishing touches were compressed as well. But now that execution is so much easier, like engineering is way less of a bottleneck because one engineer can choreograph, you know, however many agents to do stuff, right? And check their work.
SPEAKER_02Yeah.
SPEAKER_01You know, we should really be spending the time on is this the right strategy? And also on the finishing touches. And with that, when we are spending time on that strategy, we should be able to pivot really quickly too. Because it's like, okay, this isn't throwaway code anymore. That's like this huge waste of time because we can do it so quickly.
SPEAKER_00Yeah, well, you were saying about you know, anyone can build anything now. So say you you have engineering or product coming to you with prototypes, but it doesn't have any of this design built into it. I was talking to a friend of mine who's a product manager and asked her, if you could do your AI adoption differently, if you could start over, what would you do? And she said, I wish we had put in our design framework into all of the agents that we have, because people are coming to her with their prototypes, but it doesn't at all look like anything usable that she could go to market with. So I think being strategic and including design from the start would save some effort down the line.
SPEAKER_01Yeah. And I mean that's like another accountability piece, to be honest, on design. Like we need to be accountable to make sure that those kind of frameworks and things are available to our partners, right? So I I've done some of that stuff added into it, like on the IDS side. I mean, I don't, I'm not in the Intuit Design System team, but I have built some things for my team to use so that they are using actual IDS components and pulling them from our actual repos, you know, because we have MCPs and stuff that can, you know, that connect to all that stuff. And I think I guess maybe that comes back to the builder mindset too. There's like two other points there where it's one, instead of just building something, think about how it should be built and what you need up front, because that will influence how you build it, right? Like, how do I get you know the design system components in here? But fundamentally, even if you don't do that, I I'm always happy when a PM or non-designer or anyone else comes and brings me a prototype because it's so much easier to understand their thinking than something that's just in a PRD.
SPEAKER_00True. You know?
SPEAKER_01So it's like you're looking at words and you're both trying to interpret some vision in someone's head, you know? Um and it doesn't necessarily need to be a vision because design has to figure that out, but even just the basic kind of requirements laid out in a visual manner, because I mean we're all visual beings, so it really, really helps start conversations and frame them. So totally cool and fine with that. I think one of the biggest things I'm running into right now are people that like don't really want to adopt AI and don't really want to change their mind. And I think that's a bit of a friction point. You know, I assume it was the same when the internet came along and people didn't want to do that either. They didn't want to write emails. But I I'm definitely finding that that's a big, a big friction point as well.
SPEAKER_00Yeah, that that's interesting. And I would think even from a consumer standpoint, people using the tools and how they see chat agents and how that presents. But as you're building, how do you know when to trust the agent's output versus when to step in? That's probably one of the friction points where people are reluctant to just trust the AI.
SPEAKER_01Yeah. And it, you know, I mean, it comes in various forms, right? So, like in terms of coding output, I have a whole, and you can actually get a lot of this stuff on my on my website. I have a whole site just like dedicated to my method of using AI, but I'm constantly looking to see how I can be more efficient, more accurate, et cetera, with quality output. So one of them is like doing the work up front, you know. And I mean, I personally like to watch YouTube kind of instructional videos on how to do this stuff, but you can also have your agents just do it too. Like he can say, hey, Claude, go watch YouTube videos on X, Y, and Z and then come back and tell me how we should do this, right? And he could ask it for citations on like where to go watch it in the video so that you can go and see, you know, where it's at so all of its stuff is defensible, right? So I do things like that all the time. So, you know, when I go to send up a new project, whether it's, you know, uh an iOS app and it's in Swift or it's a web app, it's an X.js, one of the things up front is every time I kick off a project, the agent has to do a spec using rules and skills and my knowledge graph that has, okay, here's how to do all this stuff with, you know, modern engineering best practices, right? Because there's hundreds of books. And I mean, that's how like a lot of these models have been trained is literally scanning books, you know, into these models. And so I'm essentially doing the same exact thing, just you know, locally and it's saved in a markdown file on my knowledge graph. So that's like on the engineering side, that's how I have it. And then as part of my design tool, Raven MCP, I have a taste engine in there that's actually working on training a small local model on my home hardware. But even if you just download it and connect it, every time we kick off a project, it'll go through an interview on how do you want this to look? Who is the customer? Do you have any examples to share of what it should look like, et cetera, to try and build that taste up front? Because it's way better to get this stuff going up front so that there's a vision, like, okay, everything is spec'd out. You know, here's how you describe it visually, here's who it should talk to, the kind of voice it should have. It should, you know, be a light version of Linear's design system with, you know, one accent color or something like that. So you get all that information up front, provides way more context. Then the output is way more defensible and there's way less work for you to do on the back end. And you can kind of justify all of that because you did all the work up front.
SPEAKER_00Yeah, it's funny. Raven MCP is how I ended up reaching out to you because I saw that you made it available open source for people to use as their own design system. And immediately I had it audit my website that I built and it had really good recommendations and even identified things in the flow itself where I had some duplicate information on different pages in my site. And so I love that you've made this available for people to utilize in their own workflows. Can you tell us more on your thinking of the strategy going open source versus keeping it private and monetizing this sort of thing?
SPEAKER_01Yeah. So on the open source piece, you know, like I kind of went over earlier, I have like a very non-standard design background, right? And so when I was like building, you know, doing freelance work and building stuff for clients, like the person I was working with, we were using like every open source library possible, you know, because we were both like coming up. We didn't have a ton of money to spend. And he lived in New York, I lived in San Francisco. So it's not like we had a ton of extra money either. It's like the two most expensive cities in the United States. So we're using like every open source thing possible and learning so much from doing it, you know, things like Twitter Bootstrap. There's so many frameworks out there that that we use, free icons sets and stuff like that. And so I was like, hey, whenever I get to a point where I can do this, I want to give back to people. And so that's like one of the big ones for me. I always wanted to do that. I always wanted to make something open source that people actually wanted to use and provided a good level of utility. And then the other piece is really some of the stuff that I talked about in my recent post on Substack. And that's like this moral obligation for designers to really help people, you know, get quality stuff out there. Because one, the stuff's gonna come out, you know, whether or not we do help. And then it's either A, gonna have a poor effect on people because they're gonna spend time with bad UX and all these other things, or it might be a really, really, really good idea what this person is doing. But because the design is off in some way or another, or it looks like AI slop, people are just gonna disregard it. So I really wanted to help people get good ideas out there and, you know, get them with a high level of polish. And then as I was using it myself, I've kept evolving it. And it's kind of like why another design tool, right? Well, there's some really good ones. You know, everyone uses Figma. It's a desktop app. It's also super slow to load things and you're drawing things on a canvas, right? So there's an MCP and they have agents on the canvas. Not really the wordflow I like to use. I use Figma when I want to go like tune some really high fidelity like visual details on stuff and kind of do like animation keyframes to feed a claw. And then you have like amazing new products like paper and Dessin. They're really cool, they actually like take your own code base in, or they actually build it right there in code and you can pull it right out. And they have cool generative capabilities as well. You know, they of course all cost money. And so for me though, I was like, hey, I don't need another, you know, and there's like old reliable sketches out there, and it actually runs locally so you don't have the speed problems with Figma. But for me, I was like, I don't want another canvas, you know, like I don't want to have to open things, have more files. I just want to work right in my app. So that's why I built it, and that's kind of like why it's different from, say, a paper or a Destin or a Figma or whatever. You know, it's got design panels that. Can use and all that sort of stuff. But then to your point, you can also use it to audit all these things like accessibility and voice and tone and give you suggestions. It'll create a design system for you. Also, now there's you know functions. It's got over a hundred tools now. I started with like 12 or 15, now it's got over 100.
SPEAKER_00Nice.
SPEAKER_01Yeah, it'll like audit your designs and be like, hey, you're not using this component correctly if you have a component library, you know? And it's pretty much got all the functionality, not pretty much. I mean, I would say it's got like 75% of the functionality of your normal design tool now rendered in the canvas canvas, now like rendered in a browser. So you can just move stuff around there. I could even show that real quick too.
SPEAKER_00Yeah, yeah. No, uh, I'd love if you could show the demo mode that was on your website too. Cause I see that as a really good way for people to learn about design. Because I was clicking around and just learning about the different components and what you can do with them.
SPEAKER_02This one's let me share. Okay, here we go. I think this is the right one. Can you see my screen? That's my portfolio. All right, cool.
SPEAKER_01Let me just switch to uh my screen. All right, so this is my portfolio, and um it in itself is wildly controversial, I think. But also I make it so you can now design, redesign my portfolio. So in line. Um so I have Raven open right here. Any, you know, you can close it, open it. There's a tour as well. But just like any design tool, if I want to just, you know, click in here and see like, you know, what should we change this to? Um, and then I can change the size and all kinds of stuff over here. And then um here I can just drag it and you can just see what that looks like uh live. You know, so previously you would have to go and tell your agent, like, hey, uh move this up above this thing. You know, and that would be like one turn, or you try to package it all together and you know, it it wouldn't get it. But now you can just, you know, move it and see it. And of course, these are bad examples, but you can also give specific instructions right here, and then you hit, you know, this when you use it yourself, we'll say send to agent, not send three changes. Um, but this will send it to your agent, and then your agent will go ahead and make all those changes just as you set them up.
SPEAKER_00Um so you know, it's just the learning element I was alluding to is here, say I am a non-designer. There are probably elements here I wouldn't even think to prompt my Claude to change versus now I'm learning, oh, actually that's what that is, and here's what I actually want it to look like, and you refine it from there.
SPEAKER_01Yeah, and you know, these are things like even for someone like me, these, you know, these are all like CSS classes, and there's been so many added over the years. Sometimes I forget what's what. And so it's even good for me to see them as well. Um, and when you have different elements that have siblings too, you can select them all and change both at the same time, um, you know, making that easy as well. So, I mean, I just released this, so uh it's still a work in progress. But coming back to your question about open source versus monetization, so like I said, I always wanted to keep one open source, right? But then I have another one. Raven's got a big brother now that's kind of evolved out of it that I'm working on that's gonna be more geared towards teams. Cool. You know, so not I think a lot of the the design tools in the space are like really funny to have all the functionality to like kind of beat each other. And a lot of them, you know, pretty much all of them have a canvas still, um, even web-based ones, like a really cool animation app is Jitter, and it's still, you know, canvas-based, which works for a lot of designers and a lot of people. But also, you know, there's a lot, I think there's a lot of hybrid people out there now that are designers and engineers, and they just want to work with what's on screen. Because you're still going to be translating, even if it's rendering in a canvas and HTML, CSS, React, et cetera, it's still not the actual app itself. So I want to have somebody can do that. And then team functionality too. So, you know, a big problem always is are design tokens being followed? Like, are they being hard-coded or are they actually pulling on the right ones? Or like, is something being used that's not, you know, whatever the component is supposed to be, whether, you know, it's like a drawer or something. And like, is a drawer being used for use case that's not supposed to be used? Well, the the design team or whoever's in charge of that, back to the whole accountability piece. Like, they should be held accountable. So there's, you know, this this new, it's called Morvin, M-O-R-V-E-N. It's a Scottish word for mountain. And it's basically, you know, the mountain you don't have to climb is like all these alignments and things and like tracking down, okay, what component do they use? Do they use the right colors? Do they use the right tokens? Because it does all that for you. You know, I mean, that's the plan. That's what's I've been working on too. And then it's accountability in designers too. Like engineers could say, hey, look, I built this thing. You don't have a use case, you don't have a component that fits this, and I need one right now. And then it'll help you prioritize that stuff too.
SPEAKER_00Yeah, it's almost like they can speak the same language between the engineers and the designers now.
SPEAKER_01Yep.
unknownYeah.
SPEAKER_01So a lot of times, you know, these are conversations that like happen in Slack or in meetings and stuff like that. And then that's the other thing. Like, where do you keep the decision on where, what, why, and how? And so this helps you also hold those decisions. So it's like, why should I use this component here? And it's like, oh, you should use it here because it follows, you know, these laws, and here's why we made those decisions, you know, which is something that's like tribal knowledge that constantly gets lost.
SPEAKER_00Yeah, so true. Especially in wider teams and you're working distributed. Beyond Raven, what are you also building right now? I seen high level and an AI-narrated daily show.
SPEAKER_01Yeah. So AI by AI Daily was a cool experiment. Uh I ended up like stopping to do that because it's super expensive. But basically, I wanted, this is another like just builder mentality. I wanted to see what I could do, right? So I wanted to say, okay, can AI completely research daily AI news, right? Write a script for it, then create the video as well. And so it had an AI avatar. Her name was R, and she had like a British voice. And so she would narrate the whole thing, and then it would, you know, cut up the whole video using remote. And so there was like video from Wikipedia Commons and like video of the articles and things like that. I didn't get a huge amount of traction, you know. I mean, like most videos had, I don't know, they have like 20 views or something like that. There's so many people making content out there these days, you know. Uh, but it was just a cool experiment to do and see if I could do it. So I did that and it worked, you know. So it's something I really wanted to commit to. I could. Uh, and then yes, high level is my iOS app. So I I've for a long time I've wanted to create an iOS app. There's a lot of designers out there, like a guy like Charlie Dietz, who I've looked up to, who like has a super, super sick iOS app and you know, completely coded it by hand, etc. And it's like, well, you know, I'm a dad of two. Finally, my kids are five and seven, so I actually have way more free time than I have had over the last seven years. And they actually build stuff with AI too, especially my seven-year-old.
SPEAKER_00No way, that's yeah, yeah.
SPEAKER_01Yeah, he's made an arcade uh that's runs on iOS and iPad OS that anyone in the house, as long as you're on the tailnet, connected to the tailnet, can play in. So we can have like, you know, multiplayer family games and little like different games he's made, size throwing stuff. There's like a little plain one, it's really cool.
SPEAKER_00Fun.
SPEAKER_01Yeah. So uh I always wanted to have an iOS hat myself. And also I do jujitsu right now, and I've been a power lifter before, and like there's weight classes in both. So, like, you know, the last 20 years of my life, it's like my weight's going up or going down or whatever, and it's always about tracking calories. And there's probably a thousand calorie tracking apps out there, honestly. It's like super saturated. And I've tried a bunch of them, but none of them like were really geared towards athletes and more towards coaching. A lot of it was just like recording, like, here's how many calories you ate that day, you know. But mine has multi-model orchestration in it. So depending on the tier, you know, you get higher levels of frontier models. They'll literally help you plan out, and it's I try to keep it as simple as possible, too. There's like three pages in the whole app. There's just a chat, which is like your log. There, no, there's an actual log that shows like all your meals over every day. The chat's the main thing interact with. And then there's a weekly planner and settings, and that's it. So, like four places. And the weekly planner, it usually only touches like once a week. You usually you just plan out your whole week and you know, like people have lives. So you can be like, hey, on Saturday, I'm going to a wedding. Like, I'm not going to be following the diet. So I need you to like cut extra calories other days and it'll meal prep for you and be like, okay, here's your whole grocery list. And so one of my friends was using it and he was like, Hey, it'd be great if I could just, you know, point this at a grocery store's API. And there's really no good ones out there. So the solution for that was it'll create a grocery list for you. You copy it, open the app, whether it's Instacart, whatever, their AI agent, and just paste the list in there. And it'll just go and add all of it to your cart. So, you know, plans out your week like that based on your goals. And then, you know, you can log it via text, via type, via picture. And again, there's a lot of apps that do that, but it's also a lot of times they're approximations or it's community-entered stuff, you know, whereas mine actually talks to multiple APIs like the USDA food source API. So it's very defensible data on the calories instead of, oh, someone just, you know, estimated this, et cetera. You know?
SPEAKER_00Wow. That's super cool. And really niche for someone who's looking for that experience. I love that you've made this. Well, what is it called again? High level. High level.
SPEAKER_01H-I-G-H-L-V-L. Yeah, and I'm actually planning on with that one, scaling it to different sports and other things. Like high level is more like a brand. And this is the, you know, the diet calorie tracking app. But I think there's a lot of opportunities out there for different sports. And so you, you know, there's a lot of people who do it for lifting, like they have AI-powered workout programs and stuff like that. But as the models get smarter and they can actually coach and can critique more via vision and other things, we're gonna see it doing that live.
SPEAKER_00Uh, very cool. I meant to ask earlier when you were talking about the the AI generated show that you were experimenting with, where would taste fall in that sort of thing where a lot of it's just automated? Would it be plugged into the original infrastructure or would you like check in on it every so often?
SPEAKER_01Yeah. So that one, the taste is really yeah, up front again, right? And so you really kind of want to nail one. I don't think I ever actually nailed one. I think I probably got to like 70%, to be honest. Because it's kind of a hard problem to solve, right? Where it's like I have to have it generate everything and I can't do anything on it at all. And also I was doing that back in, it's crazy how fast stuff has evolved because I was doing that back in like April and May. And I think there was, you know, not that many frameworks available. Like Remotion was like the only one. And now there's like so many more things that have come out, or at least I didn't know about any more and I couldn't find any more at the time. But really, the taste there is like, okay, how do you set up the video? Obviously, the avatar. I worked a ton of time on the avatar. I worked on it with my daughter. She really was like interested in doing that with me. Um, so we worked on that together, getting the voice right, getting the tone right, the look, et cetera. Then, you know, it's like, okay, how should the flow go? Like, how is the intro? And then a kind of template for, you know, how the run-of show goes. And of course, I had Claude watch numerous, you know, YouTube videos with where they have a ton of followers on the channel and kind of borrow their best practices. So, like, when to cut, and like, you know, moving the avatar down to a smaller circle down here, and then you know, taking over full screen at certain points. And so there's like these different cues, you know, and then okay, what should, you know, where do we kind of have colored and different typography assets as part of the template as well? So all kinds of things like that, you know?
SPEAKER_00Yeah, interesting. And even if even if you're no longer doing it, there's still maybe a lot of learnings there that you could harness for something else, I would think.
SPEAKER_01Yeah. Yeah. I mean, there's so much stuff that I've done that have been just total experiments and I've just wasted like hundreds of dollars on in tokens that I've learned a ton of stuff from, though. And like that's been really kind of the journey. I think even though I didn't monetize it, and in fact, I lost money on it, I lot I've, you know, learned enough from it that it's just gonna carry me forward in other places.
SPEAKER_00Yeah, I think it's so cool how you started out the 82nd, and now you have your own agents, like almost your own fleet of AI. So it feels like a full circle from your where your career started and where you're at now. But uh for those who are starting their UX career, what tips would you give them in this new environment with AI in the mix?
SPEAKER_01Yeah, I mean, number one, be super curious, right? Like that's so the biggest difference. I I feel like I hadn't really evolved my tool set that much over the last few years is like basically incremental here and there, because a lot of the stuff was the same, you know? It was like, okay, I'm getting faster at Figma, like, you know, 20% faster, 10% faster, whatever, every quarter, every year, something like that. You know, better at doing user interviews and things like that. But there were there haven't been like a ton of new tools, at least that I was seeing until we really got a hold of AI. And like this is how new I am to AI, too. So I got Cloud Code in April and it's now almost August. And in that time, I have built and released my own design tool. I have an app in the App Store, I have another one in the Google Play Store that's waiting to get approved, like done all this stuff. That's how fast you can do it. And that's someone who's working a full-time job and has two kids. So if you're just getting into this, like beg, borrow, and steal anything you can in order to get token access. And uh, you know, go out there and, you know, build stuff, make stuff real. Like you can make whatever you want if you have the time and unfortunately the money because of tokens, you know. But um, that would be my biggest thing is learn all of the things you need to know, like UX best practices, really focus on your visual design because that's something where it's always going to be subjective. It's almost the number one thing people react to. Even like any design hiring manager, that's the first thing they look at because it's the easiest thing to react to. You know, like how's your spacing? How is your typography? I mean, typography is like 90% of design, right? So yeah, I would be definitely focused on those core things and then really pushing the limits on the AI stuff. And I would say try to be as much of a generalist as possible in terms of design versus engineering. If you can do both really well, like the whole design engineer role now is really taking over a lot. Like I'm pushing PRs. I pushed two PRs in the last two weeks, you know, at like the senior staff level. And so if you're a you know, a junior designer coming in and you're like really in the weeds of tactical delivery all the time. I think if you can push PRs and take that off of one, take it off your engineer's plate, and two, champion that for your design org, those are things that will really set you apart.
SPEAKER_00Awesome. Well, thanks so much, Andrew. Really great hearing your stories and all the tips. And definitely we'll share all the links to the different tools that you spoke about today. Cool.
SPEAKER_01Thanks so much for having me. I could honestly talk about this stuff for like hours on end.
SPEAKER_00To everyone listening, if this resonated, subscribe and share it with someone who needs it. I'm Christine Kahn, and this is Dream Send Delivery.