Eddie: We had no meetings. We had no textbacks. We had no Figmas. We had no Jiraboard where we tracked stories or tracked work. We had nothing. We used Cloudflare worker for the actual agent loop and Vercel AI SDK. That's it.

Claire: People get really intimidated by the idea of building an agent and I'm like literally it's an agent SDK running somewhere in the cloud and if you use AI SDK you get to switch your model. That's it. It can look up files. It can have tools. It's really not that scary and complicated.

Eddie: Co-founder was primarily built by five folks over the course of 10 weeks from initial idea, zero code to a tier one launch at Gusto.

Claire: How did you all make those precious decisions that only genius product managers can make about in or out of scope? Will this matter?

Eddie: We would build features and we'd have a discussion like does this make sense to have or not? If it is then it would get code reviewed right then and there and if not we would just delete it.

Claire: I call this the trash can method of software engineering right now where you can actually trash all the code, start like a [music] slash V2 branch, and rebuild it from scratch and it's totally reasonable to do because the cost of the code is so [music] low. Welcome back to How I AI. I'm Claire Val, product leader and AI obsessive here on a mission to help you build better with these new tools. Today I have Eddie Kim, CTO and co-founder of Gusto and he's going to show us how he three engineers and one designer completely rebuilt their app in just 10 weeks. This is not a small company and yet they're shipping like they're a startup. Let's get to it. This episode is brought to you by Magic Patterns. Today's engineers use Cursor and Claude code to ship features in hours that used to take weeks. [music] If you're a designer or PM you've probably felt a shift too. The pressure to move faster, validate sooner, and keep up with a team that's operating at a completely different speed. You've already tried AI prototyping tools to close that gap. But, if your prototypes don't look like your actual product, [music] it doesn't matter how fast you can build. You still end up redrawing it by hand. Magic Patterns takes your product team [music] from idea to production and works from your real design system.

背景音乐: [music]

Claire: When you build a prototype, what you get back actually looks like your product. You'll validate faster, get alignment sooner, and [music] when it's time to build, engineers can connect your prototype to Cursor or Cloud Code with the Magic Patterns MCP to pick up where you left off. Your eng team has their AI advantage. Make Magic Patterns yours. Try it today at magicpatterns.com/howiai. Eddy, thank you for joining How I AI. I'm so excited you reached out to chat because one of my favorite themes that I'm seeing right now is CEO, CTOs, founders, executives getting back to building product. And you're here to tell us how you built some product with the team both in a very fast and in a completely different way.

Eddie: Yeah, thanks for having me and uh great to connect again.

Claire: Yeah, so tell us the story of Co-founder and why you think this is, you know, given all your experience, you know, running this company, being a leader, being a builder, why this is so different than what you've seen before.

Eddie: So, first of all, the product is really different from anything that we've built before, but what I'm actually really excited is kind of more of how we built this and how we can take the learnings that we had um in building Co-founders, the quality and the speed in which we built this, and apply this to, you know, the the rest of our R&D organization here at Gusto, which, you know, we have over a thousand people. And so, I think we're just, you know, I feel like I just discovered on something like really incredible. And I just want to like kind of figure out how do we spread the word to not just our company today, but to any kind of engineering, um, design, or product management discipline person.

Claire: And so, you say you have about a thousand people in your R&D organization, but remind me, how many people did it take to build this product?

Eddie: Gusto co-founder was primarily built by uh five folks, myself, three engineers, I count myself as an engineer, so four engineers, and one designer. So, we had one designer and four engineers, and we we built it over the course of 10 weeks from initial idea, zero code, um to 10 weeks shipping it to a tier one launch at at Gusto.

Claire: It's got to feel so much so fun. I don't know if you're like me, I I spent the past couple years uh pre-AI spending so much of my time like prioritizing and planning out quarters and saying, "Can this come in Q1, or is this going to be a next fiscal year thing?" And now, where you can take, you know, peel off less than half a dozen folks and ship a real

Eddie: Yeah.

Claire: a real product line completely net new in less than a quarter, it's pretty amazing.

Eddie: It is incredible, and the origin of this was not on it wasn't on any road map. It was not anything that we really thought about. It kind of just came about because um, I was actually on vacation in February, and I was flying back from uh Madrid, uh my flight had a layover in London, and my the mid the flight from Madrid to London was delayed, and I I just barely missed my flight from London to San Francisco. And I had this like five-hour layover now because I missed the flight, and they had to book me on the next flight. And I was kind of pissed about it because I was so close, and so I was like, "Okay, now I have like five hours to like waste. I'm going to get home super late in San Francisco. What do I do?" And I had been like kind of playing around with Cloud Code. I was probably one of those like tech leaders that you sometimes hear about that like vibe code something over a weekend and then come to their engineering team and say like, "Look, I built this whole thing. Why can't we do it yourself? You know?" Like and then just kind of like piss people off because, you know, first of all, that was like completely vibecoded. It's not production ready. Um but that was that was my start into it. I was I started using Cloud Code a lot more for the prototype things and just kind of like, you know, materialize random ideas that I would have. And so, in this like 5-hour layover, I just took out my computer and started Cloud Coding this um idea that had been percolating in the back of my mind and just seeing how far I can get with it. And by the time I had actually landed in San Francisco, um I had this like prototype of of what ultimately became Gusto co-founder. And um I just took it to a few engineers that I talked to regularly, our our senior engineers. I talked to Katie Kovalcin, our designer. And we just started like riffing on the idea a little bit. The ultimate uh materialization of Gusto co-founder is a little bit different from that prototype that I had made. Um but that was the really the origin of this all of this all. And 10 weeks after that, like we we shipped this thing.

Claire: So, what I love about this story, I'm going to take a little detour because I've heard this over and over again from friends, is you know what I think companies need to give? Just a little bit more time off. Parental leave and like long flights. That is where I have heard the vibe coding magic happens over and over again. Like, "I was on vacation, but

背景笑声: [laughter]

Claire: I Cloud Coded something awesome." Or "I was on parental leave and I've got like a baby in one arm and like Codex in in the other and I've shipped something." I just think it's going to it's going to trickle down into like maker schedules a little bit more where if you can give people just a block of time where they're not in meetings, we can move product forward a lot more meaningfully. So, that's a little a little hypothesis that I have about how how some of our work might need to shift.

Eddie: I I totally agree. Like now after going through this experience, I mean I I kind of want to take more vacation, honestly, cuz I actually I actually went to my wife and I were going to take another vacation because like this huge thing came out of the last vacation that that that we took. And so maybe like another thing will come out of it. So, let's go like let's figure out where to go and then you know, if I have a 5-hour layover, I'm actually going to be really really happy about that.

Claire: Okay, this is how every uh CEO or executive can land the AI adoption pitch. It's like, "Look, if you go on vacation and write code something awesome, we're we're happy to give you happy to give you more." Well, let's go back to how you built this. So, you were on a layover, you have this idea was kind of like percolating in the back of your mind. You built a prototype, you brought it back to this like council I call them council of elders like this council of a product builders or engineers just saying like, "What what would it take?" Tell me though how you how those 10 weeks actually happened. And maybe show us some of the process and some of the artifacts along the way that were a little bit different.

Eddie: Yeah, I mean there was no like official council. It was like just senior engineers that I regularly like IC's that I regularly talk to and you know, I recorded a loom of this thing and I just started sharing it around. Um and these were the folks that kind of like engaged a little bit more in the conversation like, "What about that? What about that? That's pretty cool." And then so we have at at at our company this thing called anchor week uh which happens quarterly. We all like kind of the senior leaders and senior IC's um across the company they they meet in one of our offices. This particular one uh was in March uh in in Denver, Colorado in our office up there. And so we just reserved a room that Thursday. I think it was like March 20th or something like that. And we we just started like whiteboarding with this like group of five that I had sort of, you know, had that had expressed some interest in this. We just white boarded like what this would potentially look like. And it was just like literally a page of the Gusto co-founder app. And we just got to building. The crazy thing is this was more defined our build process was more defined by like what we didn't do versus what what we did. We actually just zeroed everything out. We had no meetings. We had no text backs. We had no Figmas. We had no Jira board where we tracked stories or tracked work. We had no stand-ups, no retros. We had nothing. The only thing we decided to keep was 24/7 perma-zoom, which is basically a zoom room that we just keep cuz every cuz every everybody is remote on in this particular group. And we just had this zoom room that's going 24/7. And some people like like to honestly stay in there all day and just like do their work. They just kind of like sit there quietly. And some people will like kind of pop in and out when they need something. That was the only structured thing that we had was this zoom. It was like literally just zoom, a lot of cloud code, tokens, and some like really passionate people about turning this thing in into reality. And that that whiteboard um that we took a photo of was like literally the only documentation that we ever produced in this whole 10-week process. And I can't tell you how many times I loved as people like caught wind of this thing getting built. They were like, "Oh, can you send some documentation on how this works?" And I love to say, "We don't have any documentation."

Claire: Okay, so this is the whiteboard. And just for folks that are not listening, it's a scribble a that says chat. It's a chart, it's a thing that says task name, and then like a couple components on the side that are not very well specked out. And what you're telling me as CTO is we built the entire product basically off this cloud code in a perma-zoom.

Eddie: Yeah, totally. And And if you look at obviously things have evolved a bit, but if you look at sort of the core primitive and core functionality of Gusto co-founder, um it's largely the same. They kind of like change the names of things, but one of the main things in Gusto co-founder is what we call now an automation. And it's basically just create these automations that like run these workflows for you. Um that's what you see here on this whiteboard called tasks, right? So it's like tasks have task runs. Um And that that's exactly what we have in co-founder today. Um we have assets of So tasks runs can produce these um charts or documents or markdown files. We call those in this whiteboard here assets, but eventually those got got got got renamed to artifacts. Um and then you obviously have chats. You have like suggested tasks. So like basically this like if you look at co-founder today that it is like the the actual like materialization of what you have on this whiteboard. Looking back on it, I'm actually kind of surprised right now how how close we we stuck to this.

Claire: So I have to ask you a couple tactical questions about how you pulled the this off because that's what people listening are going to want to know. So my first observation is you said you had four engineers and a designer. I don't I don't hear a product manager in there. So how did you all make you know those precious decisions that only genius product managers can make about in or out of scope, will this matter? I mean this sort of feels like the thing where you just had conviction and knew you needed to build it, but you know, in this world where there was no actual titled product manager on the team, how did you approach product decisions across this team of five?

Eddie: Yeah, I would say everybody was kind of a product manager. We would build features and we would go in that perma-zoom and sort of share it with each other. And we just have a discussion like does this make sense to have or not? If it is then like it would get code reviewed right then and there. And if not like we would just delete it. And normally that would be really really hard to do cuz like it takes so long to build something. But these days with Cloud Code we literally like write something a feature. We submit we open a pull request and this is not a draft pull request. This is actually a pull request um that is ready for a uh uh human code review. And we discuss it. We discuss if that's the thing that we want to have a functionality that we want to have. Um and we're okay with deleting that pull request if if the answer is no, right? You have a perfectly good pull request that was written that's like ready for human review. It's not slat. You actually like spend time to like you know, make sure this code works and and is written really well and then you just close it sometimes. That was like the that was like how we figured out what decided what goes in and what doesn't go in to this feature. Like the cost to write code is now so low that you can actually uh build uh products in in this way. And whereas I think you couldn't do that you know, six months ago.

Claire: Yeah, I call this the trash can method of software engineering right now where you build code and you're like actually literally okay with throwing it in the trash. And I I see two models of this. One is exactly what you say like a PR and maybe a preview branch branch to even validate is this the thing we want to build? And if the answer is no, you just close the PR. The other version of this that I do quite frequently is let's say we ship this V1 in 10 weeks. Customers start to use it and then we actually have a sense of what the product shape and architecture should be. Like it's very cheap to just build again from zero. You don't even have to build on top of what you've built. You can actually trash all the code, start like a /v2 branch, and rebuild it from scratch, and it's totally reasonable to do because of the cost the cost of the code just is is so low.

Eddie: Yeah, I have I have a funny story about that because um this prototype that I brought to the team, um I was you know, I was pretty pretty happy with it. Like and when we decided in that um that whiteboarding session to like actually like go ahead and build this, like my assumption was that we're going to continue to build on my prototype. And so someone one of the engineers brought up, "Hey, what do you think about like uh using uh building this in TypeScript and using a Cloudflare worker and kind of like making the actual agent like sort of a stateless thing that sits in its own repo." And I was like I didn't really didn't want to do it. I feel like I felt like this is my code. Like it was good code. And here they were suggesting to us start from scratch. And ultimately like I I you know, I trusted them. So I we agreed to it. We deleted my code and then we started from from scratch on that day. Um in hindsight that was like absolutely the best decision. And now but I think back then I had a lot of discomfort with that because like I had invested, you know, um that was my code. I like I don't like you don't like to delete perfectly good code. Like throwing something some good food in the trash. Uh but now like I'm totally used to it, right? It's It's there's no like sense of loss when that happens anymore.

Claire: It's It's totally It is just totally wild to me how how frequently I find myself writing lines and lines of code that ultimately are are great and never make it. Either because they don't hit my product bar or they just aren't technically how we want to implement. Um and then how often I'm just like taught like major red diffs uh all the time to just re-architect stuff. I do want to ask you a quick question about architecture. Tell us a little bit about um, co-founder and kind of some of the primitives behind the scene. I heard you say Cloudflare workers. People are actually pretty interested on how to technically build agentic products and so I'm curious kind of what stack you all landed on and and why and how you how you chose that architecture.

Eddie: Our stack is surprisingly simple. We build on we use Cloudflare worker for the actual agent loop and Vercel AI SDK. That's it. We don't have any other harness on top of that. Everything else was built in-house. You know, in the past I would have thought about how to use, you know, some third-party tool for memory or planning or things like that and it's really just, you know, memory to us is a tool that writes to a database column called memory and that simple, right? Everything is just like all the harnesses and things that we used to build like to is complex like AI agent loop stack. I think is no longer needed. It was literally just um, um, Cloudflare worker and and Vercel.

Claire: Ah, I love it. That's very similar to my stack at at Chat Purity. So it's it's good to hear that, you know, people get really intimidated by the idea of building an agent and I'm like literally it's an agent SDK running somewhere in the cloud and if you use AI SDK which which your which switch your model. That's it. Like this it it can it can look at files. It can have tools. It's really not that scary and complicated. Um,

Eddie: Yeah, it's not even I was like blown away cuz it's my first time using it that there's not even a loop. There's not a while loop somewhere, right? It's just like call stream and it takes care of the loop for you. It's crazy.

Claire: Yeah. Well, I have another question technically on how you how you ran this because PR as like the PRD is almost almost what I'm hearing, which is PR is the proposal of what to build, how it gets built, the solution, you look at it in code, and so you've almost compressed that that loop. Were you merging those into your production app and putting them behind a feature flag and just Is that how you were technically managing this development process?

Eddie: Yeah, exactly. So, I mean, that that is one of the benefits of like kind of a zero-to-one type thing. You may not be able to do that in all instances. Um but we had essentially a a hidden page in our web app, and we would land stuff into that hidden page. It would be on production, and then we would sort of like chip away at it. Like I think the analogy we use a lot was like we started with a block of marble, and we're just slowly kind of like chipping at this thing and forming it into like a piece of art over time. And we're doing that in production in place. So, one of the really cool things that we did, uh which I would highly recommend, is Katie, our designer, she actually shipped started shipping to production like a a faked experience essentially. So, it had like the UI, the the thing that we white-boarded, but if you go in and like, you know, hit put in some text into the text field and hit submit, it would just give you the same response like uh every single time, right? It's not actually going to any kind of agent loop. It's not hitting any kind of database. It's pure front end. It's kind of like what what you would maybe build as a first pass on Lovable or or something like that. And that actually got shipped to production behind this feature flag. And then in parallel, the engineers would start to build out the data models, uh they'd build the the agent loop, and they would start to connect this to the faked front end experience. And the front end would stay the same, but the canned like responses would get slowly and in place not even better. Like Like would actually be like real, right? And so you would see this thing kind of like morph from like um a a literal prototype that you would normally have thrown away because like it's just for demonstration purpose, but it would actually turn we would like like literally breathe life into this thing over time. And that's kind of how we how we built this. Um at the same time we would have engineers like build features and they would you know, have to do some front end stuff. They would they would ship the design for this, right? And the functionality on the back end would work really well, but then you know, the front end probably could use some improvement there. And we would ship that, too, right? And then and then Katie the designer would then go in and sort of like, you know, make make the experience a lot better on the front end. And so that was another example of like just shipping something into production and then like continuing to chip chip away at it to make it better over time. So if you look at it at any given time, there was always something wrong with it like in production, but then gra- over time like it turned into like this really nice thing.

Claire: And tell me a little bit more about Katie. We might have her on on the podcast, but Katie the designer is shipping stuff into production. Her code is the skeleton on which a lot of this functionality is is being built. Tell me how you know, was she was she technical? Is she technical? Was she a software engineer? How did she come to shipping code? Like what what blew your mind about her role in this team?

Eddie: What blew my mind was that she would turn into this incredible engineer. And I was just looking at our our our our PR stats, which we use this this tool called DX for it. She across our entire R&D org, she is night- in the 94th percentile of uh true throughput, which is a measure of like how many PRs you're landing into production. And that includes every single engineer, like classically trained software engineer in the company, right? She's out of 20, she's she's a top, right? Uh which is which is kind of insane and it's really good code. Like I asked her like, how how what is special about you and how do we get more of this like across, you know, design and and product management? And her answer was basically twofold. One is that she she's not an engineer, uh she's not a software developer, but she she feels like she was a little bit more like technically curious than than most designers. So she just kind of has a little bit of that more technical bent. And then most importantly is that she had a team of like three or four engineers, particularly on this team, that was willing to actually, you know, review her code, like give her feedback, um show her like how to prompt Claude a little bit better and like also how to judge um that taste of like what is actually good code that it's producing and what is not. We took the time to kind of like, you know, uh pair pair with her, right? Whereas I think a lot of engineers, what I hear them say is like, well, it just slows us down and like designers should just focus on uh producing Figmas so that we can like really focus on turning that into real products. And yeah, that's true in the short term, but like now I think once you make that investment, you have a support of an uh software developer uh around her um or any designer, I think the dividends dividends pay off really really quickly.

Claire: This makes my heart grow 10 sizes. As someone who folks don't know, I did not get a software engineering degree and yet I I'm believe I can code. Um I have been I've been coding for over 20 years. I've run very large engineering organizations and a lot of that was early in my career. Folks, shout out to like Dave and Yelland and Jeremy, like were willing to sit with me and pair with me and answer questions and tolerate where my ambition outstripped my technical ability. Now, you know, even with these AI tools, you sort of have that very patient pair programmer next to you, but I do think culturally it's very, very important for software engineering teams to extend that mentorship and guidance and feedback loop to their non-engineering partners. And this is a stress test I give to a lot of teams because I hear a lot of teams being really overwhelmed by code review and being really overwhelmed by like what may maybe they'll call like slow PRs from non-technical folks. And I just say like on average is your engineering teams PRs getting faster review than their non-engineering teams PRs? And like across the board people are like, yeah, of course they are. I'm like, that is an anti-pattern. You need to prioritize reviewing these non-engineering PRs just as high as you do your engineering ones because you can give feedback, you can create systems that improve code quality. Um a lot of those PRs are actually really good ideas. Um and so I do think there's this cultural aspect to it that I'm really pleased to hear you've you've unhooked because I think it's going to pay dividends over the long term.

Eddie: Yeah, 100% agree. And and our priority has uh always been uh PR reviews. Um I think we did an analysis and I think our median PR review time was 9 minutes.

Claire: On this team?

Eddie: On this team. Yeah. Not definitely not in like in the in the R&D org, but definitely on the team. And and the reason why is like there would always be someone on this perma-zoom. And so like you just show up in this perma-zoom and say, I have this PR ready, can we talk about it? And then someone um sometimes it's in a group setting, sometimes you just go in a breakout room with someone. We you just kind of talk through it together and and review it together.

Claire: This episode is brought to you by Jira Product Discovery. [music] AI has made individual PMs incredibly productive, but multiplayer mode is where it still breaks. Getting everyone aligned on what should actually get built. Decisions live in a markdown file from last week. The road map's a spreadsheet no one's looking at. Jira product [music] discovery is where teams actually decide what to build. Capture ideas, prioritize them as a team, and share a living road map everyone works from. It's powered by Atlassian's teamwork graph, so it can pull in customer feedback, what your team shipped, plus your goals, and suggest what to build next. And when a decision is made, you can hand it off straight to Jira, so a developer or even an agent can pick it up and start building. Teams at Canva, Deliveroo, and Toast already use Jira product discovery. Join more than 25,000 teams at atlassian.com/hia. Start building the right things together. Does this bring you back to like early founder days?

Eddie: It did feel a lot like the like startups. Actually, when we started, um also everybody coded back then. Like we hired designers at like, you know, um I mean, it's different type of code. They they wrote a lot of like HTML and CSS, and maybe some light JavaScript, but they coded, right? They they actually um uh open pull requests and and merged it. Um over time, I think that kind of went away. We said, "Oh, like everybody's got to go into these like traditional swim lanes." Um and this was like kind of almost a throwback to when we just started, right? Uh we just had maybe like a few whiteboards. We kind of like discussed in real time what we're going to build. We built it, and um yeah, and honestly, I mean, it was it was a lot of work. I I I will say that to ship something like this in 10 weeks did take a lot of like nights and and weekend time. Um but I didn't ask anybody to do it. They just did it because people were so um passionate about what we're doing, and honestly, they were having a lot of fun. It's so much fun. It was intense, but so much fun.

Claire: Uh that's that's what I hear from so many people. Um we've done so many of these I've done so many of these interviews and almost everybody says I'm working harder than I ever have and I am so much happier. I'm having so much more fun. Work is just better because the work that I'm doing is more creative. It's It feels like it's closer closer to impact. Okay. You've You've painted this vision. People are going to walk away and unfortunately their teams are going to be like, "You can ship this huge product in 10 weeks. It'll only take five people. I want nine minute per PR review review times." So you're setting the bar high. But let's like let's prove it. I mean show show us what you built and then maybe we'd love to see how you personally use Cloud Code or used Cloud Code to contribute to this product.

Eddie: Yeah, sounds great. Let me um just share my screen here and I'll I'll for the folks on audio I'll try to walk through what's on my screen. This is Gusto This is the Gusto that everybody knows, right? This is kind of uh ironically we call it classic now. Uh when you're co-founder we would never But if you're in if you're in the co-founder group like you get to call the traditional Gusto Gusto classic. This is Gusto uh co-founder here. It does the things, you know, basic stuff. I don't think any of this stuff is particularly interesting. Um it's kind of like the normal agent loop that's connected to tools. Um I would say the really interesting thing about this from a user perspective is that it comes out of the box with all of the things that Gusto is is already doing for you, right? It has all the information that Gusto already has about your business, like your employees, uh your payrolls, your schedules, uh your time off requests and and things like that. And one of the things that was really important to us was just being able to communicate it communicate with it not just through the web, but actually we want people to primarily talk to it through SMS or or Slack actually, right? So here I'm going to say something like, "Do I have any time off requests that I need to approve?" And it would be the same exact thing as if I were to like ask this in the in the web application. Um it's calling the same tools and it's going to respond to me in through the same channels. By the way, for those listening in, I'm I'm showing a screen of my phone basically, my my uh messaging app on my on my Android phone. And I just typed in, "Do I have any time off requests that I need to approve?"

Claire: What I'd imagine is this sort of like multi-channel experience is really important for kind of like small business owners in particular who are probably like running around doing stuff always operating on on their phone. Um And so I do think it's interesting as somebody who has like built B2B web apps for so long to think now my surface area is texting. It's just a very interesting product design problem. Um and then technical problem with, you know, how do you how do you show streaming or latency or make sure people understand that you are working. Um we're working on it. So I I think this whole surface area is very interesting.

Eddie: Yeah, exactly. So we have right now today Slack and SMS, but um you know, I kind of want to add like WhatsApp and Telegram. I think that's kind of one of the things I learned from using OpenClaw of like the power of like the these uh messaging channels. Okay, so it responded um actually looked up that there is someone named Todd uh who uh has a time off request. And um I could just say, "Yes." And it'll actually approve that time off request. I don't have to do that in the web app anymore. So I'll just let that run so we can come back to it to uh see that it actually approved it. But I want to show something that's a little bit more interesting and and more more complex. Uh so this actually came from a real customer. And uh this particular customer that we have is a massage spa in New New City. And what they do is they use Mindbody to track all of their work of the of their massage therapists. And they basically have, you know, how many 60-minute massages did they do, how many 90-minute massages. They also get paid like a bonus if they do certain upsells like hot stone or CBD oil. And they have like a rate for that. They also like decide how they split they decided to split their tips like as like they group their tips together and and they split it. So what this owner has to do every week is they actually, you know, they export something from Mindbody, they put it into a Google spreadsheet like this, then they actually run these calculations of like how much to pay them in terms of bonus and commission and tips. And then, and only then, they actually go into Gusto and, you know, they they go to the run payroll page and then they they actually input all that all that in, right? That part is really fast, of course, but it's all this like what I call the work before the work that a business that is not using AI has to do to run the payroll every single week. And they do this like week in and week out, right? And so in Gusto co-founder I have a set of connectors where I can Gusto co-founder can actually access third-party systems like QuickBooks, Google Sheets, Notion and things like that. And what I can do is actually just literally say my process here. So I wrote here like, "Hey Gusto co-founder, I need you to run my payroll. Look at this spreadsheet that's called export from Mindbody. And here's how I calculate it. You know, for every hot stone upsell add $15 of bonus for that therapist. For every every CBD oil add $20 bonus for that therapist. We pull tips, so just take the group tips amount and divide it by how many therapists we have." And it just goes. And it's kind of like cloud code where it'll it's going it's going to that spreadsheet, it's pulling in the data, it's running those calculations, it's updating the payroll as you see here. And then it's actually going to get to a point where it'll stop and say, "Here the amounts. Do you want me to actually submit this payroll?" And I'll say, uh, "Yes." And and it'll actually submit and and run that payroll. So, basically, you can see here that it's calculated all my payrolls, um, the hours, the bonuses, the total payroll amounts, and I'll just type in, "I'd like to submit my payroll." And it'll actually submit the payroll.

Claire: Amazing. So, and all this came out of a Vibecode prototype because you had a layover. And this is real like this is real business business data. And so, you know when I'm talking to leaders at companies, they tell me, "Oh my gosh, Claire, I am so excited about our December AI launch." And I'm like, "December? It is June. Like, what are you talking about? You need to do this today." And so, I think, you know, one of the meta stories of this is like bring forward your ambitious ideas and just executing on them quickly is really possible with not that much investment. You know, the other thing that I think about is five people for 10 weeks in an R&D organization of a thousand. Let's say at the end of the day, you ship something you're like, "Ah, I just customers don't want it." is actually not that expensive in fun outlay. Like, it feels like a lot maybe, but it's it's a very small fraction of your overall R&D investment. And the payoff can be very huge. And so, part of the advice I give people that I think you're encapsulating is you can be a lot more ambitious. Um, and you can you can you can afford a lot more risk in your product development process.

Eddie: 100% yeah. Just imagine if you had maybe 10 of those going, which even if you multiply this by 10, that's not even that's not a big investment across a company of a thousand in in R&D. You know, you're going to have like two or three of those, I think. meaningfully change the trajectory of the business.

Claire: Yeah, I love it. Well, okay, and then let's prove you said at the very beginning, well, I consider myself an engineer on the team. So, tell me a little bit about how you're using Cloud Code or maybe show show a little bit of your setup that you find useful.

Eddie: It's funny, I was looking at some user feedback um this morning and I had a feature that um I wanted to build. And so, I thought maybe we could just build it together right now.

Claire: Yeah, let's do it.

Eddie: a little overview of how how I write code, how I Cloud Code. All right. So, this came out of some user um feedback. We've talking to customers since we've launched and I actually wrote this in a GitHub issue where uh we have this issue where if Gusto co-founder says how to do something in Gusto classic, um it says, "Oh, go click on this link and that link." And that doesn't make any sense based on the page that the user is on because they're not in Gusto classic, right? So, we need to basically tell them go to Gusto classic first. Um so, I basically like copied a example um uh user conversation that happened here where it caused some confusion. So, it feels like cheating almost is um I'm going to start Cloud Code here on my terminal. Always uh you know, you can introduce

系统提示: skipping permissions.

Eddie: And I'm going to just say uh by the way, I use Whisper flow a lot. I barely type these days. So, I'm going to just say there's a customer issue that is outlined in this GitHub issue. And I'm going to paste in actually just a link to that. Can you please read this issue and come up with a fix for the problem that's outlined here? I'd love for you to first write an eval that fails to show that you can reproduce this issue. Then, come up with a solution and then prove that the solution works by uh showing that the eval now passes. And that's it. Um it's going to actually just read the whole issue. It's going to look at the conversation uh where this problem surfaced. It's going to write an eval that fails and then get that eval to pass. It's interesting. I was never actually a true like test-driven developer. Like I never was like could really get behind like writing a failing unit test first and then writing the code and getting it to pass. But when it comes to like eval stuff, like AI stuff, like it's basically kind of the only way we work now when we're trying to fix a conversation, right? So always write a eval a failing eval first, then write the code to fix it, prove that it works by seeing that the eval passes, and then seeing how the rest of the evals in your suite pass and then open up a PR. So like at this point like I like go grab a cup of coffee or or maybe I'll start a second or third work item in a in a separate cloudcode terminal and and then just wait for it to finish. Uh when it's finished, and this is honestly the part that is the most important, is actually like reviewing the code that it wrote, the eval. In this case, it's probably going to make a prompt change somewhere. Making sure that it's concise. There's like judgment on like what the prompt should look like. And and then and only then asking it to open up a pull request. Then I go to the Zoom perma-zoom and and you know, uh get someone to review this.

Claire: I I love it. So I just want to recap for everybody cuz there's so many good nuggets on how to build something in this in this new world. And so the first thing is it's okay to prototype an idea to get internal buy-in that we should get excited about something new. And you know, you did that. I think you have a lot of permission as as co-founder and leader in this organization. But I think you would extend that permission to anybody on your team, right? If you have a good idea, let's prototype it and look at it. It seems like you gathered a team around who was most excited to build on this. It's something that I sort of heard is you're like, people were leaning in and it's almost like the gravity of who was interested formed some of of the team, which I think is is quite quite fascinating. Um I heard kept the team small.

Eddie: Yep.

Claire: No dogs, no nothing. No no nothing.

Eddie: No dogs. Um Yeah, I mean there were people that wanted they like they really like what was going on and they wanted to like contribute, but we kind of like intentionally kept it as small as possible because um we just were able to move really really fast with a team of five. Now, that's not going to stay like that forever. We're starting to add more people uh to the team and more people are starting to help out. Um but I think it was critical to keep it small in the beginning.

Claire: No No one likes what I say this, but I say that my secret trick to getting things to move fast is kicking people out of projects.

背景笑声: [laughter]

Claire: I I go sign. It is an effective tactic.

Eddie: It is harsh but true. Um Yeah, and I I I do like I think so that I think there's one important um difference here that like applies to this particular project that may not apply to others, right? And it's I think it is honestly the fact that I as one of the company's co-founder was was part of this, right? So, I had like kind of the permission to you know, break all the rules that we created at Gusto, right? Mhm. If any other like teams had ever going to we're going to skip text specs and Figmas, they actually might get a slap on the on the wrist about it. Um and so how do how do I've been thinking a lot about how do you scale this to other teams? And I think um unlike me where the permission is sort of implied, like I think you actually have to go to these teams and say, we we are we want you to work in this way like where you don't do any docs, no Figmas, no text specs. We just want you to have a perma-zoom and like we're giving you permission to do it in that way. And in fact, I would even go even further and say, uh we're not giving you permission to do it in any other way. Like, if you actually produce a doc or Figma, like, you will get a slap on the wrist because we explicitly don't want you to do that, right? That's one important difference. I I I think when you don't have like a co-founder on this team, like, you actually have to be more explicit cuz I have talked to teams that want to do it in this way, but they they just don't feel like they're allowed to.

Claire: Yeah. This This reminds me of what my friend Chintan at Coinbase did is he actually he does these extreme experiments with his engineering team. One that's my favorite to reference is he's like, "Delete your IDE. Like, you are not allowed to have an IDE. Delete it."

Eddie: Yeah.

Claire: Write code. Um, and his other one that he does is he's like, "You don't touch the outputs of agent code. You only get to touch the inputs. You only get to re-prompt. You don't get to rewrite the code." It's like a very interesting just, um, tech What do they call it? Techno-psychological experiment on on your team to get people ex- get people permission to to work in a radically different way. Um, you know, that that brings me to lightning round. We'll do lightning round. We're almost out of time and I want to get you back to Claude coding, which I think is the most important thing any of us could be doing. So, my first question is, you know, coming off this conversation, do you really think docs are dead? Like, is this the way?

Eddie: I think for a subset of projects, like kind of more zero to one, I think docs are absolutely dead.

Claire: I I you know what? People don't know this about Chat PRD, but my actual intention with Chat PRD is to be the one in the product management world that legitimately kills PRDs. And And so, I'm like the anti-PRD PRD maker. So, I am all about this. I think this is truly the model. And, you know, even in my experience working with, uh, this breaking news, working with the new Fable model, is there are docs written now that are none of my business. That's what That's my new my new frame of mind is agents can write docs and they are absolutely none of my business and I will never be reading them. And that's actually a very interesting interesting model where maybe docs exist, but they're for the agents, not not for the humans. Um

Eddie: Yeah, that's awesome.

Claire: My my second question is about this like co-founder executive leader role in building teams. I hear so much or I see on LinkedIn like, "Oh my god, my CEO was like sending over Claude slop or, you know, all of a sudden my CTO thinks that he can he can um commit PRs to to main." And there's actually a lot of I think a combination of like anxiety about the shifting more hands-on role of leadership. And then sort of an leaders who have maybe in the past, depending on your culture, have been like at a little bit more at arms length with hands-on product development now are sitting with the team. And, you know, most most teams that have really healthy cultures don't actually grapple with this and so it seems like you all have a a healthy culture and it creates not so much anxiety. But for for orgs where you can imagine that anxiety exists, what perspective or advice would you give from the leadership side of that equation um that can make teams more excited to build with their co-founders, their CEOs, their CTOs,

Eddie: I think my advice to leaders is like actually get hands-on in building like production code. Don't just I think it I think it is an important first step to build a prototype and come to your team and be like, "Look, like this is actually feasible and possible, right?" And then like, you know, like that that still that still has value, right? But I do think that and I've made this mistake before in the company where I've like literally done just that. Um it can lead to a little bit of like um, under-appreciation of like the like the actual um, nuances of the work, like how much it more work it takes to actually get something into into real-life production quality like stuff, right? And so I think my advice is like don't just stop there if you're a leader, right? Actually, um, be hands-on in like merging real reviewed like high-quality code. Um, and in my case, I kind of took it to an extreme where I like went into almost IC mode for the past 10 weeks and I was like literally building I'm like 90 like fifth percentile on D DX for the past 3 months. Uh, I and just part of that was I wanted to prove that like I'm not here to just show you like prototypes and and and like tell you that you could move faster, right? I'm going to prove it by actually going and taking it all the way through.

Claire: I just I could not agree more. This is just the moment where everybody has to be hands-on and I tell a lot of executives like, "Sorry, bud. It's time for the hard skill to to show up again, not leadership, inspiration, alignment." It's like get your hands in a in a document, your hands on the code, write a campaign, build something. I think for two reasons. One, you have to be in it with your team to really understand um, how your team should be working. And two, I think it's very hard to build great AI products if you are not spending all your time using AI products right now. It's just like it's very hard to understand the primitives, the user experience, where it can solve problems, where it can't. And so a leader trying to come up with an AI AI product strategy without spending all their time practically using AI products, I think is really hamstrung.

Eddie: Yeah, I I think so one thing like that I'll disclose is that the original original idea for Gazo co-founder came from me actually just setting up openclaw myself, right? I had heard about it. Uh and I was like, "Yeah, cool. That sounds like really neat that you can like run an agent personal agent on on your own computer." Um but it really took me setting it up myself and not actually texting with it over Telegram to one like understand like the power of that like viscerally experience it. And then also like, you know, learn what's wrong with it, right? It's incredibly hard to set up. Uh you have to buy a Mac mini which you can't even get today. And that was like one of the the hypothesis of co-founder, right? Like we want it to be safe. Um you can runs in the cloud. But then also that's how we made like SMS and Slack a first-class um communication channel with it, right? And that I don't I don't think I would have gotten that insight if I didn't actually like set up an openclaw in my home myself.

Claire: I I love it. I mean, I think every leader needs to feel feel the claw um for that exact reason. I when I started using open claw, I remember I turned to my husband and like, "Oh my god, I'm having a chat GPT moment." In terms of it just changed my mental model of what product could be in a way that I just hadn't felt or experienced prior to like even since um chat GPT first came out. Even Claude code didn't like change my world in the same way that that open claw really did.

Eddie: And and you probably heard of other people talking about that that like how it's like such a crazy like innovation, right? But then you don't really you That's That's the problem like there's so much press out there and like so much hyperbole that like you kind of become numb to it. So you do kind of have to experience it yourself, I think.

Claire: Yep, completely agree. Okay, last question and then we will get you out of here. You know, when when Claude is not writing a great eval, when it's not listening, when you know you're about to have to trash a PR, what is your prompting strategy to get it back on track?

Eddie: I'm naturally a very polite person.

背景笑声: [laughter]

Eddie: Uh non-confrontational. But I and so I'm pretty nice to my Claude. I kind of like, you know, ask it nicely why did it do this way or like, could you please consider this? And actually I I for me I don't know if this is true, but I think there's a actual like practical benefit of it. I find that AI is so differential and it just kind of like defaults to doing what you want it to do, but I actually want mine to challenge me, right? And like give me different uh ways of doing things that might be better, ways of building things that might be better. Um so I have some for some reason I just feel like kind of being polite, leaving it open-ended. Uh if you think this is a good idea, could you try this? Like cuz I kind of wanted to a little bit more of like pushback.

Claire: Yep. Yep. Love it. Uh you know, most people are are very polite. I'm I'm often very polite unless it's gone real off the track. And then what I use is a 15 oh, no. I don't I'm wasting tokens, you know, it costs a drop of water, but I do it anyway. All right. Well, Eddie, this has been so fun. Where can we find you and how can we be helpful to you and the team?

Eddie: Yeah, you can find Gusto Co-Founder at gusto.com/cofounder. Um we have a waitlist, so if you're interested in it, please sign up for it. We'll give you access probably within a few days. Um and check it out. Give us feedback.

Claire: Great. I'm a I'm a very happy Gusto customer, so I'm going to ask you to to feature flag me into into the beta and I will send you all the feedback and hopefully some PRs will come out of it.

Eddie: I would love it.

Claire: Great. Well, thanks for joining, Eddie.

Eddie: Thanks, Claire.

Claire: Thanks so much for watching. If you enjoyed this show, please like and subscribe here on YouTube or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. [music] Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.