AI Can Build Your App. It Can't Make It Good. | Paul Bakaus, Impeccable

26:54That's the other thing. So intentionally, if you look at the job description, it's not boring. It's actually written the way I would write it just naturally when I explain it to you. Like, hey, this is a job where you have to get shit done, where you really have to roll up your arms and just do whatever is needed to clean, make sure that we all can spend our time on the most effective use of our skill set. That was most important to me. So it's, yeah, you need to find somebody with low ego who can naturally transition along with the company. I mean, maybe that starts being, for example, right?

27:39Initially, it's maybe a bunch of investor projects, some certifications that we still need. Maybe it's like setting up some bank account or getting more employees credit cards. But then later on, it's a special partnership with some large company, or it's something that is completely different, like set up Stripe at scale. It could be anything. But I've worked with quite a lot of operations people in larger companies like Google, who would be eaten alive in that role. So that's something I will say, right? Who have focused on just one aspect of operations and they think they can do it in the real world, but they can't.

28:26I can guarantee you they can't or they're not prepared at least. Maybe they can learn how to do it. But I had to have that conversation with a few of my friends from these larger places

28:40and give them some real expectations.

28:43Paul Bakaus:The reason I ask about this is because my guess is that when you're building your first company and you were thinking about who you're going to hire, chief of staff is probably not the second hire or the first hire. I don't even know if you had a chief of staff at that first company. I just wanted to get a little bit more about the philosophy around hiring a chief of staff for an early stage startup, especially as a solo founder, because you don't have co-founders to have sort of early division of labor. And I think it might be a role that becomes more prominent earlier on at solo founded companies than maybe it would be at traditional sort of multi-founder companies.

29:23I see. Yeah, I think that's spot on. I think the reality is I need that chief of staff to protect my time. I think that's the most important goal here. I am a very reluctant CEO, I would say. I really don't want to be the type of CEO that does everything that a CEO does, like the operations bits. I never loved it. I always loved the building. Now, there are multiple ways of solving that. One is you hire the people who actually do the building for you, and I think that's what most people do. Most people, instead of hiring a chief of staff, they hire an engineer, a designer, a product manager, whatever is needed to actually get the building off your plate.

30:06So now you're getting the building off your plate and you're the chief of staff, essentially. You're the operations person. That's now your role. I think that's a trap. I think it's really silly because if you are the one who is very good at building in the first place and that's why investors invested in you and that's why you have customers, etc., why give that up? That doesn't make any sense. So I think this is probably true for a lot of solo founders right now. And so I guess my one recommendation would be don't fall in that trap. Much better, in my opinion, to hire somebody who's very good at operations, also like a chief of staff slash operations hybrid role, who can scale up with whatever operational stuff is needed and have them be the acting person that takes all that clutter off your plate so you can actually focus on building.

31:01Paul Bakaus:Super good advice. I think it's something that we will see a lot more of. Let's talk about impeccable. Let's talk about renaissance geek. I think that the first thing to say is this has clearly struck a chord with a lot of people because it's got tens of thousands of stars on GitHub for RIP Impeccable itself. Maybe share for those who are hearing about it for the first time, what Impeccable is and then why you decided to start a company around it. Yeah, Impeccable is an agent skill, I would say, first and foremost. So skill, for those who don't know, is a way of extending your coding harness, your coding agent with some additional functionality.

31:41In this case, it's one that makes your coding agent better at design. So it makes you better at design, hopefully, but also your actual AI better design so that it doesn't produce the same bad outcome that is clearly, I mean, you look at it and it's clearly made by AI. You don't really want that. You want something intentional and you want to steer it in the proper way to get to where you need to go. So that's impeccable in a nutshell. It gives you the vocabulary of design and it also comes with a visual iteration mode that allows you to generate variants on the fly together with your coding agent and then it removes AI slop.

32:19So slop is a big topic nowadays and slop is also a moving target. So like today's slop looks very different than like the slop from a year ago. And impeccable sort of does two things. It creates a really solid baseline. So like decent, good design. And that's what I meant by like lowering the floors, basically allowing more people to participate, but also raising the baseline of, you know, decent design. and then the second one is getting from good to great. That's still where a human is needed right now and inserting that human at the right times is super important. That's also something that Impeccable focuses on.

32:58So yeah, it's been a bit crazy. I will say Impeccable was something that I've built for myself because I needed it. I became really frustrated at the time working with Cloud Code because it would never quite understand my intention. I would say like, make this bolder. And then, you know, it turned everything black or whatever, like blue. And it's like, that's not what I meant. I mean, like, no, it needs to be, you know, attracting the eye of the consumer more. And you shouldn't move away from my design system. Like, it means something specific in the world of web design. And I noticed when designers would be using Cloud Code.

33:41And I am one, but, you know, I've also worked with large engineering teams. and coach them doing these contract jobs. And I noticed when designers would use these tools, they would get better outputs. And really the type of language that you use matters. And so I packaged that up and put it into a skill. For myself, first and foremost. Impeccable was, by the way, not the first open source project that I shipped earlier this year. I released a whole bunch of others. But Impeccable was the one that had the most impact on people.

34:14Paul Bakaus:Similar to OpenClaw, how Peter had released a lot of other things before that. And they all sort of fit together, but that one was the big one. Absolutely, yeah. And so it really struck a chord with people in ways that I didn't expect. And then it just developed a life of its own. It's really cool because we use Impeccable all the time. I think the thing that we like the most about it is we talked about with regard to understanding a design system versus just being good at sort of baseline design ideas and concepts. Because as you said, usually when you give a prompt to Claude Code, unless you're extremely specific, and even maybe sometimes if you are, it still can tend to sort of ignore some of the nuances.

35:02Paul Bakaus:And it will shift away from maybe the design system that you have in place to something else. And maybe that's good from a prototyping to see what this could possibly look like perspective. but it's not a good thing from a production perspective. And I've been really impressed by just how much we can take a design system from one place that we have for solo founders and be able to spin up new applications that are in the solo founders universe that have a lot of the flavor and the verve that we've developed for solo founders, but can be translated and ported over to new systems, new user interfaces.

35:42Paul Bakaus:and I know for a fact that when we try that with just something like Cloud Code or Codex, it's actually really lacking. But when you loop in impeccable, it actually works quite well. What's actually going on behind the scenes there? What was the insight that you had in terms of how to make it so that it doesn't stray and it also sort of follows these principles? Yeah, I mean it started with the vocabulary. So basically allowing you to steer in different directions but then why does it work? Why does it stay on brand? That's a good question. The first thing is it does when you initialize Impeccable, first does a product interview with you.

36:20And so it asks you about your app or your website or what you're building, right? And it's not just superficial. It doesn't just ask you what colors does your page use or whatever. And sure, it looks through the code base as well. But it also asks about who's your audience? Who is this for? Turns out if you are building, let's say like an iPad app for kids, right? It's very important for design, let's say go to a real agency, right? Very important for designer to know, is this website that I'm building for the parents of these kids, or for the kids, right? Now, if you go to lego.com, you get like a kid's website and a parent's website, they branch.

36:57And so that's critical, like in many other design tools, AI design tools, that question never pops up and so the first is a product interview the second part is design md so google the google stitch team created this new spec called design md and that holds your tokens your topography a bunch of other things about your own brand your own design system it doesn't create it's not a substitute for your whole design system but it's good enough to at least stay somewhat on brand gives information about the components in your system. And so Impeccable creates this design MD if it's not present for your own website as well and keeps it up to date over time.

37:41So anytime you actually then use a command, it's like, well, I need to redesign this case study website, for instance, or section. It will actually load both the product MD and the design MD as context. And then there's also design hooks under the hood that if it writes something and it maybe didn't follow all rules in the skill, there is a hook that fires afterwards in the coding agent to keep it honest. So for instance, it will detect if there's certain AI slop introduced, but it will also detect, for example, let's say it introduced a different font or a color that's not in your design system.

38:18It will actually flag that under the hood without you even seeing it. It's transparent. And then sometimes Claude or Codex or wherever it's supported, GitHub Copilot, will react and say, actually, I just noticed that I made a mistake here. I'm going to revert this and stay on your system, on your brand. So there's a lot of things that have to be orchestrated to keep you on that lane.

38:45Paul Bakaus:Now, oftentimes, people come across their business idea in different ways, or they start out thinking they're going to start a business, or they start out just wanting to solve this particular problem. Well, in the case of Impeccable, and that's just one of the projects from your company, Renaissance Geek, how did you actually come to determine that this was going to be a business? Was that something that was pretty intentional from the start? Or was it more you were sort of exploring this and you said, you know what? I kind of don't want to do this management thing that I'm doing at a larger company.

39:21Paul Bakaus:I kind of want to go and start a company. So how did that actually come to be? Yeah, I think two things happened at the same time. The first one is with my partner, Anish, at A16Z, who I'm very lucky to be friends with. I went on a bunch of walks with them in the last couple of months, and I used them as a sounding board to pitch my stupid, hopefully some of them stupid, some of them not so stupid ideas to. And at some point, I think we both got to the conclusion that there's probably enough there to even if like two thirds of them don't work out, I'll figure it out. So that's the first aspect.

40:03That's beyond impeccable. And so as a solo founder and as a founder in general, you're always going to have some pivot of some sorts. I mean, it's so rare to never pivot, right? I think that was the first important realization that if I'm now creating a company, there's definitely enough there in my own arsenal of ideas that one of them will probably work. That's what I had to tell myself before starting. That was really important. That's the first one. The second one was with Impeccable. Almost every week I get somebody tweeting at me saying something along the lines of I don't understand why this is free.

40:45Like why is this free? Or like I want to pay for this. How can I pay for it? And that's a really great PMF signal. I mean, that's a signal that sometimes I would create something like a failed project and I would hope for this. I would hope for this so much for this to come. And I was convinced that it was a good idea, but I never quite got there. People would always be very nice and on a call when I were on a customer call with them, it's like, yeah, I've tried it out. It's pretty good. But you never quite got that response. And so that's happening organically now. I'm getting these tweets all the time.

41:22And that gives me the sense that there is a commercial story there. There's something that I can do here. So that was the other really important factor.

41:33Paul Bakaus:It's really interesting to release something that's just so immensely valuable for free. It clearly gains you a lot of mindshare, a lot of love. What's sort of the logic behind it? Or did you originally know that the first thing you'd be putting out for people with the company would be free? And how do you think that that maybe extends to other people who might be building things in the future? So how they should think about putting things out? There's all sorts of different opinions about focus on growth and focus on sort of distribution first versus focusing so clearly on monetization that other people say, you got to have people paying you immediately.

42:14Paul Bakaus:And I'm curious sort of what you've learned from all this and how you think that that might be something that some people can take away a lesson from. Yeah, I mean, I think maybe I'll do a hot take here, but I think sort of the PLG stuff, like product-led growth, I mean, it can work, but I do think in today's age where everybody, I mean, it's an explosion of new vibe-coded products. Everybody's building something. So many people are building something for the first time and it's exciting, but it also means it's so much harder to break through with an initial audience. I think if you have no distribution whatsoever, no followers, no outlet to at least tell a few people about it, it's super, super challenging right now.

42:57And I definitely felt that a couple months ago when I embarked on this journey again because my X account, I guess Twitter account, I don't know, I used it very actively in my jQuery days. And I was known for a specific thing. And that was sort of like UI slash front-end development. And back then, I was active. I tweeted almost every day. And it was a good social media platform for me. Then I completely neglected it for a couple of years. Then I started using it a little bit in the creator economy jobs that I had. But then people followed me for creator economy stuff. So then, a couple months ago, I start tweeting again.

43:41And I tweet about this new stuff. I tweet about AI and design and generative user interfaces and zero interest whatsoever by the algorithm, right? Zero. And I think these tweets are all certified bangers and like nobody cares, right? And it's so frustrating. And so I realized, okay, well, this doesn't, it's not going to work. Like I can put anything out there, like the best thing since, I don't know, like sliced bread, right? But if it doesn't find an audience, I was never really quite the guy who would put on Facebook ads and hope for the best. I always needed some sort of organic spark that would work.

44:21And so the first thing I did was to think through, okay, well, how do I optimize for momentum in the first couple of months and put myself on the map again? That was honestly really important to me. because it clearly didn't work anymore. The people who followed me from before, I still had a bunch of thousand followers, right? But the one that followed me, followed me for different reasons and the algorithm didn't pick it up. And so the first I did was realizing, okay, the X algorithm now rewards value. So I'm just gonna create the most valuable thing and put it out there that I can for free to see if anybody cares.

45:02And so I've done this with a bunch of projects now. Some of them worked, some of them didn't but that really was a good strategy for me because then i gained like you know a thousand followers here a thousand followers here and then at some point the algorithm changed overnight and picked up the signal and realized okay well he's clearly in it for the right reasons now with the right audience and now we will reward him or something along those lines but that's what

45:29Paul Bakaus:i tell myself it also feels like nowadays with so many people making so many things so quickly the faster you can lower the barrier, the more you could lower the barrier to somebody actually trying your thing and experiencing it and then telling other people about it. It seems like that is just generally quite beneficial at this point. Yeah, I was also always obsessed about this K-factor thing, right? True virality. And I think the most important thing is what happens, like how do people talk about your product when you're not in the room, right? So you have two people who have tried out your product and now they have a conversation about your product.

46:06That's the really interesting stuff. And that's where the magic happens or not. If they're like, no, I tried, it didn't really work. Okay, dead. But if they're like, you know what? I told five other people because this is the best thing. Yes, you have something. And I'm hearing that more and more often now about impeccable and makes me feel like I'm onto something.

46:28Paul Bakaus:What did you learn about virality and K-Factor at Zynga? You were there for a little while. Yeah, I mean, with Zynga, that was pretty much the only thing that mattered. We tried to create these games at hyperscale. And I think anything, if I remember correctly, anything that wouldn't hit at least 100 million, like DAU or something like that, or MAU, was a failure, right? I mean, it's insane if you think about it now. But that was the times where Facebook really rewarded games. and there were so many ways of gaming the profiles of people where, you know, Farmville is a great example, right? Anytime you did some action, you click on like, tell all my friends about it or something, right?

47:15And then it would post on your profile. And so you have this massive influx of profile posts that all lead back into the games. And it was very easy to gamify variety at the time. Now, that stopped though. It's not as easy anymore. I think it was really great for a while and then it became much harder again. And then there's also this thing with the whales. So I don't know if you've read about this, right? But in many of the Zynga games, there were people who were extremely active, the whales who spent a significant amount of money and they kind of carried the whole ecosystem. And so you had virality on one end and then you had these paying moguls on the other side.

48:00And it was very lucrative for a while.

48:01Paul Bakaus:How do you juxtapose sort of that kind of very consumer-oriented K-factor stuff with something that has some, I'm sure there are a lot of people who are using impeccable and non-professional ways just for passion projects and stuff like that. But how do you think about virality when it comes to more business-oriented? Yeah, I mean, so I think there's this huge difference between forced virality, and I think that's something that we did at Zynga. and I was never actually I was rarely building these games directly I built one or two games but mostly I was responsible for the technology behind the scenes but talking to those game managers who were in charge of these game studios definitely it was like it was not something that would come organically it would really make sure that that comes at all cost I mean same thing with Facebook engagement I mean, they just like not just Zynga on Facebook, but Facebook itself.

49:01That's right. Yeah. But for my business now, I'm specifically not trying to force it. I think there's no good way of forcing it in my case. I think either people find it valuable or they don't find it valuable. The only thing I can optimize for is the individual experience and then hope for the best. so it's it's i haven't done any uh outbound for example yeah like no paid ads no nothing like the

49:30Paul Bakaus:only the only thing that i've done is tweets no ugc no trying to get influencers to post about it that will some of that will happen just by nature of it being a great you know a great product yeah but it has happened exactly right i mean look fireship has done a video about impeccable or at stick you know a slice of a video about impeccable and that was completely organic and people ask me afterwards was this a paid thing no that's always a good sign when people are asking you wow that was incredible was that paid yeah it's like it's like no no they actually just really like that but i also i will say i have seen some businesses and i have some friends who are running businesses that don't work without that growth loop yeah and so you have a product you have a front door and you have to make sure there's always enough Facebook ads or always enough TikTok influencers talking about it, right?

50:20And so people come into the front door and then there's massive churn and you just get more people and fix that problem that way. That just isn't very interesting to me. I don't find that idea very pleasant. I would much rather have a really solid customer base and core who loves what I'm building than a lot of people who maybe like vibe code something with it or one shot something and then just move away again. It's just not interesting.

50:46Paul Bakaus:It also feels very aligned with how you want to prioritize your time. When you're talking about having people join the team to work on certain things while you focus on others, that seems to really be in line with it. It kind of goes back to this idea that I'm constantly thinking about, which is the shape of the business. There are lots of different business shapes and some of them can be great for one type of founder and really awful for another type of founder. I'm curious as you sort of discovered who you are and who you're becoming, what you've learned about sort of your personal relationship to the shape of a business.

51:27I think the shape of the business, but also like personal relationship to customers and to users maybe. One thing that I learned is that integrity is probably more important to me than to many others that I work with. For example, it would be devastating to me to have somebody say like, I've used your thing and you advertised something that didn't work. I would feel like a total failure at that point because then I felt like a snake oil salesman. And I never wanted that. I always wanted to be respected for my craft. and for what I put out there. And so my own sense of quality and perfection is very, very important to me.

52:14And the same is true for team building and for creating a business. I think it's very important for me. For instance, one of the sayings that I have is be nice, but be kind, not nice.

52:35and I'm trying to be kind by being not nice sometimes and I think that's a really good way to treat your co-workers with respect by giving them real feedback as opposed to just nice things that are actually not true or not meant. So I feel like, especially in the world of AI right now, there's already so much snake oil. Like skills are a great example. Most skills, I'm sure, are completely vibe coded. You created it maybe for your own project. You test it two or three times and then you ship it. And you're like, okay, here you go. And then you send it to others. But this is like building your own software, right?

53:15Compiling it on your own computer. And as long as it works on your computer, you think it works on every other computer. That's not how you distribute software, right? You have to actually test it. You have to make sure it works everywhere else. many people don't know this but impeccable every single line of the impeccable skill is tested I mean in an evals harness to actually figure out does it do something I went through great lengths and I spent a lot of tokens and money to make sure that everything is ablation tested you know if there's like an AI slot rule and it doesn't work I actually removed it from the skill because it's so important for me and still people will tell me or question and say like hey I don't know I installed the skill and then I asked to do something but maybe I also can do it without the skill and so at a conference a couple days ago I showed a before and after with GPT 5.5 and one of it was using impeccable and saying like you know make this case study section bolder and one of them was saying okay same thing same prompt, same everything, but vanilla.

54:28And the differences were pretty clear.

54:30Paul Bakaus:I think that we've covered a lot of interesting ground with regard to your philosophy, talking about things like how to focus your time as a solo founder, essentially lowering the floor for accessibility to great technology and abilities to actually make things and put them out of the world, while also increasing the sort of what's actually possible in terms of what the average like builder is capable of. We've talked about sort of how you've learned that over the years at a variety of companies. I'm really curious about your philosophy as it relates to sort of humans and like humans being in the loop for these creative practices.

55:15Paul Bakaus:Right now, you know, it is really necessary for a human to be involved on that maybe last 20%, last 10%. But I think that in a lot of ways, it seems like your philosophy is around human involvement sort of perpetually. And the idea that humans are essentially having a challenge with regard to kind of being able to pass their intent successfully to the AI to actually have the AI execute based off of what they really want. Can you share a little bit more about your philosophy around humans interacting with AI? And maybe we can go a little bit onto interface side of that, but also just more of the general prescription around what the real problem is with humans and AI right now.

56:06Yeah, absolutely. So with humans and AI, there's two conversations right now happening. and I just came back from a conference where a whole bunch of people were talking about loop maxing and software factories and some people call it orchestras. But the idea is the same. You create agents, these agents hand work to other agents and at some point you have agents working with agents working with agents and you might still say, here's my goal, but then all the work gets executed in days and days of autonomous, agentic work. That's one idea that people are pushing. Another idea is people are saying, no, don't touch AI at all.

56:50Or when you're using AI, do it piece by piece. You're just in one thread and then you observe what the agent actually creates. You look at all the code that gets created. You read every line. You validate everything. That's the other extreme. I think the reality is actually somewhere in the middle. and when I say like put the humans in the loop, I actually think when you put them in the loop matters. In my opinion, for instance, for creative design work and most creative work, I think there's ultimately three big phases and the reality is a bit more messy. You can go back and forth, but usually you have the early stage, sort of the what phase, right?

57:32What am I even making? Like you're making a website. What is even on that website? Who is it for? So this is more like product. At that point, I think it's pretty important to be in a loop. If you don't know why you need this, why do you make it? Why do you tell an agent to make it in the first place? Why does the agent determine to make it? That's one, right? Now the other one is the actual building. Now the building aspect, I think right now agents are still very easy and capable. Agents are still very much drifting from an initial vision. And they do that sometimes randomly. So sometimes an agent will work on something for an hour and everything looks great and then shoots in a complete wrong direction.

58:19So without you monitoring and checking in from time to time, you might get terrible results. That might be a temporary problem. But anyway, it happens. And then there's the last part. So let's say the agent comes back to you, produce pretty adequate website first draft. Now, how do you get it to 100 %? I think for those last 10, 20%, you need to actively iterate with the agent. You need to put in your own point of view, your own judgment, your own perspective. One example is most agentic work today, most code that's being written is actually too much code. Most articles that are being written are too long.

59:03Most design that's being created with AI is too verbose or too cluttered. So the most important skill right now where I think a human is needed is to remove things. Say like what not to do. No, no, no, actually you have to remove all of this part. Or this you have to simplify or that. And so that type of judgment is super, super critical at the current moment in time. And I also think there's kind of like a philosophical argument here to be made and then there's a practical argument. I mean, I've talked to a whole bunch of my friends at the Frontier Labs about this, and I would say overall, most people think, at the labs at least, that the models will soon be good enough to build something adequate, something pretty good.

59:51I also designed something pretty good. But the question is, in the philosophical argument, if you have agents creating things for other agents in a world made for agents, what will we do? How do we come in? What gives us purpose? Who is this all for? Where is this all going? I don't know. So there's definitely that. I think about that all the time. I actually would really like to ensure that these agents are working for humanity, not against humanity. But then there's also the practical aspect of, I think at least for the next couple of years, AI will simply not be good enough to create great creative output.

1:00:34I truly believe that. I think we're getting to like decent in some categories, but I haven't seen anything one-shotted. Not video, not images, definitely not anything written. I mean, I tried to do a first draft of my announcement post and it was so bad that I had to basically reword everything. Definitely not code and also not design. So none of these things get you to 100 % right now. And this is why I'm sometimes quite frustrated with this narrative of AI design tools and other types of tools positioned for Vibe coders to be able to make a problem go away. Just use this and you'll have a landing page.

1:01:22You just put a prompt and you get something good. That doesn't work. It has never worked so far. I have never seen anyone nail that. I think design is an iterative process and has to have a point of view of a human.

1:01:35Paul Bakaus:It feels like, in some ways, pretty good is always what a lot of people are striving for. I mean, you know, and pretty good just continues to get a little bit better, but it's incremental. I mean, look, like Squarespace, when it came out, it looked better than the other options that existed before it for the average person to set up, you know maybe a website for a business um they marketed the heck out of it they spent so much money on advertising on every podcast imaginable and you know they they did build a business off of that um but arguably i would say that it was pretty good and then also um not not impeccable um and then also uh once squarespace hit a certain saturation point of adoption everything started to look like squarespace right and similar to what you're talking about how slop is a moving target uh i remember when twitter bootstrap came out which was sort of their open source library for uh for web design like web pages um everything started to look like twitter bootstrap people would constantly complain about oh it's just another bootstrap uh themed website and you know now it's with claude and and and various various things kind of just like kind of look they're pretty good, but there's a big gap between pretty good and something that's truly amazing, that feels one of one, that feels artisan.

1:02:59Paul Bakaus:I know you've written about sort of that idea of artisan versus factory made. I think it's really important. Another thing that you touched on around sort of this idea that more is not necessarily better, you know, more words, more writing, more code, like the whole idea that you know a book most books should be blog posts most blog posts should be tweets most tweets shouldn't even exist you know there there is there is so much to be said for there's a lot of bloat out there and there's also a lot of pretty good and your goal is to essentially say to people hey if you're not just looking for pretty good we're going to help you with pretty good but pretty good is going to be even better but if you actually want to do go from pretty good to outstanding to one of one to singular uh that comes from being in the loop and we're trying to develop systems that really help you be in the loop yeah absolutely and by the way like on the on the on the thing about like removing right as opposed to adding there's this common misconception misconception that steve jobs used to be this like super creative genus but I don't think that's what made him so unique I think what made him so unique is his judgment he was just the greatest editor of all time he would just say no to 99 % of things and the 1 % that he'd say yes to made it to us the stuff that we got into our hands a lot of the work was created by other people but he just had the capability, the uncanny ability to know what not to ship.

1:04:38And I think that's just so much more important now. It's kind of crazy.

1:04:43Paul Bakaus:It's also interesting because I was just reading an excerpt from the Walter Isaacson biography of Steve Jobs. It was talking about when he was going through his medical treatments, he was complaining about the designs of the face masks. And I think it was maybe one of the machines to check your pulse or something. And he was talking about how they just weren't very good. and he wanted to see better design and he would design a better one. It's like the guy was just obsessed with elimination and with really like cutting down to the core of what it is. Another thing that comes to mind from all of that is it feels like in many ways people maybe take away the wrong lessons from people like that, which is that it's always about subtraction or it's always about minimalism or something like that.

1:05:36Paul Bakaus:I think that really what it comes down to is everybody has sort of their perspective, their thing that they put on the world. And when things don't stand out, it's often because the people are just kind of like doing whatever is on trend or what is normal. Or in this case, it's whatever the system is sort of set up as the default to do. Yeah, but it's also interesting that it's not necessarily a bad thing. It depends on what you're doing. For example, if you're building an internal tool or if you're building a system level application, use system fonts. Use fonts that are already in the system.

1:06:15Make people feel at home. If you're building something that would benefit from an enterprise customer already being able to orient themselves, copy an existing UI. Make it work like something they already used to. Great. Amazing. Because enterprise is a good example. These things are not sold by people looking at websites. These people are sold to CTOs. These things are sold to CTOs. So that's a very, very different problem than you're building something in the consumer space and you have five seconds of time to convince somebody arriving on your landing page they clicked on from Instagram that something is worth your time.

1:06:52Hugely different. And now in a sea of vibe-coded sites, sticking out is actually really, really important. So I think depending on the type of use case, you just need a completely different approach to design, which is also interesting.

1:07:06Paul Bakaus:One of my favorite things that I heard about editing, specifically when it comes to teammates, is Keith Raboy. He has this concept of when you're working with someone, you want to essentially expect that there will be times to take out the red marker and edit and put line edits and really be thorough there and correct things and give people feedback. But the idea with management and editing is that the more time goes on, you want to use the red marker less and less. I'm curious how you think, one, how that resonates with you. and then two how you think that relates to working with AI versus working with people yeah I think that's completely true and I think with AI I think the same is true I think the there was this cool like gif that somebody created or like some some meme graphic forgot who did it but it was about the process of getting to results right and so usually I have like this you build this like I don't know like parts of a car first and then you kind of built up on it until you have the final car right and so now in that picture you saw this fancy alien car first with all belts and whatever like you know like crazy extrusions right and it's blue and whatever you know and then you kind of like trim it down until it looks like a real car right i think that's certainly true and to your point if you're doing a first edit for a novel there's usually way too much in it there's usually like you're meandering in all sorts of directions there's continuity issues.

1:08:44And the same is true for design work. You're building a large landing page and you first describe what you want. I think any AI model right now, GPT may be a bit lesser than Claude, will first try to put everything on there that you might need. Not that you will need, but that you might need. And one common principle to actually counter that is to say, afterwards, you're going to cut until it hurts. You will remove more than you intended to remove because you can always bring things back but there's usually too much stuff. And then you kind of do the squint test and say, does this still work? But it is a challenging process especially because you so quickly feel ownership over something that you created even if it was created with AI.

1:09:33You create something and then you have this beautiful case study section maybe, or like this logo section that has logos of companies. And you get attached to it because it looks cool. Maybe it's a happy little accident that the AI model generated. But then to have the discipline to say, I'm going to remove this again, is sometimes really hard. So that's tricky. But I do think it's very similar. So the editing process of getting to multiple iterations until there are no red markers anymore, red lines, it's very much the same thing and in fact it's exactly the same thing if you're using impeccable because there's this design linter that runs behind the scenes in design hooks so when it edits your code and the linter runs under the hood and for instance finds 30 issues the model can then fix 20 of them so it produces 10 issues and so it continues polishing your work until it finds at least no reported issues that doesn't mean the design is great, but there's a process there that feels very similar.

1:10:38I actually called this, I was working on this other library called Radiant Shaders, and this was an open source project that didn't blow up as big as Impeccable, but it has some happy users, and it's a collection of shaders. So animated shaders that you could use on your website, and I documented my process of how I got there, and I called it Carving Shaders Out of Cloud code because I really like in the process of working with an agent as like carving something out of a stone block almost right and I think it's an interesting way to think about it it's everything is already in the model right but then you have to like shape it this is really

1:11:23Paul Bakaus:interesting because I I've I've talked about startups in a lot of ways as chipping away from a block of marble versus starting from a blank canvas because I think oftentimes startups come from revealing something that is already there and I just remember this quote or this idea that that sculptors are not necessarily starting from nothing they're actually trying to trying to free the object from the block of marble yeah and i think about startups is they're actually trying to free that what they're making from the marble it's they're trying to pull it out of there and in many ways what you're talking about with you know with an llm is there is this block of marble um or this block of information and what you're trying to do is you're trying to pull that out um and so you actually reveal it and and in many ways um you need to adjust as you go it's not it's not that you're going to have it come out perfectly um but you're going to adjust as you go and and start to see something reveal itself and say that looks good there but maybe not over there and i think that that's um i think it's a really powerful analogy for for working with ai i think so and it's also how you navigate the latent space in a model that's the other interesting thing.

1:12:46For example, there's this common saying that LLMs are not funny, that they don't have humor. But that's simply not true. They just have 500 ,000 definitions of humor. There are so many different funnies in the LLM, but they probably don't understand what you find funny. And so for you to figure out, okay, well, how do I create this camera that's like, you know, hole to pinpoint light though right that focuses on the exact definition of funny that you find funny that's an art form in itself but it's very similar right like navigating the latent space

1:13:25Paul Bakaus:yeah it's a really interesting point because pretty much anybody's definition of anything is fairly uh fairly specific to them uh and their preferences and their experiences where they grew up. I mean, just think about spicy food, right? There's different types of spicy. And some people can't handle spicy food at all. Some people love spicy food. It's just this entirely subjective, well, they try and have objectivity to it around these different units of measure. But I think that a lot of people would describe something as subjectively spicy in their own way, depending on where they're from or what they like completely i'd love to love to maybe close on you know the kind of customary questions we ask which at the end of conversations this has been super fun um you know the first question is uh sort of the bear case for being a solo founder like why you shouldn't be a solo founder we'll talk about that for a second and then we'll transition over to what the what the case for solo founding is but what's the what's the reason that somebody should maybe you consider not being a solo founder at this moment?

1:14:30Well, I think there's a saying that it takes a certain type of crazy to found something in the first place, right? But I think not knowing whether you're actually crazy or not, not having a sounding board can be really, really hard sometimes. So I think if you're a solo founder, if you're going on that route yourself, which is great, at least find some people that can become your sounding board. That can be your friends and family, but better to have somebody who's actually objective. So I would say that's probably the biggest one. And then protect your attention. Protect the thing that you are truly capable of doing.

1:15:08You might be able to do that in another way by contracting things out, etc. But I think it's a bit harder to do. I think it's possible but it's harder. Now it's also liberating because you can move faster. You don't have to negotiate with others. but those things I would say get harder.

1:15:25Paul Bakaus:When you talk about, and you touched on this a little bit about doing the things that you like and sort of focusing on them and having potentially other people who can help with the other things even if you don't have a co-founder, maybe you share just a tiny bit more about that, focusing on the right things because I think that it can be very easy to focus on the things that you like maybe sometimes to the extent that you let the things that are important sort of fall by the wayside one way to do that to avoid that is to potentially hire a chief of staff early on but you have to juggle a lot of things as a solo founder is there anything else that you've learned in this process before you brought on the chief of staff about sort of being able to focus but also getting all the other things done as well yeah i think one advice that i would give is look at your biology.

1:16:14And what I mean by that is, you know, look at your decision fatigue, for instance. In my case, the way you organize your work through the day is really important. In my case, during the mornings, I have less decision fatigue. So I do more critical meetings with people. But then in the afternoons, I'm more creative, but I have decision fatigue. So now I start billing. So for everybody, that's different. Everybody operates differently. But like, really listen to your body and what it tells you. That's one advice I would give. I think the other one is focus on creating value. We talked a little bit about this earlier.

1:16:51But so many days during my corporate jobs, I woke up in the morning. Actually, no, I came home in the evening and I thought about, well, what value did I create today? For the world, for anybody. And I didn't have an answer. And I think that's a terrible feeling. That actually leads to burnout. and it's so important to a solo founder as well. What is the value that you create on every single day? And I think it's also the key to building a small company. I think people build large companies because they couldn't build a small company instead. And so if you want to build a small company that's really effective, make sure that everybody you hire creates value every single day.

1:17:35you create that structure without middle management. So you make that true. And you have to focus on what it is that you put out there that's valuable.

1:17:47Paul Bakaus:That's a great point. The other thing about burnout, which is similar to this idea of make sure you're creating value, is that I think that one of the reasons that people experience burnout is due to a lack of momentum. Like momentum will solve for burnout very quickly. It's very energizing to have momentum. versus burnout, you're usually not very energized because you're trying to push something. And I think that if you focus on creating value, that's sort of the early sort of leading indicator for creating momentum. Yeah, momentum, I will say, I mean, it brings up a good point because there are so many things that can kill momentum.

1:18:27I mean, one thing that can kill momentum is hiring somebody into a role that cannot come in at the right elevation. But somebody who's really bad at zero to one, but who has been very good at five to 10. Super, super easy to kill momentum that way. First thing the person wants to do is create a monorepo or introduce a ticketing system or whatever it is, right? But things that worked in a large team at some big Fortune 50 company, but certainly not the right thing for an early stage startup, right? So I think that's the other thing. Figure out what feeds you, what gives you energy, what gives you momentum versus what stops your momentum, what stops you and attracts.

1:19:09There are so many things we learned about software engineering that are simply not true anymore. That's the other thing, right? The handoff process, the agile workflow, like going from design to engineering, like product design engineering, none of that stuff matters anymore. It's really interesting, like, because it cannot matter anymore in a workflow where we ship 50 ,000 lines of code every day. It's just not possible. So I'm seeing designers move into code, I'm moving engineers moving into design. I see product managers moving to everything. And we just have to work completely differently.

1:19:43So I think there's a lot of things that we assume are needed that are not actually needed in order to build something valuable. So that's one way I protect my momentum.

1:19:56Paul Bakaus:Cool. Final question. The case for solo founding. What is the case for solo founding? The case for solo founding is you only have to account to yourself. And you can move at the speed of your own thought and don't have to wait for anyone. I think it's especially important if you're in the early stages of exploration. And you also don't really need anyone when you're first building something that you envisioned in the first place. now unless you need help from outside that is a big operation in the first place but I think don't let it stop you from starting something I think it's a big mistake to wait for somebody come around the door, the right person to join you start it well this has been awesome, I'm really excited to maybe in a few months or a year come back and talk about what the the Renaissance geeks are up to, uh, in the future because I, I mean, it's incredible right now.

1:21:01Paul Bakaus:We love impeccable. We use it all the time and yeah, thanks for doing this. Thank you so much for having me. If you enjoyed this conversation, we would love it. If you could share solo founders with everyone, you know, particularly people who are considering starting companies solo or who are already doing it, you know, the entire goal of what we do is to normalize solo founding, encourage more people to solo found instead of end up with co-founders of convenience, if you'd like, and we'd really appreciate it. You can go leave a review on Apple. You can go give us a thumbs up. You can subscribe on the YouTube channel.

1:21:30Anything helps. And of course, if you are starting a company or considering starting a company, you should go look at the Solo Founders program. We work together with you in San Francisco alongside of a bunch of other Solo Founders. The idea is it's much better to be Solo together than to be Solo alone. I'll spend a lot of time helping you. And we also invest$100 ,000 in your company. So if you're excited to learn more about that, go check out solofounders.com slash program.