#232: Audacious Development

· 15 min

For a long time I’d have characterized myself as an ambitious developer, but with the range of tools now at my disposal, I find that I’m able to be an audacious developer.

Transcript

Lightly edited from the audio for readability: filler words and false starts removed.

00:00 Hello and welcome to Developing Perspective. My name is David Smith, and I'm an independent app developer, and I talk about what it's like to be an independent app developer on the show. And today I'm going to talk about levels of development and specifically ending up with where I find myself now. So let's get started.

00:17 So when I think about development, when I think about the work that I do, I think there are different ways that people could do the thing that I do, but in very different structures and in different contexts, it makes different amount of sense. And the way that I kind of think about it is there's different levels of how ambitious you are, how conservative you need to be, how your structure and the way in which you actually do the work that you do impacts the work that's being done. And so I kind of work through in my mind, there are four levels of this.

00:54 And the first is a really conservative development process. Something that is slow, is thoughtful, has a lot of process around it, is very methodical, maybe, in a positive way of saying this. This is the kind of development you would expect and in many ways want your bank, your government, systems like that that need to work with a tremendously high level of confidence because they are also being used in situations where if something goes wrong, it goes really wrong. I want the company that's making the software that does medical equipment or things like that to use a conservative process. It makes sense. It's something they should be doing.

01:36 Then you can kind of work your way up to what I would call standard development. So this is steady, incremental. This is maybe most companies, I would say, tend to fall into this category, or certainly established companies. And this is the kind of thing where there's some incremental experimentation, there may be some kind of a regular cadence of exploration and feature change, but things are done in a very straightforward way, that it's a regular, incremental...

02:08 This is the kind of company where if you look in their App Store changelog, it'll be every two weeks, there'll be an entry that says, "We are always trying to improve our application. This update includes performance improvements and bug fixes." That's the way they work. They have a process, it's very regimented, there's a lot of software engineering in that sense going into this, that they may be following a process. Maybe they're a Scrum or an Agile or I don't even know if those are things anymore. But there's a thing that is going on in that process, and that's a very standard way to do it.

02:39 It's very predictable. It makes a lot of sense. You make progress, right? You're not as stuck and slow as a conservative developer might be, but you're not making as much progress as you could potentially if you move to the next level, which I would say is ambitious.

02:53 And ambitious is probably where the vast majority of indie developers, myself included, would typically be. And this is where it's a development process that is much more experiment-driven, it's much more moving quickly, lots of ideas, lots of things that you're kind of exploring and working your way towards. You're trying a lot of things right when they first come out. You're trying ideas.

03:16 The reason that works well for indies is I think the thing that has allowed me to be successful to some degree is that I don't have the baggage of a standard development process. I'm not encumbered by that thing. So I can have new ideas, I can try them, explore them. When Apple comes out with some new device, some new product, some new something, I can be there on day one because I don't need to fit it necessarily into this big complicated software engineering cycle where the day the last GM of Xcode drops, I need to start off this big QA process. I can do the QA process myself and work on it that way.

03:52 And that's the majority of my life, that's where I would say I've been. I'm an ambitious developer. I do things that are fast and different, but not beyond that.

04:03 But what has been interesting recently, and I have a very specific example that I'm gonna go into great detail about this, but I feel like increasingly I am moving past ambitious into what I'm kind of considering audacious. So this is development that is not just about experimentation and moving quickly, but about tackling really big problems in really ambitious ways, things that may not work, things that may be taking on more than you could usually chew. It's like, but no idea is too big for you.

04:41 And clearly, this shift in my own work from ambitious to audacious is definitely driven by agentic development, about having access to essentially experts in particular fields (experts, not actually people, but models that can be an expert in any particular area that I am unable to do), and that allows me to tackle projects and to do things that I previously wouldn't have been able to do, that would have been just too much for me to do. Now I can actually give them a try.

05:12 And this makes me think a lot of an old episode of Debug that Guy English and Rene Ritchie did, the very first episode, in fact, of Debug, that great podcast from years ago, where they interviewed Loren Brichter. Loren Brichter, if you're not familiar, he's a sort of legendary developer. He made Tweetie, he made Letterpress, he made a bunch of just some really ambitious, interesting pieces of software that really do things in a way that was just sort of...in my mind still kind of hurts my brain.

05:47 And he was talking on episode one...I think specifically he and Guy were kind of geeking out about this, how when Loren was building Letterpress, which was a word game he made after he left Twitter, he built the entire UI in OpenGL rather than using UIKit, which would have been the obvious thing. So essentially...I think he describes it, in a few weeks, like a month and a half kind of thing, he re-implemented all of the parts of UIKit that he needed, which is incredibly audacious, which is incredibly beyond what most people would think is something they could tackle. Most people aren't Loren Brichter, aren't able to tackle that kind of an audacious thing.

06:31 And he talks about how there were things in UIKit he didn't like, there are things that he didn't want to do as a result, and so he just did it. And he was able to do that. And at the time, I remember listening to that first episode of Debug and just being completely in awe, like, wow, how cool would it be to be in a place where even the biggest ideas you have are things you could do. Because Letterpress was amazing. Legitimately incredible. The interaction, the way it worked, the frame rates, all of these things on these really limited devices were suddenly possible because Loren had built the entire UI layer in OpenGL directly himself. And I'm no Loren Brichter, but I feel like I'm getting able to do Loren Brichter kind of things now.

07:16 And so the specific story I want to talk about is how in the latest update to Check the Weather, which is my kind of map-first weather app that I launched alongside the iPhone Duo, I was getting frustrated with MapKit, which is the built-in Apple-based mapping system that is available to all apps for free, is built into the system, and generally is pretty good. It's fine. It does what you need it to do.

07:45 But I went on a trip to Scotland and I was up there on hotel Wi-Fi, and what I kept noticing, and I'll have a link to this in the show notes or it should be the chapter art, is what would happen is I had spent a huge amount of work putting my radar data onto that map in an incredibly performant way. I really, really wanted my radar data to show up as fast as possible. It didn't lag, it didn't...if you swipe around the map, it's not loading, it's doing really clever and interesting things where it's this custom data format that I'm pushing down to the device that is doing opportunistic prefetching, and there's some really...I put a lot of thought into making my radar data really good.

08:26 But I'd look at the map on my hotel Wi-Fi, and my radar data was loading instantly because I'd done all that work, but MapKit behind it was just gray squares, and they would eventually come in behind this beautiful radar data. And that annoyed me. That kind of just got in the back of my brain that I was like, oh, wow, that's annoying. I really want this app to feel so performant. I want the user to never see gray squares. I never want to see things that aren't loaded. It should all load super, super fast. It should be incredibly optimized, and it should be something that it is really quick as a result.

09:03 And that's a really hard problem. That's a really challenging thing. It's not like MapKit is badly implemented. MapKit is a very well-built system to some degree, but it has limitations. It has things built into it that limit how it can work and how it works.

09:19 And so I was like, hmm, what can I do? And so I explored. And I was like, I think in my mind, what I need is a different base map. I need a vector-based map that is written in Metal and uses similar custom data compression and things because my map never needs to get down to the street level. It's radar data. Why would I ever show that? So I can do clever optimizations if I only have to ever show data at a certain level.

09:47 And I went and built it. That took a few...it was not easy, it was a little bit of an audacious project, right? That's the whole idea. But I got there. I'll have a link in the show notes, go and download Check the Weather if you aren't already a user of it, because if you open that app and you swipe around the map, I defy you to find a gray square. It is so fast, so performant. It does things in a way that is so optimized for this.

10:14 But it's completely audacious and insane that I re-implemented my own version of MapKit for a weather app, right? This is not the core function of the app. The app is about telling you the weather. The base map behind that is not necessarily the kind of thing that you want to take on. But I feel empowered to do that because I knew what I wanted. I knew the road to get there. I just wasn't an expert enough in some of the specific areas, in some of the cartography, the geo kind of location and stuff like that. I definitely was not an expert in Metal or that kind of low-level performant rendering, but I got there.

10:52 And it works. It's amazing. Seriously, go download it if you aren't already and just try it out. You don't need a subscription or anything, it's just a free app that you can go and play with the map in. And it's amazing.

11:03 And I love that feeling. I've really enjoyed this process of developing an ultra-performant, low-level base map, and it allows me to do things I just couldn't do otherwise. Another thing, and I'll have a link to a video showing this in the show notes, but if you just go into the app, go to the appearance settings and change it to...Daylight is the name of the mode. Because I control now the base map, I can do stuff I couldn't do with MapKit. So on the map, it shows you, if you want, the progression of day and night on the map. So it goes from light mode to dark mode dynamically based on where the sun is rising or setting.

11:44 And it looks gorgeous. It is this beautiful thing where the map shifts throughout the day. In the morning, it kind of has this sunrise effect as the sun comes across. And that's essentially impossible to do in MapKit. I mean, maybe you could do some wild thing with two maps that are different things, but it would never look good. It would never be performant. And this is full 120 frames per second rendering everywhere.

12:07 And I love that. I love that my ideas are now...there's no idea that is too big. I can be as audacious as I want. I can think about these ideas and put them out into the world. And that is lovely. That is so exciting to me. And I would encourage people, if you're a developer who has big ideas, feel empowered to try them.

12:28 And one thing I will say with this, I feel like there's a lot of discourse that kind of goes around these kinds of topics where it's the sense of developers become lazy when they use agentic tools. And I'm sure there are developers who do get lazy when they are using agentic tools. But mostly, and I think most importantly, it's empowering for people who know what they want, but they don't know how to get there, right?

12:51 I'm not building something and it's like, I don't know what I want, I don't know how to get there. I'm just throwing in a three-line prompt into Claude and saying, build me this thing. In fact, it's quite the opposite. I am coming into Claude and saying, I know exactly what I want. I'm very specific. I know that this is exactly how I want it to work, this is the way that I want the data to flow. These are the things that I want. I don't know exactly how to do that. Can you do that part for me?

13:19 And there is a subtle, but I think important difference with that. It is not that it makes me lazy, it makes me empowered, right? There's things that I just couldn't physically do, but now I can. Not because...it's not doing the idea work for me, it's doing the implementation, doing the low-level Metal shaders that I just don't have the background or experience to know how to build. And that, I think, has been super powerful and exciting.

13:44 And I think it allows me to do other things too. I would not have taken on this audacious of a project if I didn't have agentic programming to do massive testing, massive validation across the entire world. I'm building things where they're rendering particular map frames all over the world and then running them through a bunch of rule-based and kind of model-based validation to look for errors and bugs, and I find places where a rendering thing didn't quite work right, where a lake overflows into the land around it because its polygon was wrong, or things like that. I'm able to do these things, these very big, audacious projects, because the tooling is there that allows me to do it.

14:22 And I know what I want. I know what is happening here, and I'm working it through every step of the way. This isn't the kind of thing...if you sat down and said, build me a Metal-based base map, you would get something that works. Absolutely. It would not work the way that I think something that is shippable, is going out to customers, is actually viable in the market. You would need more than that. And that's what I've been able to build, and it's just exciting.

14:49 I'm enjoying it, I'm having a great old time being audacious, thinking about these things. And I have so many other ideas, so many other things that I'm building now that are these audacious ideas that used to be too big and now are not, that now they're within reach, and all I have to do is be creative in the way that I come up with the ideas, creative in the way that I structure and do the work in a professional way, and the end results are things that previously I could only dream of. And that's exciting. That's fun. It's awesome to be audacious, and I hope there's more audacious things being built as a result.