I recently spoke with Cyrus Freshman, developer of web apps such as Minnow, the Indie Internet Index, and Lettercomb (all of which deserve to be checked out!) At the time of the interview, Cyrus was working independently; he now works for Wayfinder Collective, building a sales operations platform. The interview has been lightly edited for clarity.

Micah Blachman: What drove you to create independent projects at first?

Cyrus Freshman: I was working at Amazon Robotics on the path-planning team doing algorithmic cloud software things. I was there for about two and a half years, and I had saved up some money. Amazon Robotics was already functioning, I wasn't adding that much, and it felt like a good time to try to build out as much of my own software as I could. When I left, I had one app, a word strategy game called Wordbase. It had been discontinued, I remade it for the web, and it’s still one of the most popular things on my website. People sign up every couple of days for Wordbase. That was all I had when I left.

Another one of my games, Lettercomb, was also something that had been discontinued, and I remade it. Then, they actually re-released it, so I reached out, and they said it was okay, but that I just can’t use the real name—which I had been doing—so I changed it. Both of those games were ones that I played in high school and really enjoyed. Even those massive apps were going offline or getting discontinued, because it's just so hard to monetize.

MB: What do you think is the overarching theme behind all of your apps?

CF: I’m trying to rebuild the internet. With Instagram, Snapchat, TikTok, when you’re on your phone, you go to an app, which is killing the internet. I've been trying to build web apps that work both on mobile and desktop, that sort of rebuild the internet in a way.

I’m just trying to use the internet as it was intended. I have a Twitter-type microblogging social network, and a media upload service, and I don't let you upload things into the microblogging app. I make you put your uploads in the media uploader that I made, and then you link to it. The internet is mostly about linking to things. When you upload your photos to Instagram, you don't actually get a link for them. They're storing your image, but you can't link to it from anywhere else. You can't link to it on your website. They've kind of held your content hostage, so I'm just trying to build things that I feel like are missing. I try to spread out what I’m building into different areas.

MB: Given that you’re trying to make the web more open, how do you feel about making your apps open-source versus keeping them closed-source?

CF: At one point, I was making them all open-source, and I just closed them, because I'm a solo developer. Security-wise, it's just adding more surface area for attacks. If you're a team of people, and you have an app, it would be ideal if it's open-source. If you're one person, and you're building a coding library, open-source that. It just started to feel too sketchy—I'm sure I have some security hole in some app, and I don't want the source code to be out there to be so easily findable. I have been going through and fixing things up now that I can use AI coding tools—I was doing the development all before AI coding tools. I have gone back and asked, “Are there any security flaws with this?”

I discovered that I was building my link preview tool too simply, and that there was a flaw in it. The AI coding tools are still a little oversensitive, but I'm trying to do as best as I can with security. I did get a degree in computer science, so I have a little bit of security training. My biggest worry with open-source is that it’s opening myself up in a way that it doesn’t have to be.

I have some small libraries that are open source, and I maybe have a couple apps that are still open source. No one is opening pull requests on them, and that’s really the benefit of open-source code. The biggest interactions I get on GitHub are for the Wordle word lists I published. People will comment saying it isn’t up to date, and I’ll say that The New York Times changed it.

Ideally, there is a good open-source community, some apps aren't the best fit for that. For me, making open-source changes is something that is hard to get into because there's such a high bar when you're making a proposal, in my mind—or maybe it's easier than I think it is. I’ve always been more interested in making my own apps.

MB: How did you get into creating web apps?

CF: Web apps are really accessible. You don't need an Apple developer account, you don't need to go through App Store review processes, You can very quickly come up with an idea and publish it. I find that really ideal when I'm publishing an app so I've just continued to make web apps. I have made some iOS or Android apps, I'm mostly just making web apps.

MB: How did you build Minnow?

CF: I used to write apps myself. These days, I mostly use AI coding tools, because I can make an app ten times as quickly, and it's two or three times as good. I typically start out with a master rule set, a document that I paste into ChatGPT, which I’m using these days. I have used Cursor in the past, but it just gets really expensive. If you are spending the whole day coding in Cursor, you can spend $100 a day, which is too much. I use ChatGPT, I paste this document in that basically tells it how to bootstrap the project. I then test the basic functionality of the app, see if it’s a good start, and go from there.

MB: What was your inspiration?

CF: Nowadays especially, everyone should have a personal website. You have a computer, and you're not really fully utilizing it if you're not coding with it, and getting what you want out into the world, via the internet—if there is anything, I know, some people don't want to do that at all. 

For Minnow, the idea was, especially with LLMs that can code for you, it's really easy to make a personal website now. The only thing is that while ChatGPT or other LLMs will make you a website, it won’t publish it for you. Minnow puts those elements together: it codes the website, and publishes it on a subdomain of the main site.

MB: Where do you see the trade-offs of AI coding?

CF: If you're in a corporation, and you’re making very small, incremental changes, then hand-coding is great. Even then, you probably still want to be using AI coding tools to get ideas for how to code stuff, and then still go and hand-code it. But if you're a solo developer, and you're building web apps, which aren't mission-critical systems, if you have pushback against AI coding at this point, you're maybe not really seeing it for what it is. Even if you don't want it writing all the code for you, it's still useful for thinking through how you'll build stuff—it’s an incredible tool.

MB: From the user’s point of view, do you think there’s a difference between a hand-coded and an AI-coded app?

CF: It's more about how much effort the developer put into the tool. If someone spent one hour hand-coding versus one hour AI coding, they're probably going to get further with the AI coding app. It'll probably still be better than a one-hour hand-coded app, but it's still not going to be as good as if they put a week into the app. It’s about how much craft and skill the developer has, so, ideally, you shouldn’t be able to tell.

MB: How do you feel it is to try to get your name out there and grow your user base with so much AI-generated software out there?

CF: That's where I'm weakest. I feel like my apps are pretty good, but I still don't really know how to get my apps or name out there? I've gotten around 200K views on my site, but that’s translated to maybe 50 to 100 persistent users. I'm not currently planning to continue doing this full-time, I'm trying to go back to a company. It's still great, I'm still really happy that I spent this time making these apps, just in case they could grow while I'm doing other things. Minnow got three likes on Hacker News, but two weeks prior, I had built the Indie Internet Index, and that got to number 7 on Hacker News. The III is a directory of user-submitted websites with descriptions, and then semantic search over those descriptions to provide a cool indie internet search engine.

It's very hard to get upvotes. I got upvotes because users submit their own websites, so everyone wanted to submit their website, and then they would upvote the posts. It’s hard to engineer a mechanic like that, but it worked.

MB: How does the software you use inspire the software you make?

CF: There are a couple of different directions. There's the websites that I think back to, from when I was a kid, and the websites that made exploring the internet really fun. That's the sort of feeling that I'm trying to recreate, almost. 

Then there's apps like Twitter or Threads. Microblogging platforms are a really great software type, and photo-sharing apps like Instagram are still a really good software type. I'm also sort of inspired by great design in general. As far as my own design goes I’ll hone in on a very specific description or niche that something is going to inhabit, and then I find that really drives the design decisions, and I can take that the whole way through the app. The app will make sense, and it’ll feel correct.

There are also apps I see in the real world that inspire me by being bad in some way, that have scattered design decisions that annoy me. I take that feeling and then I go try to build very polished apps.