Maybe harkening back to the old-school days of HN, but I'd be keen to hear about the tech stack you decided to use in building the app, what your considerations are for multi-platform, and advice for start-ups getting ready to launch their own apps.
John here, one of the app engineers at Substack. Our team has lots of experience with Swift and values the level of polish you can achieve natively, so went with that. We try to avoid third-party dependencies so native implementation and customization of UIKit, URLSession, and CoreData are important to us.
Cross platform frameworks like React Native, Flutter, Xamarin, etc are super interesting but in our experience can hold an app back long term. As you're able to dedicate more resources to the product, it's more likely that you'll want custom UI (animations, transitions, etc) and functionality that require full access to native APIs. It's a tough decision to make for early stage start-ups but we also went with native at Cocoon (our last startup which was acquired by Substack). I think some important factors that should go into the decision are 1) how important you think both Android and iOS support are out of the gate 2) how soon you'd expect to be able to hire native engineers for both platforms 3) your long term vision for the product and its desired complexity. We love Android (I've got some Kotlin experience myself) and will have it out ASAP, but with the existing email/web product there is less pressure for cross platform in our case.
Yeah, we've started looking at Kotlin Multi-platform Mobile to keep our business logic cross-platform, but then native UI. The discussion we're having is if it is mature enough for our needs now, or if we write in Flutter with the expectation we throw it all way in the near future.
We all use Android while our users are mostly in single platform approach won't work for us. We're also a hardware company, so need BLE and WebBle support isn't good enough for us.
I'm curious as I think I'll be making this decision in the near future.
> it's more likely that you'll want custom UI (animations, transitions, etc) and functionality that require full access to native APIs.
Besides the custom UI stuff, as I think React Native has some decent animation support nowadays, what specific native APIs were needed in Substack's case that made React Native an ineligible choice?
As text being the main content for Substack, Its hard for me to imagine why React Native wouldn't be sufficient for a relatively simple app.
Discord, undoubtedly, has more complex UI needs compared to Substack, however it's humming along just fine with React Native.
What makes React Native capable for Discord, but incapable for your needs?
> At first, we felt that maybe doing it purely in JavaScript was futile. We spent some time trying to glue together UITableView with React Native, and while we made meaningful progress, it started feeling overly complicated. After stepping back and thinking about what else we could do — it hit us. We already solved this problem once before on the web! We already had an internal List component that virtualizes its children. There is no way we could just drop it into React Native right?
I'm also curious why it took them so many pain points to reach list virtualization as their solution. That would've been my first thought.
That was their initial problem.
Upon further reading, you will find that they found a solution in dealing with Lists with React Native. My initial comment had the answer - they brought in a component that visualizes lists from their web platform, and it worked seamlessly in React Native.
So no, while they initially implemented their core chat view natively, they are using full blown react native as of right now.
>The result was a new component we called <FastList>. The team intends to merge these together for cross-platform use and open source it for the community. With that we removed a lot of memory allocations and were down another 70-90ms of render time. We could now scroll the channel members list as fast as we wanted and it kept up admirably.
We'll get you a real answer about how the tech stack is today.
But for now I can tell a war story that might be fun. I'm one of the founders, and am in a very much non-technical role now as CEO, I did write a bunch of the very early code (some of which people curse my name for to this day.)
When we were starting, I was limited by how many new languages/frameworks I could learn at once. I started writing the backend in python, because I knew it a bit. But our first writer often needed to use Chinese characters, and in python 2.X I could never get unicode strings to work properly. I couldn't upgrade to python 3, because on google cloud I would have had to learn Docker and I was already learning too many things at once.
Eventually I got so frustrated I threw out several days work and started the whole backend over, with node + Postgres hosted in Heroku. This ended up defining much of the stack we use to this day, which might be good or bad depending who you ask. At least unicode works though :)
I’m curious: what was the issue? I’ve used python 2.7 with Unicode pretty extensively, in a wide range of languages, and have never had problems that weren’t my own fault.
I have memories of the Python compatibility issues causing huge headaches. We're back to using Python due to ML requirements now, but it kept me away for years.
I tagged in one of our lead engineers to talk about the tech stack, but can chime in re: considerations for multi-platform.
We are sprinting as fast as we can to get an Android app out the door. We're also planning on investing more in the reader experience on web. Some time ago, we launched a web reader (reader.substack.com) in beta and have some exciting ideas in the works to evolve that surface.
Beyond Android, when it smaller platforms like iPad / Desktop apps, it's mostly a matter of looking at the data and listening to users. With a small team, we have to be judicious with prioritization, and as we increase the surface area for readers it's important that the experience for writers remains clear (right now their readers can already read on email, web, and mobile).
