Console

Interview with Brooklyn Zelenka

CTO, Fission - Decentralized app backend.

What is Fission and why did you build it?

Fission is building the future of web apps at the edge- it lets you build full, complete applications with only frontend skills without having to write or maintain a server.

My co-founder, Boris Mann, and I were coming out of the Web3 space doing a lot of things with blockchains and saying, “Well, okay, this stuff’s all interesting, but very few people can actually use this. What if we applied the same techniques outside of a blockchain context to the 99% of developer problems. We thought we’d see what would happen if we completely flipped the architecture and rethought the constraints for today, because things have changed a lot since the web started in the late ’80s / early ’90s.

I’ve seen the same patterns over and over on web projects. Building a production app today requires a lot of moving parts. There are now separate disciplines for (at minimum) frontend, backend, and SREs/DevOps. Teams need to develop talent and grow to cover all of these specialties. The distance between the end user and much of the team is only growing as we have this tower of technologies.

As a full stack developer myself for many years, I witnessed the overall complexity of applications that are about a user wanting to display, edit, and share data. Teams want to iterate on problems, which requires staying nimble and close to the end user. This can become challenging when you’re waiting on a database change to be reflected in the GraphQL layer because the auth roles need to be updated after the production migration.

Fission helps you use frontend skills to ship full apps, without needing to build and maintain servers. It empowers teams to ship more with less, and “make the right thing the easy thing”.

Our hope is that we could let people do more with less. Let smaller teams compete with larger teams, and let people, even individuals, own their own data, build little websites, have small businesses, maintain things and not need a team of 30 engineers to get going.

Fission product screenshot

Fission Drive for user data storage.

How do developers approach building with Fission?

Everything runs local-first. The constraints that we’ve set is that everything has to work directly in a browser with no plugins, no extensions, without degrading the user experience or security.

An end-user shouldn’t even know that something interesting is happening, but for the developers, it should be radically easier to build and run. You should be able to have everything run totally offline.

If there’s another device making changes, when you both come online they should reconcile all of these things. The network latency really dominates when you start looking at performance in applications.

If I’m going to send something to a server to pull some data out of a database, process it and send it back to me, the long part isn’t the computation, it’s sending stuff over the wire. If we can take network latency to zero as a default, things will move very quickly.

My iPhone has way more computing power than my first laptop ever did. These things are very, very powerful, capable devices. Other than data locality, there’s no reason to ship data around, so we got rid of the data locality requirement. We deal with the vast majority of compute locally. Aside from checking if this person is actually allowed to push this new data, we essentially do nothing on the server. Everything’s done directly on the device.

What are the use cases for Fission?

Like all emerging or new technologies, they take some time to catch up, and then surpass what’s there. For anything that’s more of a transactional process or if you want to make a post on social media, that kind of thing. All of that’s doable today.

Fission lets you ship apps 10x faster. It lowers barriers to entry by only requiring frontend skills. It helps you focus on the parts that are absolutely essential: data and UI.

Due to these constraints, Fission takes a different approach to a lot of problems. This leads to several additional features, including end-to-end encryption and encryption-at-rest, user-owned accounts, passwordless logins, global serverless files, file versioning, and offline support (among others), all out of the box.

The interesting thing that this enables is edge computing. There’s some changes coming to the actual physical layer of the network in the next couple of years, and this isn’t just us hoping or guessing at this. Gartner, IBM and all these big companies are doing research saying, “Three quarters of all computing is going to happen outside of the cloud by 2025,” which is very soon. With things like 5G networks, Starlink and new systems where there’s quite a lot of bandwidth, you can start really lowering even the latency between systems.

You start hitting the physical limit of the speed of light very quickly. When you have 40 millisecond and lower latency, you open up a lot of use cases. We’re already seeing the use of sensors growing very quickly, and the volume of data is enormous, and they need to be able to react in realtime. This makes things like IoT a good use case, even though that wasn’t our original intention. Self-driving cars, extended reality, all of that work really well in this model.

When you get down to about 8 milliseconds, you can start doing all kinds of stuff where the data getting crunched locally on a device can instead be pushed out to compute resources that are literally across the street, next to the 5G receiver. It feels completely transparent because it’s an 8ms round trip. If there isn’t one, then have the device to run it locally.

All of these interesting new things fall out of this. Now that I don’t need to maintain a server, data is owned by the user, compute can happen anywhere, the end use cases beyond what’s possible today start opening up.

What does a “day in the life” look like for you?

I don’t know if I have a typical day. I end up bouncing between a lot of tasks and contexts, switching quite a lot, which took some getting used to. For deep work, I still try to defragment my calendar, so I’ll usually try to take the afternoon with as few interruptions as possible. I split my time between coding, research and calls where I do team check-ins and talk to investors.

We participate in some standards bodies for the tech that we use, such as The Decentralized Identity Foundation. A lot of our stuff is a really nice fit for a lot of the other members. We’re looking at standardizing a lot of what we’re building.

Boris, my co-founder, handles most of the external facing stuff so I’m more internal facing. I try to get chunks of unbroken time to do code or review PRs, but the most valuable time is doing the research part of R&D and figuring out, “Okay. Well, we don’t have a central server. How are we actually going to make the data sync up? What’s the most efficient way to transmit changes in some data?”

We’re fundamentally rethinking how web apps can be built. That means big chunks of this time is spent trying things out: prototyping, writing a spec, talking to the team, seeing what ideas they have, what parts can be integrated, balancing trade-offs and seeing what works best.

What do you think about the risk that you’re still quite early as a business?

From very early on, we decided that everything we would do is open source. You can run your own Fission stack. For convenience reasons, probably people mostly won’t, but if they want to, we actively encourage them to do so because of the way that a lot of the underlying technology works. It federates very easily.

With more people running both IPFS, which is the underlying networking and data stack (with our own additions), the availability of the data is growing. It lowers our hosting costs because other people are also hosting the same data. We’ve even thought about shipping people Raspberry Pis with the complete stack on it, so you can just plug it in at home and run your own stack.

Standardization would be nice to inter-operate with some of the other people in this space. If we get spec adoption, that’s just having a good impact on the world. We want people to run their own. We don’t charge people today, but the eventual plan is to charge for services, domain names etc. We see this as more of a collaborative approach to the ecosystem as opposed to a winner takes all.

What is the team structure?

There’s seven of us right now. Five engineers, our CEO and our head of product. As you might imagine, it’s very engineering heavy. People will talk to our CEO, Boris, and think that he’s a senior engineer, but he doesn’t actually do any active coding these days, he’s the least “technical” on the team, but needs to be able to think about the system in a detailed way, and communicate these ideas to folks that aren’t with us in the R&D mines.

On the engineering side, aside from myself, we have someone focused on the developer experience, someone that’s more research focused and two that are responsible for building the actual SDK and sample apps, so that people can get a sense of how to use these things. Everyone wears a lot of hats at Fission, but we’re roughly split between raw R&D to solve the underlying problems, and making it a good experience for teams to adopt.

How did you first get into software development?

I fell into it pretty much by accident! It’s a bit of a winding story, so here’s the short version:

I studied classical music in university, where I developed some Photoshop and Illustrator skills to help with concert posters and pamphlets for my classmates. After dropping out, I leveraged my skills as a graphic designer and worked a few jobs doing that. Eventually I found myself at a startup. They asked if I would be willing to pick up some frontend skills to help out, and sent me home with a couple books over a weekend. It turns out that I was a better programmer than designer, so I ended up switching careers!

That company was using a JVM-based framework for the backend — the selling point was that you could mix various languages together for different tasks. The idea was to “use the best tool for the job”, which in retrospect was more like “for everything and nothing”. I ended up picking up a bunch of languages very early on, like JS (Rhino), Clojure, JRuby, Groovy, Jython, Java, and a few others. I found that I had a knack for languages, and started to really delve into PLT and falling in love with functional programming in particular.

What’s the tech stack for Fission today?

The vast majority of things that people would actually touch are written in TypeScript. A lot of that will be getting migrated to Wasm over time. For today, we want things to also work on Node, hence TypeScript.

Most of our first party apps are written in Elm. On the backend it’s Haskell. Our infrastructure runs on AWS, and we’re using the Golang implementation of IPFS.

What’s the most interesting development challenge that you faced building Fission?

Everything we do has to work in a browser with no extensions, without UX or DX degradation, or security flaws. Architecture without assuming a central server as the source of truth is very different, and the browser is an extremely hostile environment. This has forced us to innovate in a number of areas (“necessity is the mother of invention”).

This shows up in a plethora of ways, but to name a few: Access control has to work locally, baked directly into data with encryption at rest. Accounts originate from the user’s device, not an OAuth server, and finding a consistent way to refer to data when there’s no database primary key.

What’s the most interesting tool or tech you’re currently playing around with?

Nix / NixOS. We’ve ended up using it at Fission as well, even though this was more a side thing for me. Some of the other engineers also got interested in it. and so we’ve just adopted it now.

It’s essentially a declarative package manager, but it also plugs into build tools and scripts and ops. It is, by far, the best of that suite of tools I’ve ever used. I use it to manage my Mac, and run a NixOS server.

I can’t imagine going back to ad-hoc dotfiles and conflicting dynamic libraries and changing versions of Node for different packages. Now, I’m in my Nix shell and everything works, I don’t have to mess around with anything. The main downside is the documentation isn’t quite there yet, but once you get over that hurdle, it’s really lovely.

Describe your computer hardware setup

I have two setups. Pre-COVID, I was on the road a lot, so would typically travel light with just a MacBook. I now have a M1 MacBook Air. I tend to use just the main built-in keyboard when I’m on this machine.

The other setup is actually the one that I prefer. Whenever possible, I mosh into a DigitalOcean NixOS machine from my iPad Pro using an app called Blink, using Tailscale. I have an Anne Pro 2 mechanical keyboard with the really clicky blue switches which drives everyone around me to rage, and a Microsoft Arc mouse, which is a little bit like a slap bracelet in mouse form. The M1 Air is just a spectacular machine, but the iPad setup has been my favorite, partly because I have a little stand for it, which is so adjustable.

I have an old neck injury, and so being in a more upright position really helps. I can also switch modalities to the Apple Pencil if I need to take some notes or draw a diagram and it’s just right there. I don’t have to mess around with the trackpad or a mouse which is fantastic.

I was not sold on the idea of an iPad when it was first released, but it’s really become by far my favorite device. I wish Apple would stop welding the lid on this thing shut so that we could install things on it, but that’ll never happen.

Which one are you coding on?

Both. On the iPad, I’m remoting into the NixOS machine. On the Mac, I tend to work locally. But because they’re both managed with Nix, I have the same config on both.

Describe your computer software setup

OS: macOS and NixOS.

Browser: Firefox Developer Edition with Vim Vixen so I get the Vim bindings.

Email: Gmail, plugged into Missive.

Chat: Discord.

IDE: Emacs.

Source control: Git + GitHub.

Describe your desk setup

No fancy chair. I try to keep my setup fairly light. I have an iPad stand for when I’m using the iPad, and have a little portable riser for my laptop. When it’s not COVID I like to work from cafes a lot so I bring that and just lift it up a bit and use it with the mechanical keyboard.

I have an Ikea adjustable standing desk. I used to work at a desk really consistently, and I’m finding these days, especially as I’m working from home as opposed to offices or cafes, I’m now standing, sitting and laying down quite a bit more. Reclining on the couch, and then switching to a table, and then going to the standing desk. Switching between those as needed.

This has been surprisingly good for me physically. As I mentioned before, I have this neck injury of a compressed disc in my neck. For almost a year, I couldn’t look fully to the right, and keeping my neck in a still position was required, but that made everything very stiff. I’m finding even just the change throughout the day has been very helpful.

The desk of Brooklyn Zelenka, Fission

When coding

Daytime or nighttime? Nightime.

Tea or coffee? Coffee.

Silence or music? Music.

What non-tech activities do you like to do?

My biggest hobby right now is building an 8-bit computer from scratch!

Outside of tech, I really enjoy both cooking and fermenting things. A lifetime ago I worked in a kitchen and that felt like work for a really long time. I’ve been doing a classic French cooking course which has been fun. I have also been brewing ginger beer and tepache, tweaking different spices, that kind of thing. I’m getting very excited about braising things lately.

Find out more

Fission is a decentralizd app backend. It was featured as an “Interesting Tool” in the Console newsletter on 26 Aug 2021. This interview was conducted on 21 Aug 2021.