In Conversation with Mathias Isaksen on Gumbo

Mathias Isaksen is a generative artist and creative coder based in Oslo, Norway, driven by a do-it-yourself ethos and a disregard for established practices. His work explores balancing opposing qualities and finding unconventional ways to apply commonplace techniques and methods. After finishing his master’s degree in applied physics and mathematics at the Norwegian University of Science and Technology in 2020, Mathias worked briefly as a tech consultant. This stint ended with a sabbatical to focus on his creative work, which eventually became a full-time vocation. His previous work includes the long-form project Factura, released on gm.studio in July 2022.

Jordan Kantor: Hi, Mathias. It’s a pleasure to get to chat with you in advance of the release of your debut project on Art Blocks. Before we get into talking about Gumbo, however, I’d like to take a step back and discuss how you first got into making art. Can you tell us about your creative beginnings?

Mathias Isaksen: Ignoring the drawings and doodles that everyone makes in kindergarten and primary school, I don’t have much of a creative history before 2020. (I was more interested in destroying rather than creating, to which my parents and several broken electronic devices can attest!). I bought a DSLR camera (a Canon 450D) when I was fourteen years old and dabbled with photography for a few years, before I sold the camera to pay for a Fender Jazzmaster electric guitar. As an older teenager I became interested in math, which led me to pursue a degree in applied physics and mathematics. Even though I’d probably choose a different path knowing what I know today, I use what I learned studying those fields almost every day, especially when creating generative art.

JK: Your story has some familiar elements: we find a lot of artists come to creative coding from math. How did you discover, and then get into, digital or generative art?

MI: It all started back in the summer of 2020, right after I finished my master’s thesis. This was a moment that I’d been working towards for five years, and suddenly it was all over. Perhaps unsurprisingly, I was left with a feeling of emptiness. Since this was back at the start of the pandemic, I tried to fill that vacuum by scrolling TikTok, and eventually came across a video demonstrating a pretty cool process:

1. Draw a bunch of rectangles on a piece of paper.
2. Attach a string to the center using a thumb tack.
3. Starting at the corners of the rectangles, draw lines along the string, ending in the center.
4. Finally, erase hidden lines and add color.

The end result is a drawing in one point perspective, kind of like seeing New York City from a helicopter. My immediate impulse was to try to recreate the process in R, the software I had been using in school for statistics and data visualization. That same evening, I sat down and threw together a rough sketch. I like to think the rest is history: I spent the summer exploring all of the clichés (shape subdivision, flow fields, domain warping, etc.), a phase everyone new to generative art seems to go through. The work that I produced during this period was predominantly geometric, a consequence of the limits set by the software. Eventually, I found a community for this sort of stuff on Instagram, where I was exposed to the work of artists like Joshua Bagley, Anna Lucia, Caleb Ogg, and Matt DesLauriers. 

Looking back, the most exciting aspect of this period for me was probably discovering an outlet that allowed me to leverage my love for programming to explore my creativity.

JK: A TikTok “how-to” as a gateway to instruction-based art making: what a contemporary story! So how did you discover the blockchain as a medium for art?

MI: In August 2021, I received an Instagram DM from a user with a CryptoPunk profile picture, expressing interest in buying some of my work as NFTs. At the time, I knew very little about cryptocurrencies, and even less about NFTs. I had read a few articles about CryptoPunks, Beeple, and various scams, which, honestly, left me with a negative impression of the space as a whole. Despite my suspicion that the DM was an attempted scam, I was intrigued and decided to mint a few pieces on OpenSea. In the end, it turned out the offer was legitimate, which opened my mind to exploring the NFT world further. The same day, I made a Twitter account and joined a vibrant community with many familiar names.

Understanding the connection between the blockchain and the code was a big “a-ha” moment, especially how the randomness is derived from the transaction hash. It seemed like a perfect fit for my interests, and I found the “long-form” format that arose particularly fascinating. At that point, my process consisted of developing an idea until it seemed mature enough, and then carefully curating from a large number of outputs. Seeing artists release projects capable of producing a never-ending stream of consistently high-quality work was shocking.

Going full-time in October 2021, I spent the next year working on various ideas and projects, culminating with the release of Factura, an edition of 999 I released in July 2022.

JK: I’d like to return to this distinction you just made for a moment, between using generative code to create outputs which you would curate before releasing to working on an algorithm until you were confident it could hold up aesthetically incorporating the randomness inherent across a large edition. Can you say a bit more about that?

MI: My process has, in many ways, remained the same. I still approach my work in an experimental and exploratory manner, and don’t start out with a clear idea of what a project should be or look like, but instead let it evolve organically. An important shift is being forced to think on a collection level basis. When making a change or introducing a new idea, you have to consider how it fits into the project as a whole, not just whether it works in this one output. As a result, things will tend towards being more polished, and it’s important to find a good trade-off between consistency and unpredictability. You don’t want to refine the code to the point where it no longer surprises you!

I also feel the need to mention that working on long-form projects is challenging and, at times, deeply frustrating. It can also be stressful, especially near launch, when the areas of improvement become less and less obvious. (This is probably a good sign to call it quits, which I don’t know how to do.) Many artists are familiar with the experience of looking through so many outputs that you start hating every single one, lose the ability to tell good from bad, and start making changes that exacerbate the situation.

Despite all this, working on these projects has been crucial to the growth I’ve seen as an artist the last couple of years. They’ve led me onto paths I’d never consider otherwise, and forced me to apply myself. Gumbo would not exist without having learned from these experiences.

JK: That’s the perfect segue. Let’s turn to the project at hand. Please tell us a bit about Gumbo.

MI: Gumbo evolved from a deep dive into a new area of technical interest and exploration for me, namely the combination of fragment shaders and ray marching. For those unfamiliar with these concepts, frag shaders are an important stage in the 3D rendering pipeline, used to add effects like lighting and texture to shapes created in earlier stages. Ray marching, on the other hand, is a technique used for rendering 3D scenes defined through a signed distance function (SDF), as opposed to the mesh-based approach used in conventional rendering. Piter Pasma’s Skulptuur is a great example of ray marching in action.

I find frag shaders deeply interesting in their own right, especially when considered outside the traditional rendering context (see, for example, Shadertoy). By setting things up in a certain way, the frag shader boils down to a program that takes in the position of a pixel on the screen and calculates its color, which enables us to render an entire canvas from the frag shader alone. This is not without drawbacks, however, as tasks that are usually simple become much more complicated—drawing a line segment requires computing the distance between the pixel and the segment, and using this value to blend between the line and the background. On the other hand, I find that having this level of control makes shaders able to effortlessly achieve effects that are almost unthinkable with a geometry-based workflow (i.e., one where you are essentially limited to drawing points, lines and shapes). When drawing the line segment, for example, a shader can easily add variation of color, deformation and texture in just a few lines of code. Shader code is, simply put, very expressive. You can do a lot with very little.

Because of this, I’ve been fascinated by shaders for a long time, but must admit, I’ve always found them a bit intimidating and inaccessible. Early in January 2023, after feeling burned out and taking a six month hiatus, I started studying shader-based textures. I deeply appreciate textures in generative art, both when used as a reference to things off-screen, and, even especially, when it takes on its own purely digital characteristics, independent of mimicking physical analogues. Bruce Conner, Virgil Finlay, Dmitri Cherniak, William Mapan, and Emily Xie are artists who incorporate texture into their projects in ways I especially admire.

In the study, I quickly realized that, for me, the most interesting results were obtained by taking simple methods and breaking them. The upper-left piece below, for example, involves taking the most naive type of noise and making an intentionally bad estimate of its gradient. Somehow, this leads to a texture that looks like a mycorrhizal network.

The right section of the same piece, which resembles a black-to-white gradient, demonstrates an important principle I ended up employing in Gumbo: by varying the ratio of black and white in a texture containing only these two colors, we can achieve any shade of gray.

As my investigations deepened, I started incorporating other ideas, such as ray marching. Simply, ray marching is iterative, which means, practically, that it starts with a bad guess and improves it step-by-step. It’s often lauded for yielding renders that appear physically correct, even photorealistic, as for example in the hands of someone like Inigo Quilez. I was excited to find new ways to mis-use ray marching: to break it, and fiddle with the parts that are usually kept static between different projects. An example of this in Gumbo includes limiting the maximum number of steps to a value much smaller than typical (think ten instead of one hundred), which, to my eye, can lead to a looser and more stylized look.

“I was excited to find new ways to mis-use ray marching: to break it, and fiddle with the parts that are usually kept static between different projects.”

Early in the development of Gumbo, I decided to try to approach problems from a position of naivety wherever possible, and then attempt the simplest solution. I introduced this kind of minimal limit as a corrective to my tendency to look for the most elaborate and technically satisfying solution to a problem. This often leads to overengineering and adds unnecessary complexity, making it more difficult to edit.

Take the color work, for example. My first impulse was to let the color application vary between the shapes that make up a structure. Taking the “naive” approach, however, led me to a more basic solution: assign a specific color if the position is within one of several 3D boxes, with each layer of color having its own set of boxes. I’m making it sound more complicated than it is—it required only a couple of lines of code, reusing functionality that already existed. Once implemented, I started seeing results that were completely unexpected, like the effect that occurs when the structure breaks through an area of flat color, revealing its spatial nature, or how the boxes become slivers as they are dragged along the silhouette of the structure.

While working on Gumbo, I’ve found inspiration through a wide array of sources, such as Cold War-era propaganda posters, Japanese ukiyo-e, and, of course, my Twitter feed. I’ve been particularly captivated by the amazing color work of artists like Utagawa Kuniyoshi (1798–1861) and Yashima Gakutei (1786–1868). In addition, the woodblock print method utilized in ukiyo-e, nishiki-e, builds the image by imprinting solid layers of color step-by-step. The inner workings of Gumbo are more or less a digital equivalent of this approach, a connection I realized after recently seeing a video of a contemporary master printer demonstrating this historical technique. (The same goes for the video you shared of Andy Warhol producing a silkscreen painting!).

JK: Woodblock printing to silkscreen painting to creative code creates a very interesting through-line of the use of novel reproductive technologies in artmaking. Thanks for outlining that. Tell us what collectors might look for in Gumbo as the series is revealed?

MI: Personally, I think it’s fascinating to try to identify the separate structural and chromatic components that comprise the final image and to try to see how each part contributes to the whole composition and the emergence that occurs when they combine. To help with that, I’ve tried to curate a set of features that best explain the variation across the collection. “Turbulence,” one of my favorites, adds a lot of drama, as if moving through violent turbulence in an airplane. “Iterations” corresponds to the number of steps taken during ray marching, and I find it interesting to compare how these steps affect the feel of each output. “Background coloring” and “structure coloring” are useful for unpacking how the diversity in the color application arises from a single rule. When building the compositional armature, the “structure complexity” determines how many types of shapes are used, where smaller values lead to simpler setups.

JK: There promises to be a lot to unpack. Thank you for that. Is there anything else you’d like to share that would help viewers approach and appreciate your work?

MI: Gumbo is a project where thumbnails and low-resolution renders will give an incomplete impression, as many details are lost when scaling down. I highly recommend grabbing an extra high-resolution render, just press “P” while in live mode, enter your desired size and wait!

JK: Thanks for the tip. Are there any recent accomplishments you’d like to share?

MI: Doing this full-time for a year and a half without losing my sanity feels like an accomplishment! I’m definitely looking forward to taking it easy for a while and getting a break after a long period of intensive and focused work. When I first got into the NFT gen-art space two years ago, the idea of doing an Art Blocks project, let alone a curated one, was completely unthinkable.

JK: We are delighted to host the project. Thank you again for bringing it to Art Blocks and for taking the time for this conversation. For folks who want to learn more, what is the best way for people to follow your work?

MI: For now, I’m most active on Twitter (@st4yhome_art). I’m also on Instagram (@st4yhome), and my website is mathiasisaksen.com.