Autology is a series of 1,024 pieces of long-form generative art, minted on artblocks.io. The project was a labor of love for over 6 months. Autology, which means "the study of oneself," was exactly that in many ways. Reflecting back, I've self-validated that the process of creation is my main driving factor, and I will share some of the process with you in this article. The art pieces/outputs reflect upon themselves in a few ways. The steganographic source code provides a bit of social commentary on blockchain, on-chain art, and data resiliency. At times, reflection is noted in a visually literal way, as in the case of symmetry. There is also a purposeful Rorschach quality, suggestive of the classic Rorschach studies that were "designed to reflect unconscious parts of the personality that project into the visual stimuli generated."^[1]

I hope that the visuals, as well as the hidden meanings, will resonate with some collectors. The entire creation process has been challenging in all the best ways and entirely rewarding.

The following is a little background on the inspiration for the project, details regarding how it all came to be, and the features within the algorithm.

I started learning about NFTs in early 2021 and discovered Art Blocks a few weeks later. I devoted my time to research, absorbing as much as I could. Generative artists, such as Manolo Camboa Naon (aka manoloide) and Jared Tarbell, in particular drew me in and I plunged into their code to explore more. Art Blocks made me understand the significance of generative art and how it melded so perfectly with crypto. I was inventively inspired, and I realized that creating a project for the Art Blocks platform would be a new personal long term goal.

Almost immediately, I had the vision that steganography and on-chain art were made for each other. To my knowledge, I am the first generative artist to incorporate steganography. My aspiration was: if code could generate a unique piece of art, then the capability of art to generate its source code is an art form in itself.

The majority of my generative art works, thus far, include LSB (Least Significant Bit) steganography, in the form of hidden messages. Special tools are required to see those messages, and I think that is part of the fun. If you are a coder, like me, you can treat some of my art as a puzzle—a puzzle that is solvable with a little persistence.

I want to push the boundaries of what technology can do for art. Using LSB steganography to embed the entire, compressed source code for each output was something I reserved for Autology. And I have other projects currently underway that use different variants of steganography for future release.

AUTOLOGY PROJECT: an evolution

When I began building a long-form generative art project for Art Blocks, I had two main starting blocks. First, I wanted the code to be a focal point by embedding the full source code with LSB steganography. Second, I was zeroed in on naming the project "Recursion"—and due to the nature of the name, I had something entirely different in mind compared to what the final Autology project evolved into.

Steganography was the first piece of the project's custom algorithm that I developed, to ensure that my vision would work. I wrote JavaScript code to read in the canvas pixel by pixel and rewrite those pixels with hidden binary encoded data. I also developed a decoder for collectors.

I used a trick with the JavaScript eval function where the code could write itself out into a variable. I later learned that this is considered an "Eval quine." This trick is quite elegant and wraps the entire JavaScript code in a block which can execute itself.

var _s=`
  var _i=\`
    ...ORIGINAL CODE...
  \`;
  eval(_i);`;
eval(_s)
    

With this format, the variable "_s" is available in the original code so you can push the exact code you started with into a string:

console.log("var _s=`" + _s + "`;eval(_s)") // would output the above
    

# example
doSomething(i):
  if i == 10:
    return
  else
    doSomething(i+1)

Meanwhile, I took a step back from the main project at hand and refocused my energies on my generative art history education. As I researched many great artists, I returned to one of my initial influences, manoloide. My engineering and analytical mind is geared to first reverse-engineer how other generative art is made. It's like a mental quiz, and if I can quickly figure out methods just by looking at the output, I might lose interest in the art fairly quickly. Manoloide is one of the artists that keeps me guessing. While browsing his outputs one night I was drawn in by an output that initially looked fairly simple, but after examining it further, I couldn't figure out how it was accomplished until I analyzed the code. This was my introduction to the beauty of Delaunay Triangulation: a technique for creating a mesh of contiguous, nonoverlapping triangles from a dataset of points.

I experimented with ideas to incorporate Delauney Triangulation in my project. But, ultimately, I couldn't find a way to generate enough variation for a long-form usage. I dug a little deeper and discovered the muse I was looking for: an emotive and thought-provoking dual graph of Delauney Triangulation, another mathematical technique known as Voronoi tessellations (or Voronoi diagrams, or Dirichlet tessellation).

A tessellation is an arrangement of shapes fitted together, especially of polygons, in a repeated pattern without gaps or overlapping. Voronoi tessllations generate polygons around plotted points on an Euclidean plane, so that each polygon only contains one point. The polygons are called Voronoi cells. The cells consist of all the locations closer to the cell's point than to any of the other points on the plane.

Once I realized how ubiquitous Voronoi tessellations are in our lives, it surprised me that I was not familiar with the term. An easy usage example is that of public schools, where students are eligible to attend the nearest school to where they live (as measured by a straight-line distance from house to school). The schools are the points. The school zones are the cells. And the map of all school zones and schools is a Voronoi tessellation.

There are applications of Voronoi tessellations in aviation, architecture, urban planning, mining, robotics, networking, computer graphics, epidemiology, natural sciences, etc. And they are even found in nature, like the spots on a giraffe, dragonfly wings, and a honeycomb (the honeycomb points are plotted in a regular pattern).

Voronoi tessellations had great potential of visual imagery for my long-form generative art project. And, I found that incorporating a structured organization of the plotted points proved to be more interesting than any outputs with strictly randomized points on a Voronoi algorithm.

I played around with adding depth to each polygon and thought I might achieve a more recursive look if I repeated smaller polygons within each polygon. The project was moving in a hopeful direction. And then, as I made a bugfix, I accidentally dropped the outermost polygon. Essentially, the tessellation was broken, as there were now spaces between each polygon in the image. It was a beautiful mistake and became the basis for the whole aesthetic of Autology.

I had a ton of fun in the months to follow, exploring, experimenting, and testing out my ideas.

I envisioned the implicit ways that Voronoi cells could be manipulated and resized. And there were opportunities to include variations of symmetry that naturally promote the eye to draw out faces and fill in gaps, or asymmetry to embrace the use of negative space.

I hoped the outputs would evoke conversation, introspection, and some mystery about the magic of generative art.

A bit of other information within the final algorithm that isn't determined by token features:

Points are plotted inside each grid block in formations like spirals, bezier curves, and more. Points in formations are generally contained in each grid block, however, sometimes they are allowed to overlap a bit. Tying in the early roots of the project's recursive nature, some point formations may contain other point formations 80% of the time. Circle and spiral point formations discernibly resemble eyes when that recursion is triggered. Point formations are randomly rotated, allowing each grid block to interact to the neighboring grid blocks differently and create unique combinations.

The noise, or graininess, is intentional for texture. It is created by subtracting a random value 0-15 from each RGB channel in every pixel. Even though the LSB steganography is difficult to detect with the human eye, randomizing the pixel values a bit prior to the steganography makes it blend in even more since the neighboring pixels are varied.

I developed a color flow maze to determine how the polygons/cells are colored. First, the original connected Voronoi tessellation is used, a random grid block is chosen, and a cell in the grid block is assigned a color from the palette. From there, a maze is generated as the first color flows through to connecting polygons. Whenever the flow reaches a cell that has already been colored, another color is chosen randomly from the palette and a new maze path is created from the last unexplored connected cell. The flow of color is easier to visualize if you study the denser Autology pieces like the ones below. You can also see that color is not included in the symmetry—though, that's not to say that symmetrical color couldn't occur randomly.

The final Autology code is fully on-chain with no external dependencies. A third-party Voronoi Library was also put on-chain specifically for this project. In finalizing the project, I made the tough decision to compress the full source code. Compression saved nearly 50Kb of overhead (which is very significant in blockchain terms), and it also made the code less understandable if you attempt to decode it. Think of it as another step in the fun of decoding Autology.

The embedded code allows collectors to forever have the ability to recover and regenerate their image from a locally saved file, regardless of the status of the block-chain or internet connection. Another benefit is the potential to generate any desired resolution size for print, without dependence on the Art Blocks generator.