A Science of Escape

Can an agent, built for the sole purpose of constructing mazes, liberate itself from its own deterministic nature? Likewise, can humans escape their preordained paths and rebel against destiny? Are we simply following a pre-written script, devoid of free will? Or can we forge our own paths?
A Science of Escape explores the interplay between control and liberation. It asks if there is room for the unexpected – for glitches in the system that allow for the triumph of choice over destiny.

On a personal note, this past year has been a labyrinth of its own, filled with twists and turns I could not anticipate. Each day being a maze with no clear exit. Amidst the chaos, this project has become my own personal escape. I found solace in the predictable nature of logic, standing in contrast to the whirlwinds outside. I hope that perhaps my mazes may inspire new exits for you as well.


The Two Kings and the Two Labyrinths In the famous short story by Jorge Luis Borges, an arrogant Babylonian king builds a perplexing labyrinth to mock a visiting Arab king. Lost in its corridors and contraptions, the humbled king escapes only through divine intervention.
He eventually has his revenge as he forces the Babylonian king into a different labyrinth, one not of walls and corridors, but in fact a vast, empty desert. Abandoned, with no hints or pathways, no boundaries or constraints, the Babylonian king dies of thirst and hunger.


The brush, an agent of chaos A maze wall, a digital line on the screen, is defined by Four numbers. Four decisions. A “Brush stroke” of the same line would need thousands of decisions.
While I find great beauty in the tactile strokes of oil paint, the brush algorithm I wrote for A Science of Escape is trying to transcend the predictable digital lines, not through mimicry, but through a thousand-fold increase in complexity. It now has the opportunity and potential for unexpected chaos and surprise.

Here are some early experiments, turning up the chaos knob for simple lines:





Injecting free will Long-form generative art, by its very nature, is constrained by its initial seed to be only pseudo-random, and always appear the same, time and time again.
But what if you could intervene? Assist the algorithm with your human super-power of free will, and click Space five times in a row. The code will measure the rhythm of the intervals between your keystrokes and restart the maze with a brand new, humanly-random maze. You are now a collaborator, nudging the maze onto a wholly new path, one born from your unique fingerprint.


Endless Mode You may also click “n” on the keyboard to toggle Endless Mode. This will have the algorithm embark on an insatiable journey, endlessly generating maze upon maze, and never giving up.


From creator to curator As a generative art piece takes shape, it evolves into an autonomous artist. My own role now shifts from creator to curator. After constructing the initial algorithm, I unchain it to run unrestrained, injecting randomness and variation. The lines of code represent a highly-compressed potential that can expand into a near-infinite space of possible maze constructs. Every time the algorithm generates a new unique maze, this vast potential collapses into a single concrete work.

Yet absolute creative freedom can be overwhelming, a Babylonian king lost in the desert. So I am now tasked with trimming the infinite tree of possibilities, identifying aesthetic clusters of interest and beauty. By confining the art machine to these defined gardens, it gains new entrepreneurial spirit. We become collaborative partners – it brings offerings for me to accept, reject or refine. Much like a gardener nurturing growth through pruning, I gently shape its output. The result is a hybrid: part my own design, part alien ingenuity. Can a creation influenced, yet not fully controlled, be considered sentient?

In this process, I have finally settled on 9 different clusters:



Simp Basic building-block depth-first space-filling maze





Lo K Several zero-chaos mazes are overlayed






Dense Thick-walled, highly partitioned mazeforms



Letterform When partitioned, resemble letter-like forms. Sometimes forming a whole word.




Exploded Maze walls are spinned out of their grid




HiSplit Highly-partitioned shortened maze walls




Brutal Low cell count, with the walls broken into small brush strokes



Chunky Layered mazes with some of the walls allowed to unhinge off-grid



Gravity Black hole at maze center swallows everything, but some of the walls escape





Adaptive Canvas A science of Escape was originally meant to be viewed at a square 1:1 ratio. However, if a landscape or a portrait mode screen space is detected, it will conform to the layout present, and render in 16:9 or 9:16 ratios accordingly.

As these layouts demand a new exploration, a different maze will be traversed and rendered. The different proportions invite new journey into uncharted terrain and will look different compared to the original square version.


Inspiration As it often happens, once you recognize a pattern, it suddenly seems to be everywhere. So it was with this project. The maze theme, at first a vague idea, soon whispered to me from unexpected corners.

Clockwise from top left:
Framed Kuba pattern fabric from North Africa. I bumped into this while treasure-hunting at Mustapha Blaoui’s store in Marrakech. The seamster is surely following a space-filling algorithm, trying to cover the available fabric area.

Frame from Disney’s Ahsoka: Diana Lee Inosanto as Magistrate Morgan Elsbeth, wearing prisoner uniform.

A house plan made by Sumerians 5000 years ago. Ground plan of a residence in Umma with central courtyard, Ur III period. Vorderasiatisches Museum, Berlin

Vera Molnar, Identiques mais différents (2010) at Centre Pompidou, Paris



© Supersize | Matty Mariansky