Crescent Loom is a game about creating life. Knit bones, stitch muscles, and weave neurons into a biologically-realistic simple creature.
[ Online Demo ] [ Dev Log ] [ Early Access ]
Hi! My name is Wick Perry. I’m a neurobiologist-turned-game-developer based in Oakland, CA, and I study the neural circuits that move bodies. I believe that in order to make intelligent machines, we have to first master simple brains.
In March 2017, I ran a Kickstarter to make Crescent Loom, a game to bring neuroscience to everybody. Following that first year of making the prototype, turning it into a full-fledged game is now a matter of finding funding & collaborators.
How do bodies move?
There’s so much that goes on beneath our awareness. We don’t need to concentrate on our lungs to breathe, our teeth to chew, our legs to walk. Instead, we have special circuits of neurons called central pattern generators that produce these rhythms of motion for us.
(If you’ve played QWOP, you know how impossible walking would be if we had to control our muscles ourselves.)
Crescent Loom is a game that lets you — in a hands-on way — discover and get an intuitive sense for how small networks of neurons control bodies.
The first step to playing Crescent Loom is to knit a body.
Movement is entirely dictated by the physics simulation and force of the water; there are no typical video game controls that’ll make your creature swim. You have to think about the placement of joints, muscles, and fins to generate a forward force.
The primary way to move is with muscles, which contract to pull points closer together. For example, activating muscles along one side of a spine will cause that spine to bend.
After you master the basics, you can add begin adding organs: jaws to eat other creatures, suckers to latch onto walls, and harpoons to capture prey. Adding senses can allow your creature to begin reacting to the world around it.
In Crescent Loom, you don’t directly control your creature. Instead, you “program” movement by weaving neurons into a brain.
Neurons are modeled using a branched compartmental RC circuit, with ion channels as resistors, concentration gradients as batteries, and sections of cell membrane as capacitors. (Modelling algorithm developed by Gabriel Barello at the University of Oregon)
By running the simulation significantly slower than in real life, Crescent Loom allows you to see biophysically-accurate dynamics within a neuron (an actual action potential takes only a handful of milliseconds, while in Crescent Loom you can actually watch one travel down an axon).
In brains, there’s only one final output: muscles. Everything that goes in a brain is ultimately for coordinating the activation of muscles. Activating a motor neuron in the brain activates its corresponding muscle.
One of the most useful types of cell are pacemaker neurons. They rhythmically turn on and off and are key for any kind of repetitive movement. Here, a pacemaker neuron (green circle) connects to a motor neuron (red diamond).
An important concept is the difference between connections that are excitatory (which activate their target) and inhibitory (which silence their target).
Here is an excitatory connection (shown by a small triangle) that activates an otherwise-quiet neuron:
And here, an inhibitory connection (shown by an empty circle) silences an otherwise-active neuron:
By connecting different neurons in different ways, you can get more complicated and coordinated behaviors. My personal favorite is reciprocal inhibition; when a pair of pacemaker neurons inhibit each other, they naturally begin to take turns firing:
A comment I get frequently is “This is awesome! …what’s the goal?”
First off, Crescent Loom is in open development, which means the design can dramatically change as I see how people respond and follow the fun (& funding).
Right now, it’s a sandbox with an online ecosystem populated by player-made creatures. A creature’s fitness is measured by races along a long corridor; creatures that win more races will be more likely to show up in another player’s game.
Regarding future gameplay, there’s two likely paths:
- Open-world exploration. Double-down on the sandbox. Players explore the ocean to unlock new body parts and locations. Add more fitness challenges to enrich the dynamics of the online ecosystem. This has the advantage of potentially generating many hours of play, which does well in entertainment markets like Steam.
- Level-based puzzles. Carefully introduce concepts and abilities via designed levels. This gives me a lot of control over being able to teach how to play the game, and provides an easy-to-understand progression. It also has the advantage of being able to potentially fit into a classroom session, which is required if I want it to do well in an educational market.