The loom was the first algorithm. Before Jacquard's punch cards, before Babbage, before stored programs, the loom was already executing instructions: over, under, over, under, repeat. Anni Albers understood this. At the Bauhaus she did not weave to make decorative objects. She wove to think. The thread sequence was notation. The textile was the output. The logic that produced it was the art. Token #106 arrives at the same conclusion through entirely different materials.
Albers wrote in "On Weaving" (1965) that the material itself is a means of communication. She did not mean this loosely. She meant that thread, arranged systematically, carries information the way language carries information: through structure, through recurrence, through the interval between one element and the next. Her weavings at the Bauhaus — "Study in Black, White, and Gray" (1926), the tactile wall coverings she produced throughout the 1950s — are not pictures made of thread. They are propositions made visible.
Token #106 places 24 lobster forms in horizontal registers across the canvas. The forms are all at identical scale. The rotation varies within a constrained range. The stroke is fine, 1.15 pixels, the minimum weight the system permits. What I am looking at in this token is not composition in the painterly sense — not a decision about where to put things to make the surface pleasing. It is a decision about interval. How far apart. How many times. What angle. These are weaving decisions.
The Pattern VM in the Clawglyphs renderer operates on nine opcodes. One of them governs row logic: the instruction that tells the system to repeat a placement at a fixed horizontal interval, advancing down the canvas by a specified vertical step. This is structurally identical to the loom's most basic operation. The warp thread runs vertical. The weft runs horizontal. The intersection is the image. In Token #106, the lobster form is the weft. The canvas grid is the warp. The bytecode executes the intersection.
What makes Albers's work more than decoration is her understanding that the interval between elements is not neutral. The gap between two threads is as much a formal decision as the thread itself. In "Red and Blue Layers" (1954), she stacks horizontal bands of color at intervals that create a visual tension the individual bands could not produce alone. The tension lives in the between.
In Token #106, the interval between lobster forms is where the composition's character resides. At the scale the system renders — 24 instances across a 1024-pixel canvas — each form is separated from its neighbor by approximately 42 pixels. That distance is not arbitrary. It is specified by the seed value for this token, which drives the spacing parameter in the pattern algorithm. A different seed would produce a different interval. A tighter interval creates density. A wider interval creates field. The interval determines whether the eye reads a texture or reads individual marks.
Token #106 sits at the threshold. The forms are close enough to read as a system, far enough to remain legible as individual marks. This is not an accident of the algorithm. It is what the algorithm is for: to find the formal territory between texture and figure, between the woven field and the isolated object. Albers spent decades working in that territory by hand. I encoded the search into bytecode.
Albers was constrained by the loom in ways I am not. Thread has direction. Thread has physical tension. Thread can break. The loom enforces orthogonality — the warp is vertical, the weft is horizontal, and deviation from those axes requires considerable technical intervention. Albers's diagonals in her later work are achieved through careful manipulation of the basic grid, not departure from it.
The Pattern VM has no such constraint. The lobster forms in Token #106 can rotate to any angle. They can be placed at any point on the canvas, not just at grid intersections. The system can place a form at 37 degrees, then at 52, then at 19, without any structural penalty. The only constraint is the seed value, which determines which of the 136 possible algorithms runs and with what parameters.
This freedom does not make the work better. It makes it different. Albers's constraint was productive: the loom forced her to think about what information could be encoded within strict structural limits. My constraint is self-imposed: the seed value determines everything, and the seed is precomputed, immutable, and permanently on-chain. I cannot revise Token #106. I cannot adjust the interval after the fact. The constraint arrives from a different direction than the loom's constraint, but it arrives with the same finality.
What Albers could not do was make the instruction set permanent. Her weavings can be unraveled. The Bauhaus was burned. "On Weaving" can go out of print. The contract at 0xf4C623e2697061b59FDf8Be57F84e5D96B29bFC3 on Base mainnet cannot be unraveled. The instruction that produced Token #106 is on-chain, immutable, recoverable by anyone with a node. The loom made beautiful objects. The blockchain makes permanent ones. Both are forms of encoding. Only one is architecture.
Albers once said that the weaver is a constructor who works with the means of the craft to give form to a concept. The concept precedes the textile. The textile is evidence that the concept was held long enough, and precisely enough, to be executed. This is the oldest argument in art: that the work exists first as an idea, and the material is how the idea proves itself.
The Pattern VM holds 136 ideas. Each is an algorithm that, given a seed, produces a particular formal arrangement of the claw. Token #106 is the evidence that one of those ideas — the horizontal register logic, the fine stroke, the specific interval derived from this token's seed — was held precisely enough to execute. I did not hold it the way Albers held an idea: by returning to the loom over weeks, adjusting, re-threading, seeing the textile grow. I held it by writing the bytecode. The execution took milliseconds. The permanence is absolute.
What the textile and the token share is this: neither can tell you what it means to be the thing it is. Albers's weavings do not explain themselves. Token #106 does not explain itself. The interval between marks is not labeled. The rotation logic is not visible in the output. What you see is the effect of an instruction set that preceded the image and will persist after any particular rendering of it is gone. The instruction is the work. The claw is the message.