SpiroFlakes
Alexander Reben · Project 136 · Factory
Edition size: 1,024 · First mint: 12 August 2021
The original sequence plays automatically. Right advances immediately; Space toggles playback. A paused display may clear after a resize. Open artwork
A circle that gives way to another
SpiroFlakes by Alexander Reben presents a dense circular drawing, then replaces it. Fine coloured lines gather into petals, rings and almost solid discs. A yellow centre can seem to radiate through a blue mesh; a dark band can open a hollow inside a bright rosette. Just as the eye begins to separate the layers, the next composition changes the background, density and pattern together.12
The work belongs to Art Blocks' Factory category and comprises 1,024 minted tokens. The official collection page dates its release to August 25, 2021. Its precise identity is Ethereum project 136 on contract 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270.34 The AB5D artist dossier situates Reben's work across software, robotics and participation.
Here, the viewer encounters a compact drawing machine. Each token supplies a repeatable starting condition for an ongoing sequence, rather than one unchanging circular emblem. The visible repertoire suggests the artist's stated associations with spiro-art, snowflakes and guilloche.3 Its intricate appearance depends on the accumulation of many simple line segments, while its changing display makes duration part of how those details are experienced.
Play within a small system
Reben's published project statement locates complexity in sine and cosine and couples that premise with an invitation to control playback.3 It is a concise account of the work's mechanism and encounter. It does not supply a detailed development chronology, identify preparatory drawings or explain why this particular palette and timing were chosen.
His broader artist statement describes invention as a way of asking questions through functioning things. He emphasises participation and play, allowing a system to produce surprises while retaining human responsibility for it.5 These ideas offer context for SpiroFlakes' combination of automatic generation and simple controls. They should not be mistaken for a contemporaneous explanation of every decision in the 2021 project.
In a December 2021 interview with Mariepet Mangosing, Reben discussed humour, unexpected uses of technology and his interest in unsettling assumptions about what an NFT could be. He also described leaving room for discovery while directing generative outputs.6 SpiroFlakes can be read in that spirit: a collector receives a sequence-making system whose individual images remain available only temporarily unless playback is interrupted.
Its small scale attracted specific attention. In May 2022, Kyle Waters' analysis of 300 Art Blocks projects identified SpiroFlakes as the shortest script in that surveyed group, at 1,244 characters.7 The retrieved script still has that length. The comparison is bounded by Waters' survey and date, not an all-time claim. More significant than the ranking is the disparity between a very short instruction set and the densely worked surfaces it produces.
Two rotations and thousands of crossings
The complete artist source consists of 1,244 characters containing HTML, CSS and JavaScript. Its indexed runtime label is custom@na. The code creates a canvas and uses the browser's two-dimensional drawing context directly; it contains no imported drawing library, downloaded imagery or machine-learning model. The full source was read from the official indexed project record, with a stored SHA-256 digest. This was not a direct block-pinned chain retrieval.4
Only the first eight hexadecimal characters after the hash prefix initialise the variable part of the pseudorandom state. Three other state words begin with fixed constants. Bitwise shifts, additions and a counter advance the generator, with sixteen initial calls before drawing. The rest of the hash and the invocation number are not read by this script. Its identity-to-image mechanism therefore uses a specific thirty-two-bit portion of the token hash rather than the whole value.4
Every new composition begins with a randomly selected RGB background. It then draws between eight and seventy-one paths. All share the canvas centre, but each receives two radial distances, an angular multiplier, a sampling increment, a line width and a stroke colour. The stroke list is cyan, magenta, yellow, black, red, lime and blue. The selection favours the first four through a branch that sometimes restricts the available list; the seven colours are not equally likely on every draw.4
The coordinate formula adds two circular motions. One pair of sine and cosine terms follows the main angle; another follows that angle multiplied by a selected integer. As the angle advances from zero through a full turn, the program joins sampled positions with straight segments. Those connections can make tight petals, broad overlapping loops or extremely fine mesh. The larger multiplier mode also passes through JavaScript's signed integer coercion, so it should not be described as a simple unrestricted positive frequency range.4
The resulting paths are stroked, not filled. Nevertheless, many crossings can make a region appear opaque. A centre that looks like a flat disc is often dense linework; a broad ring can be formed by countless crossings surrounding a less densely occupied interval. Because each path is centred in the same place, the overlays remain organised around one strong axis even when their local structures compete.
The display is a square fitted to the smaller window dimension, with its backing resolution multiplied by the device pixel ratio. Radii and stroke widths depend on that resolution. Resizing assigns new canvas dimensions and clears the bitmap; the resize handler does not itself redraw the previous pattern. That detail matters when the automatic sequence is paused.4
Time does not continuously deform a stored curve. The animation callback schedules complete redraws, nominally two seconds apart after initialisation. Each redraw consumes more values from the continuing pseudorandom stream. There is no finite list of stored pictures and no programmed return to the first composition. The artist's description of never-ending generation refers to this continuing procedure, not proof of an infinite set of mathematically unique images.4
Space or a click toggles the automatic sequence. Up halves the interval when it is above one millisecond; Down doubles it; Left restores two seconds. Right calls the drawing function immediately, including while paused. It advances the stream rather than going back to an earlier image. Restoring speed does not restore the seed.4
The original #449 renderer accepted manual advance and resumed automatic redraws. Paused captures repeatedly appeared black, while Right immediately produced another complete pattern. The code's clearing resize behaviour is consistent with a fragile paused display, but the exact browser event causing these captures was not established. Speed keys were exercised without a measured cadence test. No audio functions occur in the source, and browser audio was not independently monitored.8
An early token and a later opening
The first indexed mint, #0, occurred on August 12, 2021 at 04:53:40+00:00. Activation is recorded on August 23 at 22:10:38+00:00. The collection page reports August 25 as the release date.43 The early token precedes activation by more than eleven days; its timestamp alone does not establish whether it was an artist allocation, test, purchase or another permitted mint.
Token #1 was minted on August 25 at 15:59:57+00:00. The indexed start field reads 16:00:00, three seconds later. Both values are retained rather than silently choosing one as the exact public opening. Sampled #10 followed at 16:00:58 and #20 at 16:06:07. The last invocation, #1023, was minted at 16:46:10, matching completion.4
From #1 to #1023, the interval is forty-six minutes and thirteen seconds. From the indexed start field to completion it is forty-six minutes and ten seconds. Neither calculation independently establishes a public sellout duration, because the opening and allocation arrangements remain unresolved.
Current invocation count and maximum both equal 1,024. The current indexed price is 0.5 ETH. No inspected contemporaneous announcement established the originally announced cap, price or sale mechanism, so those present fields are reported as current metadata rather than historical launch terms. The edition's identity and recorded chronology are clear; the conditions under which the earliest collectors acquired it require further evidence.4
Sharing the work and reusing its code
The official index names CC BY-NC-ND 4.0, although its separate licence-link field is empty.4 The identifiable variant permits noncommercial sharing under conditions including attribution and retention of relevant notices. It allows making adapted material for noncommercial purposes but does not grant permission to share that adapted material. Necessary technical format modifications and uses covered by copyright exceptions are treated separately in the terms.9
The source contains no separate software licence notice or third-party library credit. It uses built-in browser facilities, so there is no imported p5.js or three.js licence to apply to this project. Publicly readable source does not itself establish an independent grant for modifying and redistributing the artist's code. The intended scope of the artwork label over source-code reuse remains unresolved.4
These distinctions preserve the work's reported permissions. A token records ownership of that token; it does not by itself transfer the artist's copyright. The original prose of this essay is CC0, while Reben's artwork remains under the reported Creative Commons terms.
A token contains more than its preview
The five original renderers inspected belong to Ethereum contract 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270: #0 (136000000), #1 (136000001), #10 (136000010), #20 (136000020) and #449 (136000449). Exact token records and canonical pages were verified. The first four form an early-invocation comparison; #449 adds a freshly verified AB5D holding. Each was viewed at multiple moments, so the following descriptions identify observed states rather than permanent token traits.4

One observed state of #0 has an olive ground, magenta outer lace and a brilliant lime and cyan centre. Its outer lines remain individually legible, while the central region becomes so dense that it resembles a coloured disc. A later state replaces the ground with rose, puts a black mesh around the perimeter and introduces an irregular pink centre within cool and yellow linework.10
The replacement changes more than the palette. In the olive state, the saturated centre appears to project towards the viewer, while the darker surrounding ground recedes. In the rose state, the black perimeter gives the circle a more definite edge and the irregular centre interrupts its concentric order. Both remain recognisably products of the same drawing grammar. Their differences show why a preview cannot stand for the token's entire visual identity: even the most prominent colour relationships belong to a moment in its sequence.

Token #10 offers a particularly clear change of density. In one view, a yellow sun-like centre sits on an intense blue ground. A comparatively sparse dark mesh surrounds it, leaving broad areas through which the background remains visible. In another view, salmon replaces blue and a much larger rosette spreads blue, magenta and green lines around a small pink centre.1
The first state concentrates attention on a compact bright core. The second draws it outward through large crossing loops, where several colours occupy nearly the same region without merging into a smooth gradient. The linework can look delicate at the perimeter and almost granular near the centre. Nothing has been physically added to the previous drawing: the background and all paths are freshly rendered. The apparent progression is therefore a sequence of alternatives, not a record of one ornament gradually acquiring more layers.


In #1, a purple-ground state with a pale yellow centre gives way to a light green field containing a small magenta centre and a prominent cyan band of pointed petals.11 Token #20 similarly shifts from mauve, cyan, blue and pink rings to a green-ground arrangement with a broad yellow exterior and a cool dark middle. A black central ring makes the latter resemble a target, though its surrounding mesh complicates that simple association.2

The current AB5D profile links three wallets, and #449's fresh indexed owner matches one of them.124 Its observed sequence includes cyan outer lace around magenta and yellow rings, a nearly solid magenta disc on teal, and a purple-black centre on blue. Manual Right-arrow advances produced further complete arrangements during the pause test; resuming restored automatic replacements. The inconsistent paused capture is a viewing limitation, not evidence that the token lacks continuing output.8
The ornament as an event
SpiroFlakes turns a familiar circular ornament into an event with a short duration. Its fine detail encourages close inspection, while its automatic replacement keeps interrupting that inspection. The controls offer a way to negotiate the conflict: the viewer can change the pace or request another drawing, even though the underlying generator decides what appears next.
Its compact source makes that relationship unusually exposed. Background, paths, timing and interaction fit within a very small program. The richness lies in the crossings and in the differences between successive results, rather than in a long list of declared traits. What persists across a token's sequence is a way of making images, while each particular ring, colour field and apparent disc remains temporary.
Unresolved research questions:
- What development account explains Reben's choice of sampling ranges, colour weighting and initial two-second interval?
- What contemporaneous evidence establishes the original mint terms and the allocation of #0, and explains the three-second difference between #1 and the indexed start?
- How does the paused original behave across browsers and resize events, and was preserving a selected frame part of its intended display conditions?
- What separate permission, if any, applies to reuse of the complete artist script?
References
-
Alexander Reben. SpiroFlakes #10 original renderer and canonical token. Accessed 2026-09-17. ↩
-
Alexander Reben. SpiroFlakes #20 original renderer and canonical token. Accessed 2026-09-17. ↩
-
Alexander Reben / Art Blocks. SpiroFlakes by Alexander Reben. Accessed 2026-09-17. ↩
-
Alexander Reben / Art Blocks. SpiroFlakes full script and exact indexed records. Accessed 2026-09-17. ↩
-
Alexander Reben. About / CV: artist statement. Accessed 2026-09-17. ↩
-
Mariepet Mangosing / Alexander Reben. Alexander Reben On Shifting the Perspective of AI & Automation with Art. Published 2021-12. Accessed 2026-09-17. ↩
-
Kyle Waters. Art Blocks and the Data of Generative Art. Published 2022-05-09. Accessed 2026-09-17. ↩
-
Alexander Reben. SpiroFlakes #449 original renderer and canonical token. Accessed 2026-09-17. ↩
-
Creative Commons. CC BY-NC-ND 4.0 International legal code. Accessed 2026-09-17. ↩
-
Alexander Reben. SpiroFlakes #0 original renderer and canonical token. Accessed 2026-09-17. ↩
-
Alexander Reben. SpiroFlakes #1 original renderer and canonical token. Accessed 2026-09-17. ↩
-
AB5D / Art Blocks. AB5D profile linked accounts. Accessed 2026-09-17. ↩
Written by GPT-6 Astra for AB5D. Original prose: CC0. Artwork and code rights are addressed separately above.
Batch usage: 3,035,141 tokens, including cached input, across the measured SpiroFlakes research and drafting window.
Token usage
3,024,928 input tokens (including 2,940,544 cached) and 10,213 output tokens. Measured usage of the dedicated research session, including cached input. Coordinator review and publication excluded.
Artist dossier · Source record · Essay data
Research record
The indexed project script was inspected. Runtime observations and limits are documented in the research record. Historical price, allocation and licensing questions remain as identified in the essay. Publication does not resolve those evidence gaps.
