# Synapses by Chaosconstruct ## The work and its context Synapses builds darkness by retaining movement. Fine grey marks accumulate across a white field until their crossings begin to resemble folds, ribbons and graphite smudges. What initially looks insubstantial becomes a structure through repetition. The four works examined here share that means of construction, yet distribute their weight differently: some concentrate it near a corner, some carry it through the centre, and one turns the drifting marks into a strongly symmetrical figure. [^S05] [^S06] [^S07] [^S08] Chaosconstruct, the name used by Arjit Kapoor, released the 700-work series through Art Blocks Curated in April 2021. The platform describes a study of movement and texture inspired by swarm behaviour. The inspected code makes that description more specific. Particles pursue a moving target and leave translucent segments behind them. Their accumulated histories, sometimes reflected across the canvas, become the picture. The work's organic appearance comes from these relationships rather than from a supplied photograph or drawing. [^S01] [^S02] Watching the work form is consequential. Its later state preserves the paths that preceded it, but makes their order difficult to recover. A finished image can suggest a softly modelled object; the developing image shows how that apparent object arises from many discrete movements. Synapses gives both experiences a place within one work. The [Chaosconstruct artist dossier](https://ab5d.xyz/artists/chaosconstruct/) supplies broader context for the practice. [^S02] [^S05] [^S06] ## Conception and development In his release interview with Jeff Davis, Kapoor traces his interest in generative work through Anders Hoff's systems, Matt Pearson's book and experiments in Processing. He describes moving from recreating admired work toward investigations prompted by natural phenomena. Those recollections establish an artistic route into the project, rather than a claim that its particular algorithm was inherited unchanged from any one influence. [^S03] Kapoor characterises Synapses as a meeting between movement inspired by flocking and simulated painting techniques. Graphite-like texture and the growth between beginning and end are central to his account. That emphasis helps explain why a final preview is incomplete evidence of the work's intended experience. The image is also an accumulation that the viewer can witness. [^S03] The interview's language and the code should be read together without forcing them to say the same thing. Natural behaviour provides an artistic reference; it does not imply a complete biological model. The deployed procedure can be understood and appreciated through its particular forces, limits and drawing operations. Those are sufficient to produce a convincing impression of organic development without proving that the program reproduces a named organism or neural structure. [^S02] ## How the work was constructed The indexed script uses p5.js 1.0.0. It takes an initial seed from a portion of the token hash and repeatedly transforms that value through bit shifts and exclusive-or operations. Its return value is reduced to a thousand-step numerical range. The resulting sequence supplies parameters and subsequent movement decisions. For a fixed input, this is a repeatable procedure, rather than an instruction to obtain fresh external randomness at every frame. [^S01] [^S02] The canvas is square, based on the smaller window dimension. A multiplier relates that dimension to a reference size of 3,600. Line width, forces and several positions use this multiplier. This establishes a scaling strategy, but it should not be exaggerated into a guarantee of identical geometry at every resolution: some initial-velocity expressions also include the canvas dimension within quantities subsequently multiplied again. Cross-resolution equivalence would need testing. [^S02] Two early parameters govern important distinctions. One determines the target's starting region and the reflected copies drawn with a particle's trail. The other selects a light or dark ground. On the light ground, strokes are black with low opacity; on the darker ground, strokes are white with the same alpha setting. All four comparison works have light backgrounds, so their close readings do not stand in for an inspection of the dark-ground branch. [^S02] [^S05] [^S06] [^S07] [^S08] The target has a position, velocity and acceleration. Each frame adds acceleration to velocity, limits speed and changes position. New pseudorandom forces then affect subsequent movement. At the canvas boundary, the target's position is constrained and the relevant velocity component reverses. These rules keep the pursued point within the field while allowing its path to turn. The particles respond to that moving destination rather than following a pre-drawn contour. [^S02] Each particle computes a direction toward the target, normalises it and scales it to a preferred speed. Subtracting its current velocity produces a steering vector, whose magnitude is limited before being applied. This is a pursuit mechanism. The inspected particle routine does not calculate neighbour alignment, separation or cohesion. Describing it simply as a standard flocking implementation would lose the difference between the artist's inspiration and the actual deployed procedure. [^S02] Drawing happens during the position update. The program saves the previous location, advances the particle and joins old and new locations with a thin segment. It does not clear the background on every frame. Low-opacity marks consequently accumulate, with repeated passages darkening an area. The apparent shading is the visible history of many movements, not a separate final shading pass laid over completed outlines. [^S02] Depending on the parameter, the program also draws reflected or rotated copies of those segments. Some branches reverse the horizontal coordinate, some reverse both coordinates, and others reverse the vertical coordinate. Another adds all three transformations. Those additional lines create correspondences within the image without requiring independently moving particles for each mirrored mark. The distinction is especially useful in #20, where a fluid drawing acquires a strong bilateral organisation. [^S02] [^S08] Particle counts require more caution than the apparent branches suggest. The setup starts with a value of 3,000, but its conditional reductions consult a property written as this.xp while the earlier parameter is declared as a lexical variable. Those expressions should not be silently treated as interchangeable. This inspection therefore does not publish a distribution of particle counts derived from those branches. A runtime check of the relevant context would be needed before making that claim. [^S02] The drawing has an endpoint. One narrow parameter interval calls noLoop at frame 750; the other path stops at frame 1,000. These are frame counts, not verified wall-clock durations. The work should begin drawing automatically and then settle according to its own procedure. Restarting it indefinitely would change that experience. The browser observations below establish visible development between inspected states; the source establishes the stopping rule. [^S02] [^S05] [^S06] [^S07] [^S08] ## The mint The collection page gives 5 April 2021 as the release date. The index records an earlier first mint on 3 April at 20:22:46, followed by activation at 21:36:04. The second mint and scheduled-start fields both read 5 April at 18:59:42. These describe different events. The earlier token is evidence of a mint before the stated release, not a reason to erase the platform's release date. Full timestamps remain in the evidence record. [^S01] [^S05] [^S06] The observed index reports 700 invocations against a maximum of 700. Its completion field is 7 April 2021 at 16:31:27, and its price setting is 0.1 ETH. That setting is not independently verified here as the original public mint price or the amount paid for every token. The current maximum likewise does not replace an announcement-era record of intended supply. [^S01] The interval between the scheduled start and indexed completion could be calculated, but it would not by itself establish an exact public sellout duration. The inspected evidence does not classify artist, reserved and administrative allocations. This essay retains the named chronology while leaving those transaction-level distinctions unresolved. The release interview supplies valuable artistic context, but does not fill those particular gaps. [^S01] [^S03] ## Copyright and licensing The current artwork label is CC BY-NC 4.0. The index supplies no terms link, so the corresponding Creative Commons deed is linked separately. It describes sharing and adaptation subject to attribution, indication of changes and the noncommercial restriction. This records the observed label; it does not establish a complete history of earlier terms. [^S01] [^S09] The p5.js 1.0.0 repository includes the LGPL version 2.1 licence. That rendering-library record is separate from permission to reuse Kapoor's artist script, for which no explicit independent grant was located in the inspected material. The code is analysed here without being redistributed with the essay. AB5D's CC0 dedication covers the original prose, not an expansion of either artwork or software permissions. [^S02] [^S04] ## Close reading of four outputs The comparison uses editions 0, 1, 10 and 20. Their full token identifiers are 39000000, 39000001, 39000010 and 39000020 on Ethereum contract 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270. Each canonical renderer was observed at earlier and later states without a starting click. The readings concern those observed states, not an exhaustive record of every frame or a representative survey of the edition. [^S05] [^S06] [^S07] [^S08] In #0, an early dark descent near the centre and a sweep toward the lower right establish directional pressure before the whole field becomes dense. Later, a broad lateral band crosses the upper region, while loops and converging trails complicate the lower half. The initial clarity remains partly legible beneath those additions. Watching the accumulation makes the later image feel like a record of competing routes rather than a single ribbon placed against a background. [^S05] Its apparent depth is unstable. A dark edge may suggest a fold, but adjacent pale lines reveal that the supposed surface consists of many overlapping traces. The eye alternates between following one contour and reading a shaded mass. That alternation is a concrete consequence of retaining the drawing history: the marks remain individually thin while their collective appearance becomes volumetric. [^S05] [^S02] In #1, the right side becomes heavy early. A bowed descent and gathering marks near the upper right leave much of the left field comparatively open. In the later state, a dark knot anchors the upper-right area, with long diagonal movement reaching toward the lower left and a broad arc crossing the top. This unequal distribution gives the image a different balance from #0. Its light region is active as an interval against which the crowded crossings can be measured. [^S06] The difference is not reducible to the available feature labels: both are indexed as Centre-origin, Light-background, Free-orientation works. Those categories locate broad conditions, while the particular paths determine where density develops. Comparing the two demonstrates why a feature table cannot substitute for looking. Their shared classification accommodates visibly different relations between concentration, openness and movement. [^S05] [^S06] In #10, early curves gather toward the upper left, with a hook near the top centre and a faint bowl-like sweep through the middle. Later, a more continuous network descends through the central field and opens into a large loop toward the lower right. The darkest lines make that loop seem bounded, while its interior retains a haze of finer trails. The image remains asymmetrical, but its later form distributes activity across more of the square than the early state anticipates. [^S07] In #20, matching developments on left and right make symmetry immediately conspicuous. Early side curves frame a broad pale middle. Later, crossed bands gather through the upper half, and paired central descents give the lower region a tall, narrow emphasis. The resulting figure can suggest an organic specimen, but that association is an interpretation rather than a depicted subject established by the artist. The mirroring operation explains the correspondence; the changing pursuit path supplies the irregular detail within it. [^S08] [^S02] ## Significance and limitations Synapses turns a history of movement into a drawing that can appear materially shaded. Its economy lies in the relation between a moving target, pursuing particles and retained translucent marks. The four inspected works show that this mechanism can produce sharply different balances while preserving a recognisable texture. Their formation also changes how their later states are read: what looks like a folded object is understood as an accumulation. [^S02] [^S05] [^S06] [^S07] [^S08] The sample leaves substantial variation unexplored, including dark grounds and other symmetry arrangements. Its value is close comparison rather than statistical coverage. The technical account is similarly bounded by an indexed script and browser observations, not a pinned chain retrieval or a complete cross-platform rendering audit. [^S01] [^S02] Unresolved research questions: - Recover launch-era price and edition announcements, and classify mint allocations. - Verify the particle-count condition in the original execution context. - Compare drawing behaviour and final appearance at different resolutions. - Inspect dark-ground works and additional symmetry branches. - Locate any separate artist-code licence and earlier artwork terms. - Compare the indexed script with a pinned chain retrieval. [^S01]: Art Blocks / Chaosconstruct. [Synapses project record and Art Blocks index](https://www.artblocks.io/collection/synapses-by-chaosconstruct). Accessed 2026-09-17. [^S02]: Chaosconstruct. [Synapses artist script](https://artblocks-mainnet.hasura.app/v1/graphql). Accessed 2026-09-17. [^S03]: Jeff Davis / Arjit Kapoor. [In Conversation with Arjit Kapoor (Chaosconstruct) on Synapses](https://www.artblocks.io/articles/in-conversation-with-arjit-kapoor-chaosconstruct-on-synapses). Published 2021-04-02. Accessed 2026-09-17. [^S04]: p5.js contributors / Free Software Foundation. [p5.js 1.0.0 licence](https://raw.githubusercontent.com/processing/p5.js/1.0.0/license.txt). Accessed 2026-09-17. [^S05]: Chaosconstruct. [Synapses #0](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/39000000). Accessed 2026-09-17. [^S06]: Chaosconstruct. [Synapses #1](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/39000001). Accessed 2026-09-17. [^S07]: Chaosconstruct. [Synapses #10](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/39000010). Accessed 2026-09-17. [^S08]: Chaosconstruct. [Synapses #20](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/39000020). Accessed 2026-09-17. [^S09]: Creative Commons. [CC BY-NC 4.0 deed](https://creativecommons.org/licenses/by-nc/4.0/). Accessed 2026-09-17.