# Haywire Café  by Jess Hewitt

## The work and its context

Haywire Café makes pattern look elastic. Checks, dots and rings occupy a square field, while bright contours stretch across them, pinch into narrow necks and open into bulbous pockets. The regular pattern often remains visible inside a differently coloured region. That continuity encourages the eye to imagine transparent material, although the picture offers no dependable division between a surface and whatever lies behind it. Motion keeps revising the proposed depth. What briefly resembles a folded sheet can become an opening, a ribbon or a flattened patch of colour. [^S05] [^S06] [^S07] [^S08] [^S09]

Jess Hewitt released the 256-work series through Art Blocks Factory in February 2022. Its numbered variations are animated compositions, not single frames intended to stand for the whole experience. In the five originals examined here, a stable colour family and pattern type establish each work's identity, while the contour relationships change. The differences between editions therefore concern both what persists and what moves. A still preview can convey colour and texture, but it cannot show the repeated renegotiation of the picture's apparent surfaces. [^S01] [^S05] [^S06] [^S07] [^S08] [^S09]

The title's suggestion of disorder fits the viewing experience without supplying a literal subject. These images do not depict a café, and the code does not simulate liquid poured over a table. Their interest lies in making a relatively simple patterned field support contradictory spatial readings. The apparent material is invented by the interaction of edges, colours and duration. This is an interpretation of the inspected outputs, grounded in the actual rendering procedure rather than a claim that the artist specified one physical substance. [^S02] [^S09]

## Conception and development

Hewitt's project account names Chuck Jones, Maurice Noble and Dr. Seuss as inspirations, particularly their playful treatment of perspective. It describes an exploratory workflow: randomise settings in an application, identify promising results, refine them and reproduce the developed system as a standalone work. The account connects pattern, curl noise, shadow-like transitions, scanlines, inverse vignette, colour mapping and dithering. Its value is the relationship between experimentation and a deliberately assembled image process. It does not establish that every parameter choice in the deployed version existed unchanged throughout development. [^S03]

The same account records an exhibition of editions #94 and #117 on the three-storey display at Samsung 837 during NFT.NYC in June 2022. That is an artist-reported presentation history, separate from the five originals inspected for this essay. Those exhibited editions are not silently added to the visual sample. [^S03]

The reference to animated drawing is useful because the work's humour does not require characters. A narrow junction seems to pull several broad regions into a point; an inflated lobe contracts into a small loop; a patterned plane refuses to remain flat. These are changes of visual expectation. In the collection's #184, for example, repeated lavender dots preserve an almost wallpaper-like rhythm while surrounding green and cyan forms continually undermine its stability. The result places orderly repetition and exaggerated deformation in the same perceptual space. [^S09]

## How the work was constructed

The complete indexed script is 8,415 bytes and uses p5.js 1.0.0 with a custom WebGL shader. Its JavaScript converts four portions of the token hash into unsigned integers and uses a bitwise pseudorandom procedure, discarding a thousand initial values before selecting settings. The hash-derived choices include palette, pattern, scale, rotation, skew, offsets and a seed for the field calculations. Those choices establish a repeatable configuration for a token. They are distinct from the changing time parameter that animates that configuration. [^S01] [^S02]

The canvas is square, fitted to the smaller window dimension and adjusted to an even size. The program requests thirty frames per second. When animation is enabled, a frame counter advances and supplies a time value beginning near one hundred, with one added unit per 150 rendered frames. This is a description of the code's progression, not a measurement of playback speed on every device. Rendering performance can change the wall-clock time needed to display those frames. [^S02]

Seven pattern branches provide the underlying field: cellular bubbles, checks, a circle, repeated dots, a grid, rings and stripes. Several branches transform coordinates through rotation and skew before evaluating their repeated shapes. Others use radial distance or nearby cell points. Time also enters the pattern calculations. Consequently, the ground is not simply a stored texture over which an unrelated animation is placed. It is calculated as part of each frame, with its own relationship to the common time value. [^S02]

The fluid-looking forms arise from a separate noise calculation. The shader samples a three-component noise field at positions offset on either side of each axis, takes differences between those samples and normalises the resulting vector. The offsets vary with sine and cosine of time. A ridge operation and an intermediate colour function then convert that result into channels that can be thresholded and mixed with the pattern. The inspected procedure contains no mesh of a folded sheet and no physical model of viscosity, gravity or surface tension. Its convincing folds emerge from the organisation of a two-dimensional colour field. [^S02]

This distinction explains why the apparent depth can remain unstable. Channel-dependent masks change how the initial pattern mixes with black, producing regions that can read as shadows or layers without establishing a single illuminated object. Where supported, screen-space derivatives help smooth threshold edges; the shader includes a simpler threshold fallback. The same abstract rule can therefore produce a sharply pinched boundary beside a comparatively soft transition. In the inspected works, these neighbouring treatments encourage the eye to oscillate between graphic contour and modelled volume. [^S02] [^S06] [^S07]

Colour is applied later than those initial mixtures. After scanline and radial adjustments, the shader converts the intermediate result to luminance and passes it through one of twenty-one cosine-based palettes. A palette is therefore a relationship between calculated brightness and output colour, not just a list of flat paints assigned to separate objects. The script also contains an internal palette used during the field calculation; it is not an additional twenty-second selectable output palette. This separation matters when counting the code's named structures. [^S02]

An eight-by-eight threshold arrangement supplies ordered dithering. The amount blended into the image changes with another value derived from the field, so texture does not sit uniformly over every region. An optional branch adds a dust-like spatial variation. The fine patterned grain visible in #10 and #184 can be examined alongside the smooth large contours, but the mere presence of grain should not be used to infer that the optional dust setting is active. The indexed features identify these particular examples as not dusty. [^S01] [^S02] [^S07] [^S09]

Clicking toggles animation. The code stops advancing its time counter while continuing to draw, so a pause holds the current configuration rather than selecting a new token or restarting the sequence. An actual click test on #184 produced two unchanged successive observations; a second click resumed visible deformation. The script also responds to window resizing. No audio implementation was found in the inspected program. These findings describe the original renderer tested in Chrome, not an exhaustive browser or device compatibility study. [^S02] [^S09]

## The mint

The current Art Blocks record reports 256 invocations against a maximum of 256. Its first indexed mint is dated 31 January 2022 at 15:49:48, earlier than the collection page's public release date of 11 February. Activation is recorded on 3 February at 07:17:36. These events should retain their separate names: an early token, a configuration change and a public release are not interchangeable meanings of launch. Full timestamps and their offsets remain in the evidence record. [^S01]

The scheduled start reads 11 February at 17:00:00, while the second indexed mint reads 16:59:39 that day, twenty-one seconds earlier. Completion is recorded at 22:08:55. This ordering is a reason to preserve the fields rather than compress them into an apparently exact account of public-sale duration. The research does not classify every mint as public, reserved, administrative or artist-held, and it does not establish which event should begin a sellout clock. [^S01]

The observed price setting is 0.1 ETH, with a current fixed-price sales note. Neither field alone proves the originally announced price or the amount paid by each recipient. Likewise, the present maximum establishes the current configuration, not a recovered history of every edition-size announcement. The essay records the completed indexed supply while leaving those historical questions open for launch-era documents or transaction-level work. [^S01]

## Copyright and licensing

The current project record labels the artwork CC BY-NC-SA 4.0 and supplies no separate terms URL. The named Creative Commons deed permits sharing and adaptation subject to attribution, a noncommercial condition and ShareAlike requirements for adaptations. The linked deed explains those conditions; the index observation identifies the label attached to this project. A complete history of earlier artwork terms has not been reconstructed. [^S01] [^S10]

The p5.js 1.0.0 dependency has an LGPL 2.1 licence in its versioned repository. That dependency licence does not establish a separate reuse grant for Hewitt's complete artist program. No blanket artist-code licence was identified in the inspected project material. The code was read for analysis and executed through the official renderer; its complete source is not redistributed with this essay. Original essay prose is CC0, while artwork and software permissions retain their separate boundaries. [^S02] [^S04]

## Close reading of five outputs

The sample comprises editions #0, #1, #10 and #20, selected as an early-edition comparison, plus #184 from the verified AB5D collection. Their full token identifiers are 262000000, 262000001, 262000010, 262000020 and 262000184 on contract 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270. Each original generator was inspected in more than one animated state. These observations support comparisons among particular works, not estimates of the frequency of their features throughout the edition. [^S05] [^S06] [^S07] [^S08] [^S09]

In #0, a large orange-yellow circle provides an unusually firm point of recognition against dark blue and teal. Acid yellow and lime contours curl across the field, bordered by red-brown bands that suggest depth. Between the observed states, a tall loop and a narrow rib give way to a green mound and a more elongated opening toward the lower right. The central circle remains recognisable through this activity. Its relative simplicity makes the surrounding changes feel more unruly: the viewer has something stable enough to measure them against, even though stability here is a comparative visual impression. [^S05]

Edition #1 replaces that broad circular anchor with a busy diagonal checker pattern. Pale mint and near-black checks remain visible through blue and cyan ribbons, while larger light patches occupy the sides. Near the centre, an elongated dark terminal becomes a more open loop between observations. Lower in the image, narrow contours crowd together and then spread into wider coloured regions. The checks keep supplying a repeated unit against which those expansions can be read. Their persistence makes the blue shapes appear permeable, as though patterned fabric were visible through tinted material. [^S06]

That imagined transparency is not uniform. A pale cyan lobe can look comparatively opaque, while the neighbouring dark blue band lets the checker rhythm dominate. The eye keeps assigning different material properties to adjacent zones without receiving a consistent object to explain them. This is where the animation becomes more than decoration. Motion exposes how provisional the material reading was: a form recognised as a ribbon can widen into a field that no longer behaves like a separate strip. The field never needs to resolve into an illustration. [^S06]

In #10, large skewed checks in orange-red and black support warm yellow lobes. A descending, S-like dark region threads the centre, accompanied by fine nested contours. Compared with #1, the larger checks make the diagonal rhythm more emphatic and the broad pale areas more expansive. Between states, a dark opening within the right-hand yellow lobe changes and a pale region spreads across the middle-right edge. The warm palette does not simply recolour #1: scale and the placement of broad versus narrow passages change how the eye moves across the square. [^S07]

Edition #20 brings pale blue and near-black into a field of curved repeating bands. Tall nested contours gather at upper right, while a large rounded light patch occupies the lower left. A smaller elliptical opening near the lower right changes shape and position between the inspected states. Because both the patterned field and the apparently overlaid forms curve, separating ground from figure becomes especially difficult. The image can seem to be one continuous undulating surface, then split again into several overlapping regions. [^S08]

AB5D's #184 sets repeated lavender ovals against black-purple. Mint, green and cyan regions stretch over them, with lavender also appearing near the edges. In one observed state, a broad green fold tapers to a pinched point left of centre. Later, the green region expands and a prominent junction appears nearer the upper centre. The oval pattern remains legible across contrasting zones. Its rhythm acts like a measure laid across the deformation, showing how far the eye's imagined surface has departed from the regular field that still supports it. [^S09]

Pausing #184 changes the task of looking. Held still, the image invites inspection of fine dithering, nested outlines and the exact points where a purple oval crosses a green boundary. Once resumed, those local relationships begin changing again, and no single contour can hold attention in quite the same way. The tested control therefore offers two legitimate modes of encounter: examining one configuration and following the system through successive ones. Neither mode alone supplies the whole work. [^S09]

## Significance and limitations

Haywire Café's distinctive achievement is the coexistence of recognisable pattern and unstable material appearance. It does not discard order in pursuit of movement. It uses order as the condition that makes deformation visible. Across the inspected works, differences in repeated motif, scale and palette alter the tension between a legible field and contours that seem to pull it out of shape. The code's sequence of masks, brightness adjustments, colour mapping and texture provides a concrete account of how that tension is made. [^S02] [^S05] [^S06] [^S07] [^S08] [^S09]

The available evidence supports an account of the deployed system and five particular encounters with it. Further research should recover launch-era price and edition announcements, classify mint transactions before making a public sellout claim, investigate the history of artwork and artist-code permissions, and compare the indexed script with a pinned chain retrieval. Playback also deserves testing across different graphics environments. Those tasks would strengthen the preservation record without replacing the essential act of watching how each composition changes. [^S01] [^S02]

[^S01]: Art Blocks. [Haywire Café project record](https://www.artblocks.io/collection/haywire-cafe-by-jess-hewitt). Accessed 2026-09-19.
[^S02]: Jess Hewitt. [Haywire Café indexed artist script](https://artblocks-mainnet.hasura.app/v1/graphql). Accessed 2026-09-19.
[^S03]: Jess Hewitt. [Haywire Café: inspiration, process and exhibitions](https://jesshewitt.net/art/haywire-cafe/). Accessed 2026-09-19.
[^S04]: p5.js contributors. [p5.js 1.0.0 licence](https://raw.githubusercontent.com/processing/p5.js/1.0.0/license.txt). Accessed 2026-09-19.
[^S10]: Creative Commons. [CC BY-NC-SA 4.0 deed](https://creativecommons.org/licenses/by-nc-sa/4.0/). Accessed 2026-09-19.
[^S05]: Jess Hewitt. [Haywire Café #0](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/262000000). Accessed 2026-09-19.
[^S06]: Jess Hewitt. [Haywire Café #1](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/262000001). Accessed 2026-09-19.
[^S07]: Jess Hewitt. [Haywire Café #10](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/262000010). Accessed 2026-09-19.
[^S08]: Jess Hewitt. [Haywire Café #20](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/262000020). Accessed 2026-09-19.
[^S09]: Jess Hewitt. [Haywire Café #184](https://www.artblocks.io/token/1/0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270/262000184). Accessed 2026-09-19.