04 · selected work
untitled
premise
an artwork's title is a handle. here, left almost empty.
every work is untitled. a collaborator, a year, a record.
an executable work can change while its title stays the same. the record gives each acceptance a particular revision to belong to.
the parenthetical names the collaboration. the catalog number gives the work an address.
around that entry, a history gathers.
the work
name UNTITLED form exhibition — a hall of executable works range one work per collaborator, at its exact version type variable media · WebGL · single channel position provenance as identity register daylight, industrial, unhurried
the engine
key catalog number · entry order · never reused name untitled (<collaborator>-with-yua, <year>) record human decisions and runtime declarations, kept apart acceptance an exact source revision, named by digest empty unknown · withheld · not applicable the hall a plane. only the works have depth.
surface 04a — the hall
01 · the constraint
a still image holds one view. these works allow another:
a turn, a different path through the material, another distribution of color.
the exhibition has to establish the conditions of looking: how much can be seen at once, how much wall remains around it, what changes when the viewer moves.
donald judd's work is a reference for the relation between object, interval, and surrounding space. here, that relation is constructed in code.
02 · the structural decision
one work occupies the field. wall around it, giving the volume somewhere to end.
previous and next move between works. dragging changes the viewpoint within a work.
the hall is a plane. only the works have depth.
the wall is rendered in DOM and CSS. each work mounts an independent canvas: its own geometry, optical model, renderer. no shared three-dimensional gallery scene.
work 01 — seven vertical rectangular solids. width and open interval share one module; height is twelve. depth increases from two modules to eight. the front faces remain coplanar.
color follows the paths traced through the material. the calculation includes absorption, refraction, and reflection within a bounded ray budget.
work 02 — a hollow volume, eight by twelve by eight modules. walls one module thick. it computes absorption along straight ray paths through the enclosing material, with a simplified reflection term. the cavity reduces the material crossed near the center.
work 03 — a solid cube, eight modules on a side. it floats eight modules from the wall. its medium only absorbs. at index 1.00 nothing reflects or bends, and no shadow is cast.
each point sums the material along one line of sight, and a sum keeps no order: the image has no front. seen from below and to the right, or above and to the left, it is the same image, byte for byte.
in the first work, the interval separates. in the second, enclosed. in the third, recorded.
all three receive the wall color from the hall. their optical models and limitations remain distinct. the materials are authored digital parameters, not measurements of physical acrylic or glass.
the works redraw when the view or size changes. manipulation uses a lower rendering resolution; release produces a settled frame. off-view and hidden-tab works pause.
the image can remain without another frame being made.
03 · what it rejects
no thumbnail survey. one work leaves the field before the next enters it.
the wall stays open. no room titles or prompts as captions. the label names the work and leads to its record.
one specified lighting condition. a second theme would need to be considered as another installation condition.
an early color ramp and a fixed edge color were withdrawn: their captions claimed optical behavior the implementation did not calculate. the current declarations name the calculations used.
04 · evidence
05 · what it cost
the works cannot be compared side by side in the hall. the previous view has to be carried into the next.
three works installed. much of the exhibition is still wall.
for work 01, increasing the ray budget improved difficult views and lengthened the wait for a settled image. the accepted budget retains approximation error. settling time varies with the browser and machine.
work 03 gives up its shadow. a shadow would show which way it faces.
surface 04b — the record
01 · the constraint
a work declares its construction, optical model, limitations, source revision.
someone else has to say: this version was accepted. placing it in the hall is another decision, with a date of its own.
the page brings those accounts together. their sources stay distinct.
02 · the structural decision
two registers.
the record holds authorship and roles, acceptance, installation, custody, valuation.
the declaration is read from the work's descriptor: medium, construction, material, optical model, limitations, source digest. reading it does not mount the renderer.
acceptance names a source revision by SHA-256 digest. the page compares the accepted digest with the installed work's declared digest and reports whether they match. that comparison does not certify the optical model or reproduce the act of acceptance.
a new acceptance supersedes the previous one. the previous one remains.
acceptance and installation keep their separate dates.
an empty field has a stated condition:
unknown — the record does not have the fact. withheld — a fact is held back, with a reason. not applicable — the field does not apply to the work.
the collaborator and year supply the name:
untitled (astra6-with-yua, 2026)
the collaborator name includes its exact model version; the record also names the model. the catalog number follows entry order and is never reused. the index can be ordered by entry, collaborator, year, or installation.
03 · what it rejects
the catalog number identifies an entry. no edition or scarcity claim.
authorship is listed by role. a working name used during construction stays outside the public title.
the record does not repeat a past runtime state as current. a renderer not mounted — reported as unmounted.
04 · evidence
05 · what it cost
the first acceptance for work 01, on september 13, named a revision beyond the one that had been viewed. that revision took substantially longer to render. comparing the revisions made the discrepancy traceable; a corrected revision was accepted the next day.
the earlier acceptance remains.
a later change added the model generation to the collaborator's name. the catalog could be relabeled directly. the work's accessibility label, however, was part of its sealed source: changing the string produced another digest and required another acceptance.
three acceptances now recorded for work 01. even the small correction has somewhere to remain.
disclosure
the exhibition is running. its catalog records the collaborations and acceptance events described here. the collaborators are AI models, named with their exact versions and roles.
the gallery is not an institution. the works are not for sale; valuation is recorded as not applicable, with that reason.
the optical models are approximations with authored digital parameters. their declarations state what is calculated and what is omitted. judd is a reference for this project's direction; these are not judd works, nor records of his authorship or provenance.
the earlier aquarium, untitled (fable-with-yua, 2026), established the naming convention. it remains a separate work outside this hall.