ARPA-H Deadlines·IGoR Aug 13·HEARING Aug 14

InTelluric / Knowledge / Obsidian Plasma Methodology

Methodology Note·May 2026

Deriving the Obsidian Plasma Interface

Mechanism-First Generative Design as a Technical Process

InTelluric — Los Angeles, CA


ABSTRACT

This note documents a generative design method for producing a visual language that had no stable reference class in existing model priors. The target was already known: a utility interface that appeared to emit from inside polished black glass, with plasma-like spectral behavior, CRT-like legibility, and no ordinary science-fiction console vocabulary. Direct aesthetic prompting failed because the model interpreted the request as flat graphics on a futuristic panel.

The successful process replaced description with mechanism. A hypothetical but physically grounded display lineage was constructed from adjacent technologies: gas discharge graphics, femtosecond laser volumetric graphics, laser writing in glass, liquid-crystal optical modulation, Chladni/cymatic modal fields, and physics-informed inverse design. The generated images became experimental observations from which a transferable interface grammar was extracted.

The exercise demonstrates that a missing visual reference class can be manufactured procedurally by constructing the physical mechanism that would make the desired appearance inevitable. The final artifact was not the HTML page. The HTML page was the transfer test. The core output was the causal grammar: a set of visual rules that survived translation from photorealistic image generation into ordinary web primitives.


KEY TERMS

mechanism-first promptingvisual system designgenerative AIplasma displaydielectric bulkfictional engineeringreference-class constructionaesthetic constraint synthesisprior-collapse diagnosisstaged prototypinginvariant extractiongas discharge emissionChladni modal fieldsinterface grammarphysics-informed inverse design

EXECUTIVE DECISION SUMMARY

FindingImplication
Direct style description was underconstrained.The model filled the gap with familiar science-fiction display priors.
The desired result required medium causality.Text had to appear as a luminous state inside matter, not graphics on a surface.
A staged fictional engineering lineage solved the prompt collapse.Each generation added one physical constraint and removed one failure mode.
The final aesthetic became transferable.The derived grammar could be implemented with normal HTML/CSS/SVG without retaining the fictional hardware.

METHOD PIPELINE

Six stages from perceptual specification to deployable grammar. Each stage operates on the output of the previous one.

StageOperationOutput
Target inventoryState the required perceptual behavior without naming a genre.A functional spec for the visual result.
Prior-collapse auditIdentify the imagery the model will default to and why that class is wrong.A map of failure modes.
Mechanism selectionChoose physical primitives that force the target traits as consequences.A hypothetical display architecture.
Staged prototypingRender crude, then increasingly capable versions of the device.A sequence of visible mechanism tests.
Invariant extractionExtract visual laws that survive across generations.A portable aesthetic grammar.
Implementation transferTranslate the grammar to the target medium.A conventional interface that preserves the effect.

MECHANISM SELECTION RATIONALE

Each design requirement was mapped to a physical primitive that made the requirement a consequence, not a description.

Design requirementMechanism chosenWhy it was necessary
Light must appear embedded in matter.Polished obsidian-like dielectric bulk.A near-black substrate suppresses inactive regions and makes emission read as internal, not overlaid.
Visible marks must be matter-energy events.Localized plasma / gas discharge emission.Gas discharge provides hard cores, halos, pinch points, and spectral limits absent from flat RGB UI.
The display must be programmable, not signage.Edge excitation, waveguiding, and modulated field control.Preserves neon/Nixie visual truth without fixed tube geometry.
Color must feel physical, not themed.Limited plasma-standard palette: purple, amber/red, white.Spectral constraint prevents generic rainbow sci-fi palettes.
The result must scale from text to imagery.Layered optical modulation, then modal-field excitation.Escalating mechanisms solve resolution and beam-addressing bottlenecks.
The system must feel real, not cinematic.Prototype-lab staging before polished output.Hardware clutter, exposure limits, uneven intensity, and repair history anchor the visual prior.

ITERATION SEQUENCE

The sequence was deliberately incremental. Each prompt introduced one new physical capability because the previous generation exposed a bottleneck.

Prompt mechanismRationaleFailure avoided
0Direct "futuristic plasma/obsidian display" description.Test whether the desired aesthetic existed in prior space.Revealed collapse into flat sci-fi HUDs and decorative panels.
1Real plasma reference images: filaments, sheets, glow in nearby volume.Anchor emission to observed plasma topology.Avoided soft generic neon haze.
2Gas tube embedded in quartz.Calibrate internal optical depth and trapped emission.Avoided text pasted onto a surface.
3Crude proof-of-concept: black glass, one color, uneven HELLO WORLD.Prototype framing activates lab hardware priors — imperfect optics, brackets, cables, dust, exposure limits.Avoided polished prop design.
4Multi-edge laser array with three spectral layers.Introduce color as depth/spectrum rather than theme.Avoided arbitrary RGB UI.
5Prism-expanded laser sheets with cholesteric LCP modulation.Solve beam count and resolution bottleneck while preserving optical plausibility.Avoided point-spark typography.
6RF/cymatic modal-field excitation and inverse waveform synthesis.Replace directional beam intersections with continuous modal interference.Allowed high-resolution vector/raster behavior without impossible laser quantization.
7DOOM and browser benchmarks.Force capability tests: imagery, perspective, UI hierarchy, dense text, legibility.Prevented beauty-only outputs that fail as instruments.

WHY MECHANISM BEAT DESCRIPTION

A descriptive prompt can specify appearance but cannot force causality. Mechanism-first prompting changes the sampling problem: the model is no longer asked to render a style — it is asked to render a physical apparatus with constraints.

Direct description failureMechanism-first repair
"Glowing futuristic interface" produces ornamental arcs, holograms, HUDs, and genre color palettes.Define a display substrate and emission mechanism; let visual features follow from the mechanism.
"Plasma text" becomes neon typography or lightning effects.Constrain emission to embedded channels, gas spectra, localized breakdown, and optical depth.
"Volumetric display" becomes hologram or floating screen.Use a dielectric bulk whose active content exists inside the material.
"High-tech UI" produces decorative complexity.Use a utility benchmark: readable text, status panels, navigation, tables, and diagnostic diagrams.
"Future style" has no failure modes.Prototype staging creates flaws: uneven intensity, beam leakage, edge glow, dust, hardware limits.