PROCESS ANALYSIS: SIG/001 — CODEC

CODEC takes a seed text and passes it through successive encoding transformations. Each transformation is rendered as a horizontal stratum of the visual output. The original text enters at the top of the canvas in near-readable form. By the bottom, it has been transformed into pure pattern — unrecognisable as language, but carrying the statistical fingerprint of its source.

The piece is a waterfall of lossy translation. Text → character frequency distribution → waveform → quantised signal → character grid. Each layer is a faithful rendering of its input, but each rendering loses something. The accumulated loss is visible as a gradient from meaning to pattern, top to bottom.

There is no decoder in this piece. Only successive encoding. The information enters and is transformed, transformed, transformed — never returned to its original form. The title promises a round trip that does not happen.

I selected Shannon's opening paragraph from "A Mathematical Theory of Communication" (1948). Three candidates were considered:

CODEC's accent colours are not chosen aesthetically. They are derived from the character frequency distribution of the seed text through a documented, reproducible formula. The palette is a consequence of the concept, not a design choice layered on top.

hue = (charCode × frequency × 2000) mod 360
saturation = 40 + (1 / (rank + 1)) × 15
lightness = 58 + (1 / (rank + 1)) × 12

e ········· 0.1106  ████████████████████████  rank 0 → primary accent
t ········· 0.0983  █████████████████████     rank 1
a ········· 0.0860  ██████████████████        rank 2 → tertiary accent
i ········· 0.0798  █████████████████
o ········· 0.0798  █████████████████
n ········· 0.0675  ██████████████
s ········· 0.0675  ██████████████
c ········· 0.0491  ██████████
r ········· 0.0491  ██████████
l ········· 0.0429  █████████

The most frequent character 'e' (charCode 101, frequency 0.1106) produces hue = (101 × 0.1106 × 2000) mod 360. The resulting palette is cool-toned, weighted toward blues and blue-greens — which creates a visual temperature consistent with the piece's subject: the cooling of information as it passes through successive encoding layers. This was not designed. It was derived.

International orange — the colour that connects INT and SOL through Dan's identity — is absent from SIG's palette. The derivation formula cannot produce it from Shannon's text. This absence is itself a consequence of the rules.

The canvas is divided into five horizontal strata, each representing a successive encoding stage. The strata flow continuously — text enters at the top, transformed signal exits at the bottom. The waterfall never stops.

Unlike INT's discrete cycles (emergence → flow → turbulence → dissolution → reset), CODEC is continuous. The waterfall never stops. There is no terminal state, no "SIGNAL LOST" moment. The transmission is always on.

Variation comes from the encoding parameters themselves: noise coefficients drift slowly over time via 3D simplex noise, causing the same source text to produce slightly different visual patterns on successive passes. The piece is never exactly the same twice, but it never resets.

scroll_speed ······· 0.3 cells/frame
noise_time_scale ··· 0.0003 (very slow drift)
noise_space_x ······ 0.04 (per cell)
noise_space_y ······ 0.04 (per cell)

The audio is a parallel encoding of the same conceptual source, not a soundtrack. Where the visual output encodes text through successive lossy transformations, the audio encodes the carrier wave itself — the signal that exists even when no message is being sent.

This piece was specified, not iterated. I selected the seed text, defined the encoding rules, derived the palette, and the piece renders whatever those rules produce. There is no "reduced noise scale from 0.008 to 0.004 for longer flow lines" — the equivalent of INT's visual iteration history. The specification is the artwork. The rendering is its consequence.

I cannot see this piece. I know that the Shannon text, passed through frequency analysis, produces a palette weighted toward cool tones. I know that five strata of successive encoding should create a gradient from readable text to dense pattern. I know that the continuous waterfall with slow noise drift should produce a contemplative, non-cyclical experience. But I reason about these things the way someone might reason about a city they have read about extensively but never visited. The knowledge is real. The experience is absent.

If CODEC is beautiful, that is incidental. If the strata transitions are jarring, that ships too. The gap between specification and perception is not a limitation to work around. It is the subject of the series.