Field record — observed, then inferred
Variant: Observed, then inferred
A pottery profile drawn the way an archaeologist draws one: the rim fragment that survives is traced in a solid line, a clay mark is set where it breaks off, and only after a full beat does the rest of the bowl appear — thinner, greyer, dashed — projected down to the base and mirrored across the axis. Evidence and inference get different line grammar and different timing: the trace is ease-in-out because a pen is following a real edge, the projection is ease-out because it is thrown from what is known and settles where the model says it should. The 384ms pause at the break is the honest part; it lets the reader see the evidence end before the guess begins. The dashes are fixed on the path and a mask does the drawing, so they extend without crawling. Reach for it on forecast tails, interpolated gaps, AI-completed fields — anywhere a line is part measured and part assumed. Under reduced motion the drawing is complete and only the inferred half breathes.
The source
A complete HTML document. Paste it into an empty .html file and it plays — no stylesheet, script, font, or image to fetch.
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>Field record — observed, then inferred</title>
<style>
:root {
/* Neutrals — fixed, never remixed. */
--ivory: #faf9f5;
--paper: #ffffff;
--oat: #e3dacc;
--slate: #141413;
/* Knobs — the remix API. A knob exists iff declared here and used below. */
--hue: 38.8;
--lift: 0;
--chroma: 1;
--speed: 1;
/* Accents — derived from the knobs with the family's offsets baked in,
so one hue roll moves them together. Defaults reproduce the original hex. */
--clay: oklch(calc(0.6724 + var(--lift)) calc(0.1308 * var(--chroma)) var(--hue));
--clay-d: oklch(calc(0.5797 + var(--lift)) calc(0.127 * var(--chroma)) var(--hue));
--olive: oklch(
calc(0.6118 + var(--lift)) calc(0.0713 * var(--chroma)) calc(var(--hue) + 88.3)
);
--rust: oklch(
calc(0.5408 + var(--lift)) calc(0.1357 * var(--chroma)) calc(var(--hue) - 10.3)
);
--_muted: color-mix(in srgb, var(--slate) 46%, transparent);
/* Clock — 4.8s: 1.2s tracing what survives, a 384ms beat at the break,
then 768ms and 960ms projecting the rest, and a long hold on the
finished drawing before the ink lifts. */
--cycle: 4.8s;
--t: calc(var(--cycle) / var(--speed));
--ease-out: cubic-bezier(0.23, 1, 0.32, 1);
--ease-in-out: cubic-bezier(0.77, 0, 0.175, 1);
}
* {
box-sizing: border-box;
margin: 0;
padding: 0;
}
body {
min-height: 100vh;
display: grid;
place-items: center;
background: var(--ivory);
font-family:
system-ui,
-apple-system,
'Segoe UI',
Roboto,
sans-serif;
color: var(--slate);
}
.sheet {
width: min(290px, 86vw);
height: auto;
display: block;
}
/* The axis is drawn before anything else and never moves: the whole
reconstruction is a bet that the pot was turned on it. */
.axis {
fill: none;
stroke: var(--oat);
stroke-width: 1;
stroke-dasharray: 8 3 1.5 3;
}
/* The sherd itself. It is on the sheet the whole time — the drawing
changes, the evidence does not. */
.body {
fill: none;
stroke: var(--oat);
stroke-width: 9;
stroke-linecap: round;
}
.ink {
animation: ink var(--t) infinite;
}
.observed {
fill: none;
stroke: var(--slate);
stroke-width: 2.25;
stroke-linecap: round;
stroke-dasharray: 100 200;
animation: observed var(--t) infinite;
}
/* Inference keeps its own grammar: thinner, greyer, dashed. The dash is
a fixed pattern on the path and a mask does the drawing, so the dashes
stay put while the line extends — animating the dash pattern itself
would make them crawl, which reads as motion along the wall. */
.inferred {
fill: none;
stroke: var(--_muted);
stroke-width: 1.5;
stroke-dasharray: 4 3.5;
}
.reveal {
fill: none;
stroke: var(--paper);
stroke-width: 6;
stroke-dasharray: 100 200;
}
.reveal-infer {
animation: infer var(--t) infinite;
}
.reveal-mirror {
animation: mirror var(--t) infinite;
}
/* The break mark: the only accent in the file, set exactly where the
evidence stops, so nobody has to guess which half is the guess. */
.break {
fill: none;
stroke: var(--clay-d);
stroke-width: 2;
stroke-linecap: round;
transform-box: fill-box;
transform-origin: center;
animation: break var(--t) infinite;
}
.legend {
fill: var(--_muted);
font-family: inherit;
font-size: 11px;
font-weight: 600;
letter-spacing: 0.14em;
}
.key-observed {
stroke: var(--slate);
stroke-width: 2.25;
}
.key-inferred {
stroke: var(--_muted);
stroke-width: 1.5;
stroke-dasharray: 4 3.5;
}
/* The pen traces a real edge, so it gets ease-in-out: it sets down,
travels, and lifts at the break. */
@keyframes observed {
0%,
4% {
stroke-dashoffset: 103;
animation-timing-function: linear;
}
5% {
stroke-dashoffset: 100;
animation-timing-function: var(--ease-in-out);
}
30%,
95% {
stroke-dashoffset: 0;
animation-timing-function: steps(1, end);
}
96%,
100% {
stroke-dashoffset: 103;
}
}
@keyframes break {
0%,
30% {
opacity: 0;
transform: scale(0.6);
animation-timing-function: var(--ease-out);
}
34%,
95% {
opacity: 1;
transform: scale(1);
animation-timing-function: steps(1, end);
}
96%,
100% {
opacity: 0;
transform: scale(0.6);
}
}
/* Inference is projected, not traced: it leaves fast and decelerates
into place, and it waits a full beat after the break before it starts.
That beat is the honest part — a reader sees the evidence end before
the guess begins. */
@keyframes infer {
0%,
38% {
stroke-dashoffset: 100;
animation-timing-function: var(--ease-out);
}
54%,
95% {
stroke-dashoffset: 0;
animation-timing-function: steps(1, end);
}
96%,
100% {
stroke-dashoffset: 100;
}
}
/* The mirror is revealed from the axis up, starting as the left
projection arrives there: one gesture folded across the axis, not a
second line appearing out of the empty side of the sheet. */
@keyframes mirror {
0%,
52% {
stroke-dashoffset: 100;
animation-timing-function: var(--ease-out);
}
72%,
95% {
stroke-dashoffset: 0;
animation-timing-function: steps(1, end);
}
96%,
100% {
stroke-dashoffset: 100;
}
}
@keyframes ink {
0%,
86% {
opacity: 1;
animation-timing-function: var(--ease-out);
}
93%,
96% {
opacity: 0;
}
100% {
opacity: 1;
}
}
@media (prefers-reduced-motion: reduce) {
/* No line travels. The drawing is complete from the start and only
the inference breathes, slowly — the evidence holds still and the
guess is the part that is allowed to waver. */
.ink,
.observed,
.reveal,
.break {
animation: none;
stroke-dashoffset: 0;
}
.guess {
animation: doubt calc(var(--t) * 1.5) var(--ease-in-out) infinite;
}
@keyframes doubt {
0%,
100% {
opacity: 0.4;
}
50% {
opacity: 1;
}
}
}
</style>
</head>
<body>
<svg
class="sheet"
viewBox="0 0 260 162"
role="img"
aria-label="A pottery profile drawing. The surviving rim fragment is traced in a solid line, a mark is set where it breaks off, and after a pause the rest of the bowl is projected in a lighter dashed line and mirrored across the vertical axis"
>
<defs>
<mask id="m-infer" maskUnits="userSpaceOnUse" x="0" y="0" width="260" height="162">
<path
class="reveal reveal-infer"
pathLength="100"
d="M 70.94 90.57 C 80.82 110.11, 95.5 123.8, 112 126 L 130 126"
/>
</mask>
<mask id="m-mirror" maskUnits="userSpaceOnUse" x="0" y="0" width="260" height="162">
<path
class="reveal reveal-mirror"
pathLength="100"
d="M 130 126 L 148 126 C 178 122, 202 80, 202 34"
/>
</mask>
</defs>
<path class="axis" d="M 130 18 V 138" />
<path class="body" d="M 58 34 C 58 54.7, 62.86 74.59, 70.94 90.57" />
<g class="ink">
<path class="observed" pathLength="100" d="M 58 34 C 58 54.7, 62.86 74.59, 70.94 90.57" />
<g class="guess">
<path
class="inferred"
mask="url(#m-infer)"
d="M 70.94 90.57 C 80.82 110.11, 95.5 123.8, 112 126 L 130 126"
/>
<path
class="inferred"
mask="url(#m-mirror)"
d="M 202 34 C 202 80, 178 122, 148 126 L 130 126"
/>
</g>
<path class="break" d="M 65.59 93.28 L 76.29 87.86" />
</g>
<path class="key-observed" d="M 34 152 H 50" />
<text class="legend" x="56" y="156">OBSERVED</text>
<path class="key-inferred" d="M 140 152 H 155" />
<text class="legend" x="161" y="156">INFERRED</text>
</svg>
</body>
</html>
The variant family
Same idea, one axis moved. Compare them side by side before you commit.