Eight hand-crafted demonstrations of what one English prompt to an AI agent can produce when the AI knows how to write Glyph. Each page picks a subject anyone can recognize — a NASA mission, a heartbeat, a leopard, a sunflower, a steam engine, a pendulum, a rotary engine, an ancient gear-computer — and shows the full chain: prompt → JSON spec → byte-locked SVG → page. Designed to be enjoyed by a 5-year-old and a 70-year-old in the same sitting.
NASA's Artemis I, 25 days, three phases. Earth → outbound transit → distant retrograde orbit → return → splashdown.
data.shape: "function"
Read →
FitzHugh-Nagumo (1961) — the relaxation oscillator that produces the spike-and-recover rhythm. 60 BPM. Lub-dub.
data.shape: "trajectory"
Read →
Kipling's 1902 just-so story, Turing's 1952 chemistry. A single seed in a uniform field, the Gray-Scott PDE, and 400 steps later: spots.
data.shape: "pde-solve"
Read →
Vogel's 1979 model — every seed placed at √n and n × 137.5°. The only angle that fills a disc without leaving gaps.
data.shape: "recurrence"
Read →
The slider-crank that converts piston motion to rotation — and started the Industrial Revolution. Non-sinusoidal kinematics in closed form.
data.shape: "function"
Read →
The 1656 design that gave humanity navigation-grade time. A damped harmonic oscillator, integrated for 60 seconds.
data.shape: "trajectory"
Read →
The 1957 design where the chamber wall and the moving part are the same epitrochoid. Three explosions per revolution instead of one.
data.shape: "function"
Read →
~150 BCE — the world's oldest known analog computer. Bronze gears, ground by hand, that predicted lunar eclipses 1,600 years before the next mechanical computer.
data.shape: "function"
Read →
| Page | Glyph shape | Math | Year |
|---|---|---|---|
| Orion | function (parametric) |
Three-phase ternary x(t), y(t) | 2022 |
| Heartbeat | trajectory (ODE) |
FitzHugh-Nagumo coupled differential equations | 1961 |
| Leopard | pde-solve (RD) |
Gray-Scott reaction-diffusion on a 32×32 grid | 1952 / 1993 |
| Sunflower | recurrence (discrete iteration) |
Vogel's golden-angle phyllotaxis | 1979 |
| Machines of Wonder | |||
| Steam engine | function (parametric) |
Slider-crank piston kinematics, closed form | 1769 |
| Pendulum clock | trajectory (ODE) |
Damped harmonic oscillator | 1656 |
| Wankel rotor | function (parametric) |
1:3 epitrochoid housing curve | 1957 |
| Antikythera | function (parametric) |
Compound epicycle, 1:13 frequency ratio | ~150 BCE |
Eight pages exercise four of Glyph's six data shapes. geodesic (light bending around mass) and streamline (vector fields) live in Joy of Math.
Most viz grammars are designed for charts — bars, lines, scatter plots. They handle survey data and revenue dashboards beautifully and stop there.
Glyph is designed to handle math, physics, biology, and engineering too. The same JSON grammar that draws a bar chart can also integrate an ODE, solve a PDE, walk a recurrence, or describe a 1769 slider-crank in closed form. The output is always the same kind of object: a byte-stable SVG that an AI agent can author from English.
"Life in Glyph" is eight hand-crafted proofs of concept, in two halves. The first half — Life — picks subjects from nature: a spacecraft's path, a heartbeat, a leopard's coat, a sunflower's seeds. The second half — Machines of Wonder — picks subjects from human engineering: Watt's slider-crank (1769), Huygens' pendulum (1656), the Wankel rotor (1957), the Antikythera Mechanism (~150 BCE). Each page walks the reader through: the English prompt a user might give Claude → the JSON spec Claude writes → the SVG the Glyph compiler produces → the animated page the user sees.
None of these required code from the user. All of them are deterministic — same prompt, same render, every platform, every time. That's the contract this grammar makes.