Interactive Lab + FAQ

Oct 13, 2025

Generative Radial Geometry (GRG)

A discrete–exact grammar for curves and surfaces using turns, chords, and radial projections. π is not imported as a magic constant; it emerges from dyadic steps and converges in the classical limit.

Turns-first definitions

We discretize the unit turn; dyadic refinement keeps operations exact on the mesh and drives accuracy.

$$ b \in \mathbb{N}_{\geq 0}, \qquad N = 2^{b} $$

$$ \Delta u = 2^{-(b+2)} \quad \text{(quarter sampling)} \qquad \Rightarrow \qquad \Delta u = \frac{1}{4N}. $$

For a circle of radius $R$, the inscribed polygon at $N$ turns has:

$$ P_{N} = 2 N R \sin\!\left(\frac{\pi}{N}\right), \qquad A_{N} = \frac{1}{2} N R^{2} \sin\!\left(\frac{2\pi}{N}\right). $$

Turns, chords, and dyadic rays. Crisp vectors at any zoom.

Interactive Lab

Switch presets and adjust the dyadic exponent $b$. For the star case $r(\theta)=1+a\cos(k\theta)$, the exact area is $A=\pi\!\left(1+\frac{a^{2}}{2}\right)$.

Preset Dyadic exponent b (N = 2^b) Amplitude a (star-shape) Frequency k (lobes)

N (turns)

—

Perimeter (approx)

—

Area (approx)

—

Perimeter true

—

Area true

—

Errors (|P|, |A|)

—

Notes. Circle & atlas use exact finite-step formulas (chords) and converge monotonically. Star-shape uses inscribed polygons; C-shape uses odd–even radial sweep for area (perimeter shown exact for reference).

Critical FAQ (condensed)

Q: Is GRG just numerical integration with a trick?
A: For circles with dyadic bisections steps are exact at finite $N$ (chords & projections). The limit recovers classical values; nothing is imported.

Q: Where does $\pi$ come from?
A: From the discrete perimeter/area growth of dyadic polygons. $\pi$ appears as a rotational-symmetry shadow as $N\to\infty$.

Q: Non-convex shapes “jump across the outside”.
A: We prevent that with visibility-aware schemes: (i) odd–even radial sweep; (ii) radial atlas (locally star-shaped patches with ± signs for outer components/holes).

Q: Convergence?
A: $O(N^{-2})$ for $C^{2}$ boundaries; $O(N^{-1})$ near corners (mitigated via adaptive dyadics). Inscribed perimeter/area are monotone.

Q: Why is GRG pedagogically useful?
A: You can teach Euclidean geometry from bisections and chords; radians & trig tables emerge rather than being prerequisites.

Technical Appendix — Non-Convex & Complex Parametrizations

Star-Check. If a center sees the whole boundary (star-shaped), sample $\theta$ in dyadic turns and use $r(\theta)$. Inscribed area/perimeter are monotone $\uparrow$ with $O(N^{-2})$.
Odd–Even radial sweep. For each $\theta_{k}$, intersect the ray with the boundary, sort radii $r_{k,j}$, and add inside segments $[r_{k,1},\, r_{k,2}],\ [r_{k,3},\, r_{k,4}], \dots$.
Radial atlas (± signs). Decompose into locally star-shaped patches; outer components $+$, holes $−$. Sum areas with signs and perimeters by absolute length. Same $O(N^{-2})$ patch-wise.
Robustness. $C^{2} \Rightarrow O(N^{-2})$; corners $\Rightarrow O(N^{-1})$ unless adaptively refined. Self-intersections: use winding rule instead of parity.

All computations use turns, chords, projections; operations are exact at finite dyadic steps for circular primitives and exact on the mesh for polygons.

Team

Roles are presented by responsibility area. Individual bios can be added later.

Role-based structure. Short scope, clear ownership.

Leadership & Research

Sets the research agenda and publishes specifications with reproducible examples.
Maintains the “turns-first” glossary and consistency across demos and notes.
Approves public statements and abstracts before release.

Methods & Algorithms

Designs visibility-aware procedures (odd–even sweep, radial atlas with ± signs).
Owns convergence studies and baselines for circles, ellipses, and polygons.
Runs complexity/accuracy benchmarks; documents limits and mitigations.

Systems & Frontend

Implements interactive canvases; keeps demos accessible and performant.
Maintains build tooling and test fixtures for numerical correctness.
Coordinates localization and analytics (non-invasive, anonymized).

Editorial Review & QA

Checks clarity, tone, and scope; enforces the evidence-led style.
Validates formulas and captions (MathJax + SVG semantics and accessibility).
Controls versioning of figures; ensures non-repetitive hero imagery.

Collaborations

We welcome collaborations with educators and researchers. A public contact form will be added once finalized.

Critical FAQ

Sep 14, 2025

Critical FAQ — Generative Radial Geometry (GRG)

This FAQ clarifies how GRG treats circular geometry using turns (period-1 angles) and chords in a discrete, inside-out manner. Classical constants reappear as limits; they are not assumed as inputs.

What does “without π” really mean?

We do not claim a “π-less Euclidean geometry.” In GRG, π is not imported as a primitive; it emerges from a refinement process. At finite $N$, lengths and areas are computed with algebra (products and radicals) using turns and chords—no radians or pre-baked trig tables are required. In the limit, the classical constants reappear exactly.

Notation: let $s(u)$ denote the sine in turns, i.e. $s(u)=\sin(2\pi u)$. The chord corresponding to a step of $1/N$ turns on a circle of radius $R$ is

$$ s_N(R)\;=\;2R\,s\!\Bigl(\tfrac{1}{2N}\Bigr)\;=\;2R\,\sin\!\Bigl(\tfrac{\pi}{N}\Bigr). $$

Then the inscribed $N$-gon perimeter $P_N$ and area $A_N$ are

$$ P_N(R) \;=\; N\,s_N(R), \qquad A_N(R) \;=\; \frac{N}{4}\,s_N(R)\,\sqrt{\,4R^2 - s_N(R)^{2}\,}. $$

Consequently, π is determined by the limits

$$ \pi \;=\; \lim_{N\to\infty}\frac{P_N(R)}{2R} \;=\; \lim_{N\to\infty}\frac{A_N(R)}{R^{2}}. $$

Numerical sanity check (unit circle): $A_4=2.000$, $A_{12}=3.000$, $A_{128}\approx 3.1403$ — monotone convergence with error $O(N^{-2})$.

Isn’t this just numerical integration with a different skin?

Radial partition: pieces are internal chords/projections (not rectangular strips), naturally adapted to circular/elliptic symmetry.
Closed forms at finite $N$: for many $N$ (notably dyadics) you get nested radicals from geometric bisections—exact symbolic values, not only floats.
Monotone bounds & rate: inscribed/circumscribed constructions give clean bounds; the error decays like $O(N^{-2})$.
One grammar, many shapes: the same operators scale from rough to fine and extend to circles, ellipses, and spheres without changing the language.

So what is actually new?

The novelty is methodological and pedagogical: a single, discrete, inside-out grammar (turns, chords, projections) that does not import π/radians/trig as primitives, yet recovers them in the limit. Archimedean ideas are acknowledged; GRG reframes them for modern discrete computation and teaching.

Why do some diagrams mention “90°” if GRG uses turns?

Turns ($u\in[0,1)$) are the formal unit. Occasional “90°” labels are didactic visual cues for a quarter turn. All definitions and formulas are stated in turns and internal measures.

Pedagogical & computational takeaways

Conceptual clarity: learners see π emerge from an internal process rather than being assumed.
Computational simplicity: increase only $N$ to refine accuracy; same grammar from coarse to fine.
Extensibility: the approach transfers to ellipses and spheres with minimal changes to the operators.

Rate vs. step (dyadics)

With dyadic refinement $N=2^{b}$ (where $b$ is the number of bisections/bits), the turn step and the error scale as

$$ \Delta u \;=\; \frac{1}{N} \;=\; 2^{-b} $$

$$ \text{err}(P_N),\ \text{err}(A_N)\;=\;O\!\big((\Delta u)^2\big)\;=\;O\!\big(2^{-2b}\big). $$

Interactive: inscribed polygon and convergence

Use the slider to vary $N$ on the unit circle ($R=1$). For an inscribed $N$-gon, both $P_N$ and $A_N$ are monotone increasing and stay below the true values ($2\pi$ and $\pi$).

Number of sides $N$
12

Inscribed $N$-gon; chord $s_N=2\sin(\pi/N)$. Increase $N$ to see $P_N\!\nearrow 2\pi$ and $A_N\!\nearrow \pi$.

Computed (unit circle): $s_N=\;$—, $P_N=\;$—, $A_N=\;$—. Relative errors: $\tfrac{2\pi-P_N}{2\pi}=\;$—, $\tfrac{\pi-A_N}{\pi}=\;$—.

Dyadic refinement note: with $b\in\mathbb{N}_{\ge 0}$, $N=2^{b}$ gives especially clean nested-radical evaluations; quarter sampling uses $\Delta u=\tfrac{1}{4N}=2^{-b-2}$.

Floating metallic ring casting mismatched shadows on digital planes (perspective can mislead)

Paradigm Shift — Core Framing

GRG invites a change of lens: do not chase π; see the circle. Compute with the geometry directly (turns, dyadic refinement, additive projections). Where π appears, read it as a signature of circular symmetry — not a prerequisite for calculation.

Myths & Facts

Myth: “You must use π to define or measure a circle.”
Fact: GRG generates and measures the circle via a radial sweep with additive projections. π is a shadow of circular symmetry, not the ingredient.
Myth: “Discrete means coarse.”
Fact: In GRG, discrete means targeted precision. You choose the resolution appropriate to purpose; refine dyadically when needed.
Myth: “Turns are just degrees renamed.”
Fact: Turns compose cleanly with dyadic refinement and quarter-turn sampling, matching how GRG computes.

Short Answers (for the critical reader)

Q1. If π shows up in unrelated places (Wallis, ζ(2), Euler), what does GRG say?

Where π appears, a circular/rotational structure is at work — sometimes implicitly. GRG computes from the cause (the circle), then reads π as the signature of that structure, not as a required input.

Q2. How does GRG handle accuracy without limits or π-digits?

By resolution. You pick dyadic refinement and recompute; the method is stable and uniform across refinements. Precision is a knob, not an infinite chase.

Q3. Why “turns” first?

Because one full rotation is 1, halves and quarters are exact dyadic fractions, and sampling on a quarter-turn aligns with GRG’s construction.

Turns & Dyadic (at a glance)

$$ b \in \mathbb{N}_{\ge 0}, \quad N = 2^{\,b}, \quad \Delta u = 2^{-\,b-2}\ \text{(quarter-turn sampling)}. $$

Increase b to double the resolution — same procedure, tighter result.

Go deeper

See the Paradigm Shift

Prefer a quick recap of objections? Read the myths above, then continue with the full Critical FAQ below.

GRG — Objective Evaluation & FAQ

Generative Radial Geometry (GRG): reproducible results, comparisons with classical geometry, and clear answers to common objections.

Discrete-first construction
Product-only 2D
Turns (fractions of a full turn)

Principle: verifiable radial constructions

In GRG the radius is the generative object. Circumference and area emerge from radial sums over discrete turns. No claim of infallibility: for finite N the error is controlled; in the limit we recover classical values.

Figure — Ellipse circumference error; expected order ≈ 2.

⬇️ CSV 🔗 Notebook

Figure — Sphere surface area error; expected order ≈ 2.

⬇️ CSV 🔗 Notebook

Circle — exactness at finite N

For a circle of radius R, radial sums for circumference and area are (machine-)exact for any N: the integrand is constant, so rectangle/trapezoid rules reproduce the exact integral.

⬇️ Circle circumference CSV ⬇️ Area (circle/ellipse) CSV

Methods (concise, rigorous)

2D area: A \approx \tfrac{1}{2}\sum \rho(\theta_k)^2\,\Delta\theta.
Perimeter: L \approx \sum \sqrt{\rho_k^2 + \big(\tfrac{\Delta\rho}{\Delta\theta}\big)_k^2}\,\Delta\theta.
Sphere: A \approx \sum 2\pi R^2 \sin(\phi_j)\,\Delta\phi (midpoint).
Order: O(N^{-2}) for periodic C^2 integrands (trapezoids/midpoints).

$$ \text{Turns-first:}\quad b \in \mathbb{N}_{\ge 0},\quad N = 2^{\,b},\quad \Delta u = 2^{-b-2} $$

Frequently Asked Questions

“Are you just approximating π in another way?”

Yes, but with a constructive and controlled method: some quantities (circle) are exact at finite N; for smooth periodic integrands the error decays as O(N^{-2}). π reappears as a property, not as an input.

“What is the error order?”

With the trapezoidal rule on a uniform periodic grid and a C^2 integrand, we get O(N^{-2}). See charts and CSVs.

“Does it work in 3D?”

For the sphere, midpoint latitude-ring summation converges to 4\pi R^2 with order ≈ 2. Generic surfaces require metric extensions.

“When should I avoid GRG?”

Non–star-convex shapes and complex surfaces may need multi-valued \rho(\theta) or alternative parameterizations; classical/hybrid methods can be preferable there.

“Where are the code and data?”

This demo links the CSVs of our experiments. For public replication, attach a repository (e.g., Zenodo DOI) and a notebook that regenerates all figures.