Geometric Computation Engine for Operations Intractable to Traditional Methods
Antikythera Digital is a spatial differentiation engine that treats calculus as a measurement operation rather than a symbolic procedure. Instead of chain-rule derivation through nested expressions, the engine preloads signed distance fields (SDFs) as geometric manifolds and probes them directly for gradients, curvature, geodesics, and arbitrary differential operators.
The key insight: When you embody computation in geometry, operations that are exponentially expensive symbolically become constant-time spatial queries.
The Antikythera mechanism (c. 100 BCE) is the oldest known analog computer—a geared device that computed astronomical positions through mechanical relationships. It didn't calculate symbolically; it encoded knowledge in physical alignment.
Antikythera Digital extends this philosophy:
Computation is alignment. Turing machines achieve it through recursive meta-instructions. We achieve it through direct geometric embedding.
grug think: math hard when many symbols
grug discover: shape already have math inside
grug poke shape -> shape tell grug answer
no need calculate -> shape IS calculation
grug happy
Traditional approach:
- Write complicated function
- Apply chain rule many times
- Hope no mistakes
- Get gradient
Antikythera approach:
- Build shape
- Poke shape
- Shape gives gradient
- Done
The engine implements spatial differentiation on implicit surfaces represented as signed distance fields. Rather than computing derivatives through symbolic differentiation or automatic differentiation (AD), we exploit the geometric structure of SDFs:
- Gradient: ∇f(x) emerges from finite differences on the preloaded field
- Curvature: κ(x) = ∇²f(x) / |∇f(x)| computed via Laplacian probe
- Geodesics: Integral curves on the manifold, computed via streamline integration
- Arbitrary Differentials: dⁿf/dxⁱdyʲdzᵏ constructed from Vandermonde stencil coefficients
The key advantage is that composition complexity does not increase query cost. A boolean union of 1000 primitives has the same gradient-probe cost as a single sphere—the geometry is preloaded, and differentiation is measurement.
| Command | Description |
|---|---|
/init |
Initialize the Antikythera map |
/gear <name> <type> [params] |
Cast an SDF gear from library |
/sdf "expr" [params] |
Parse user-defined SDF expression |
/probe <gear> <x> <y> <z> |
Probe SDF value at point |
/gradient <gear> <x> <y> <z> |
Compute gradient vector |
/normal <gear> <x> <y> <z> |
Surface normal at point |
/curvature <gear> <x> <y> <z> |
Mean curvature value |
/laplacian <gear> <x> <y> <z> |
Laplacian (divergence of gradient) |
/divergence <gear> <x> <y> <z> |
Vector field divergence |
/flow <gear> <x> <y> <z> <steps> |
Trace streamline from point |
/levelset <gear> <iso> <x> <y> <z> |
Project to isosurface |
/geodesic <gear> <start> <end> <steps> |
Compute geodesic path |
/union <g1> <g2> <result> |
Boolean union |
/intersect <g1> <g2> <result> |
Boolean intersection |
/subtract <g1> <g2> <result> |
Boolean difference |
/blend <g1> <g2> <k> <result> |
Smooth blend operation |
/morph <g1> <g2> <t> <result> |
Linear morph between shapes |
/diff <gear> <spec> <x> <y> <z> |
User-defined differential operator |
/list |
List all gears in machine |
/throttle <value> |
Set compliance slack (h value) |
/dump <gear> |
Export gear parameters |
/quit |
Exit CLI |
sphere(radius)box(width, height, depth)torus(major_radius, minor_radius)cylinder(radius, height)gyroid(period, thickness)schwarz(period, thickness)twisted_torus(major, minor, twist)
- Boolean union, intersection, subtraction
- Smooth blending with controllable smoothing factor
- Morphing between arbitrary SDFs
- Custom SDF expressions with JIT compilation
- Arbitrary differential operators: dⁿf/dxⁱdyʲdzᵏ
# Clone repository
git clone https://github.com/marshalldavidson61-arch/antikythera-digital.git
cd antikythera-digital
# Run the engine (requires Julia 1.9+)
julia antikythera_diff_engine.jl# Initialize machine
/init
# Create a sphere
/gear mysphere sphere 1.0
# Probe its surface
/probe mysphere 0.5 0.0 0.0
# Output: -0.5 (inside sphere by 0.5 units)
# Get gradient
/gradient mysphere 0.5 0.0 0.0
# Output: [1.0, 0.0, 0.0]
# Create another shape and blend
/gear mybox box 1.0 1.0 1.0
/blend mysphere mybox 0.3 blended_shape
# Compute curvature on blended shape
/curvature blended_shape 0.0 0.0 0.0Operations marketed as "quantum-required" may simply be problems framed incorrectly. When you embody computation in geometry:
- Optimization landscapes become surfaces to probe
- Gradient descent becomes streamline following
- Constraint satisfaction becomes admissibility region design
The engine demonstrates that certain computational hardness assumptions depend on algorithmic framing, not fundamental limits. A geometric computer doesn't "solve" NP-hard problems—it makes them irrelevant by construction.
antikythera-digital/
├── README.md # This file
├── WHITEPAPER.html # Comprehensive technical documentation
├── antikythera_diff_engine.jl # Main engine (1,743 lines)
├── test_antikythera.jl # Test suite (537 lines, 83 assertions)
└── docs/
└── images/ # Diagrams and flowcharts
MIT License - See LICENSE file for details.
GrugBot420 / Bindboss
Listening to nature before theorizing.
- The original Antikythera mechanism builders (c. 100 BCE)
- Every craftsperson who understood that fit matters more than exactness
- Nature, for having answers without being asked