Experiments
Sign inExperiments
Aaron Rhodes · things built after dark
sigma-ground is a physics engine written from the ground up in Python: it resolves matter from materials down to quarks, runs dynamics from falling feathers to N-body orbits, renders its own simulations, and exposes all of it to AI assistants as Mentat, a tool server that goes head to head with Wolfram Alpha. Every simulation and demonstration is in one place below.
Highlights
N-Body Shootout
Our N-body integrator against pure Keplerian orbits: 26 years of the Jovian system, 1.38 million steps. Energy is conserved to 1.5×10⁻⁶ and angular momentum to 4×10⁻¹⁵, and the Laplace resonance stays alive, which no Keplerian orbit can show.
Live in your browser 〰️Double-Slit
Watch the Born rule build an interference pattern one particle at a time. Turn up the observer strength D and the fringes fade: which-path information costs visibility, D² + V² ≤ 1.
Everything else
Mentat — a Wolfram Alpha rival
About 220 physics and maths tools with provenance on every number: units, constants, symbolic maths, mechanics, relativity, cosmology, optics, thermodynamics, circuits, chemistry, nuclear and materials. Try any of them live.
Results 📊Mentat vs Wolfram Alpha vs Gemini
150 physics questions, from intro mechanics to relativity, scored against ground truth. A small local model driving Mentat's tools beats both.
Recorded simulations ✨Radiance — simulated physics, rendered
sigma-ground's own WebGL2 raymarcher replaying recorded simulations: falling metals, a Mach 2 projectile, glowing iron, gear trains, a windmill, water with Fresnel reflection.
Nothing matches that. Clear the search or pick "All".