The AI that's grounded in the physical world.
One short film. No villain, no competitors named, no "what if." Math-Ground AI walks in, says what it is, and leaves.
Built on math. For the physical world.
Language AI rose on the internet of words. The world isn't made of words — it's made of forces, motion, sensors, and joules. Math-Ground AI is the physical-AI library that answers to the world, every frame.
- 00:00 Opening The world isn't made of words.
- 00:10 Physical Robot. Drone. Car. Humanoid. Physical AI answers to the world — every frame.
- 00:22 Name This is Math-Ground AI.
- 00:29 Math Every primitive — closed-form. Every envelope — derived. Every controller — Lyapunov.
- 00:40 Joules Every call emits a picojoule receipt. Before you ship. Not after.
- 00:51 Compose Eighty composable primitives. One workspace. One
cargo add. - 01:03 Live In your browser. WebGPU. Under four hundred kilobytes.
- 01:13 Close math ground dot AI.
Production
Made from the same primitives it presents.
The film is eight segments, each matched to a cinematic beat: library, physical-AI quadrant, name lockup, math overlay, picojoule counter, compose graph, WebGPU viewer, outro card.
Narration was generated by ElevenLabs (George voice, same storyteller as corridor.openie.dev). Imagery by BFL Flux 2 Pro with text-free prompts. Math and wordmark overlays rendered in Manim, composited with ffmpeg timeline `enable=` filters for the narration-synced 2×2 spotlight on segment 2.
Total production cost: ~$1.15 in API spend, ~30 minutes of wall time. Full pipeline is open — narration script, BFL prompts, Manim scenes, and assembler all live in the repo.