---
title: README
emoji: 🤖
colorFrom: blue
colorTo: gray
sdk: static
pinned: false
---

# Adaptive Machine Patterns
Adaptive Machine Patterns builds small, understandable embodied-AI models, and the tools to look
inside them.  Check out these:

- **[AMP Tadpole](https://huggingface.co/Adaptive-Machine-Patterns/amp-tadpole)**: a 6.5M-parameter
  vision-language-action model that drives a robot arm from three camera
  images and a plain-English instruction. It runs on a laptop CPU.
- **[AMP Toy Robotics Benchmark](https://github.com/Adaptive-Machine-Patterns/amp-toy-robotics)**:
  a light, seed-generated benchmark for small VLA policies. It needs only numpy and Pillow, its training data is unlimited
  and reproducible, and its grounding tests check that a policy really uses
  its cameras and its words. Open source (Apache-2.0).
- **[AMP Embodied AI Workbench, Demo](https://huggingface.co/spaces/Adaptive-Machine-Patterns/amp-embodied-ai-workbench-demo)**: drive the arm with
  words (or your voice), then fly through AMP Tadpole as it thinks, all in
  your browser. A Pro edition, with live tracing of larger models and of
  your own, is in development: contact us to hear first.

[adaptive-machine-patterns.com](https://adaptive-machine-patterns.com) ·
amp@adaptive-machine-patterns.com

Adaptive Machine Patterns Ltd, registered in England & Wales, no. 15618869.