Zerolaw Blog

Notes from building data infrastructure for Physical AI.

Practical perspectives on ego data quality, human demonstrations, hardware systems, and the training workflows behind embodied intelligence.

Latest articles

Research notes and engineering perspectives from Zerolaw.

ZL Capture V1 wearable egocentric capture system

Hardware selection7 min read

How to Choose Egocentric Data Collection Hardware for Robotics

Evaluate visibility, depth, IMU, timing, calibration, and exports before choosing a wearable capture system.

Read article

Multimodal engineering6 min read

RGB-D and IMU Wearable Capture for Robot Learning

What appearance, depth, and motion contribute to an episode, with practical checks for alignment and data quality.

Read article

Synchronized visual, pose, and motion data streams

Physical AI fundamentals6 min read

Physical AI vs Embodied AI: a practical distinction

The terms overlap, but they emphasize different scopes. Here is what the distinction means for systems that must learn from and operate in the real world.

Read article

Synchronized Physical AI observations, pose, and motion signals

Field guide7 min read

What is Physical AI data?

How observations, motion, actions, geometry, task context, and outcomes become trainable episodes.

Read guide

ZL Capture hardware used to illustrate robotics data collection

Field guide7 min read

How robotics data collection works

Protocols, instrumentation, field operations, quality gates, and model-ready episode delivery.

Read guide

Wearable hardware illustrating human demonstration data collection

Field guide6 min read

Human demonstration data for imitation learning

Preserving expert perception, motion, task context, action, and outcome in one episode.

Read guide

Data infrastructure6 min read

What makes Physical AI data trainable?

Why alignment, calibration, task context, and provenance matter more than raw footage volume.

Read article

Lightweight egocentric data collection hardware

Human demonstrations4 min read

Why ego data and egocentric demonstrations matter

First-person data preserves the relationship between perception, intent, and action.

Read article

High-dimensional wearable data collection hardware

Hardware systems5 min read

Designing hardware around the data contract

Sensor choices should begin with the learning objective, not a maximal specification sheet.

Read article

Build with Zerolaw

Turn your data requirements into a working collection program.

Start a conversation