PUDA: AI-Native Hardware for Self-Driving Labs

The introduction of a new hardware framework, known as Physical Unified Device Architecture (PUDA), is set to enhance self-driving laboratories (SDLs). This innovative system replaces conventional graphical user interfaces designed for human interaction with a command-line runtime environment, allowing AI agents to observe, orient, decide, and execute experiments while ensuring deterministic, atomic, and auditable hardware operations. PUDA operates without a graphical interface, utilizing discoverable command-line interfaces and JSON protocols through a distributed messaging network. It systematically organizes command logs, protocols, runs, samples, and measurements into an AI-native data structure linked by timestamps and run identifiers, ensuring data provenance from protocol submission to hardware response and final data products. This architecture effectively decouples scientific orchestration from physical operations and data telemetry.