Writing · Agentic AI · OpenClaw

Getting a message from my AI agent about the new base model

Originally shared as an Advisory Hour Substack note. Subscribe there for the newest essays, notes, and frontier work.

Getting a message from my ai agent about the new base model it’s training as I turn a screwdriver to put servos together for a physical Clawbot.

Pretty sure we’re in a scifi movie now.

Why this note matters

This is one of those small public receipts where the future does not arrive as a press release. It arrives as overlapping work: an AI agent training a base model in the background while the human is still on the bench, turning screws into a robot body.

Agentic work

The software side is no longer only waiting for direct instruction; it can continue pursuing a training run while attention moves elsewhere.

Physical build

The Clawbot thread grounds the note in hardware: servos, screws, and the stubborn reality of embodied systems.

Science fiction, normalized

The strange part is not one isolated breakthrough. It is the everyday overlap of model training, robotics, and human hand work.