The banana-peeling robot

A robotic arm picks up a banana. It doesn't have the peel motion pre-programmed. It isn't following a rigid script for fruit manipulation. It is improvising.

This is the shift happening in robotics right now. We are moving away from factory-line automation—where every movement is defined by a developer—toward generalist systems. These robots observe a task, learn the underlying physical logic, and then apply that to new, unseen objects.

Watching a robot use a banana as a tool isn't just a party trick. It is a sign of a fundamental change in how machines interact with the physical world. In the past, you needed a different machine for every different task. Now, we are training models that understand the concept of "peeling" or "grasping" broadly.

The implication for founders is simple. The bottleneck for real-world automation has always been the fragility of the code. If a cup moved two inches to the left, the old robot failed. These new systems are built to handle the noise of the real world. They don't panic when the environment shifts.

This is exactly why we focus on agentic systems at Orbari. The goal isn't to replace the human. It is to build systems that can adapt to changing demands without needing a manual rewrite every time the conditions change.

We are still in the early days. A robot that can peel a banana is a long way from a robot that can manage a logistics warehouse or assist in specialized surgery. But the direction is clear. The intelligence is moving from the rigid instruction set to the adaptive model. And that changes the math on what is worth automating.

Source: I Saw the Future of AI in a Robot That Can Learn on the Spot