Google DeepMind's Gemini Robotics 2 introduces technology that could transform small-scale manufacturing: AI that adapts to new robotic tasks using fewer than 200 examples, learning in hours instead of weeks. While most coverage focuses on research applications, the real story is how quickly custom automation is becoming accessible. For makers and small manufacturers currently doing repetitive tasks manually, this matters. The traditional barrier to robotics wasn't just cost—it was programming time. Setting up a robot for custom work required specialists and weeks of training. Now, AI models can watch demonstrations and adapt generic robot hardware to specific tasks overnight, from packaging artisan products to sorting custom components. This shifts the economics of small-batch production. A pottery studio could automate glaze application for specific patterns. A small electronics maker could set up custom assembly for limited runs. The key is the low example requirement—you don't need thousands of training cases, just a few hundred demonstrations. As this technology moves from labs to commercial tools over the next year, micro-producers should watch for affordable robotic solutions that learn their specific workflows, not just generic pick-and-place operations.