The Rise of Agentic AI and 'End of Manual Work'

The push towards agentic AI is fueled by massive investment in "foundation models" for robotics. NVIDIA's Project GR00T, a general-purpose model for humanoids, aims to let robots learn skills by observing human actions, moving beyond explicit programming. This is powered by new hardware like the Jetson Thor, a computer designed specifically for running multimodal generative AI models on a robot. Humanoid developer Figure AI's collaboration with OpenAI to create next-gen AI models for its robots, backed by $675 million from investors including Microsoft, NVIDIA, and Jeff Bezos, highlights the industry convergence. The partnership leverages Microsoft Azure for the immense AI infrastructure, training, and storage required. While Figure later ended the partnership to focus on in-house AI, the initial deal underscored the trend of pairing advanced robotics hardware with leading AI model builders. In industrial automation, companies like Covariant are deploying models such as RFM-1, an 8-billion parameter transformer that gives warehouse robots human-like reasoning and language understanding. This model allows a robot to understand ambiguous commands, ask for clarification if its path is blocked, and even generate a video of a predicted action to choose the best course. The hardware is evolving rapidly to keep pace. Tesla's Optimus Gen 2, for example, is 10kg lighter and walks 30% faster than its predecessor. Its new hands feature 11 degrees of freedom with tactile sensors on all fingers, enabling delicate tasks like handling an egg without breaking it. Major research hubs are also being established to solve core challenges. Hyundai and Boston Dynamics have invested over $400 million to create the Boston Dynamics AI Institute. Its mission is to make fundamental advances in areas like "athletic AI" and "cognitive AI," focusing on creating robots that are safer and more perceptive. For aspiring engineers, this trend demands a hybrid skillset. Proficiency in C++ for performance-critical control loops and Python for its extensive AI/ML libraries like TensorFlow and PyTorch is essential. Experience with systems integration—making sensors, actuators, and software work together reliably—is becoming as crucial as expertise in a single domain.