Unlike LLMs that train on the existing internet, robotics lacks a pre-training dataset for the physical world. This forces companies like Encore to build a full-stack solution combining a software platform for data management with human-led operations for data collection, annotation, and even real-time remote robot piloting for exception handling.

The practical path to automating heavy industry is the difficult engineering task of retrofitting decades-old, non-digital machinery with sensors, compute, and actuators. This approach respects customers' massive existing capital investments and provides a viable path to adoption.

The true value of autonomy is not just making one truck self-driving, but creating system-level intelligence where a heterogeneous mix of machines in a port or mine can communicate and optimize operations collectively. This unlocks efficiency gains far beyond single-agent automation.

Adopting the true 'foundation model ethos' from LLMs is difficult for roboticists. It means a warehouse automation company should collect data from kitchen robots. This breadth, while seemingly unrelated, builds a generalist model that better handles the weird edge cases in the target domain than a narrowly trained specialist model.

A major hurdle in robotics is the laborious collection of real-world training data. Atlas accelerates this by creating high-fidelity simulations from sparse real-world images ("real-to-sim"), enabling rapid training and randomization of robotic policies without extensive data capture.

A major bottleneck in deploying industrial automation is that existing heavy machinery is often not 'drive-by-wire.' This necessitates the difficult and time-consuming process of installing mechanical and hydraulic actuators to allow software to control physical systems. This retrofitting reality is a core challenge for the physical AI industry.

Unlike the automotive industry where cars have standardized systems like CAN bus, forklifts lack internal standardization, even within the same model and year. This makes retrofitting for autonomy unreliable, forcing serious players like Victor Boyd's company to design and build their own forklifts from the ground up.

General-purpose robotics lacks standardized interfaces between hardware, data, and AI. This makes a full-stack, in-house approach essential because the definition of 'good' for each component is constantly co-evolving. Partnering is difficult when your standard of quality is a moving target.

Beyond general training, world models enable a "real-to-sim-to-real" workflow. A user can capture a new environment with photos, instantly create a simulation, fine-tune a general-purpose robot for a specific task within that sim, and deploy it, enabling robot onboarding to new environments in minutes.