Boston Dynamics has deployed its Spot robot at a working copper mine in Utah, equipped with Google DeepMind's latest AI reasoning model, to autonomously monitor equipment and read analog gauges with 93% accuracy. This marks a significant shift from one-off robot demonstrations to integrated, continuous monitoring systems that directly improve mining operations. The deployment at Copper One, operated by Mariana Minerals, represents the second phase of a broader initiative to build the most technologically advanced copper mining operation in the United States .

What Changed in Google's Latest AI Model for Robots?

Google DeepMind released Gemini Robotics-ER 1.6 on April 14, 2026, a specialized AI reasoning model designed to help robots understand their environment and plan complex tasks. The "ER" stands for "Embodied Reasoning," which means the model handles perception, spatial understanding, and task planning rather than controlling the robot's motors directly . This separation matters because it allows robots to use different manufacturers' low-level controllers while benefiting from the same advanced reasoning brain.

The headline capability in this new version is instrument reading. The model can now interpret analog pressure gauges, thermometers, sight glasses, and digital readouts with remarkable accuracy. In testing, the baseline version achieved 86% accuracy on gauge reading, jumping to 93% when using a technique called "agentic vision," where the model autonomously zooms into critical image regions and uses code to estimate needle positions . This is a dramatic improvement from the previous version, which managed only 23% accuracy on the same task.

"Capabilities like instrument reading and more reliable task reasoning will enable Spot to see, understand, and react to real-world challenges completely autonomously," explained Marco da Silva, Vice President and General Manager of Spot at Boston Dynamics.

Marco da Silva, Vice President and General Manager, Spot at Boston Dynamics

How Is Spot Actually Being Used at the Copper Mine?

At Copper One, Spot doesn't operate as a standalone inspection robot. Instead, it functions as a mobile sensor platform integrated directly into MarianaOS, Mariana Minerals' proprietary software system that connects all aspects of the mining operation . As Spot navigates the mine site, it continuously captures thermal, acoustic, and visual data. This information flows into the broader system, where it's combined with production, maintenance, and environmental data to power alerts, predictive models, and operational decisions.

The practical benefits are substantial. Spot can detect overheating equipment, identify gas or fluid leaks, and read analog gauges without human supervisors on-site. Rather than conducting periodic inspections, the system enables continuous monitoring. When Spot collects more data over time, MarianaOS improves its predictive models, increasing accuracy and reducing downtime . Human operators shift from routine data collection to supervising autonomous systems and making high-leverage decisions.

Why Does This Matter for U.S. Copper Supply?

The United States currently imports approximately 50% of its refined copper, despite domestic demand projected to nearly double by 2035 due to artificial intelligence infrastructure, electrification, energy storage, defense systems, and grid modernization . Restarting and scaling permitted domestic mining operations quickly, safely, and reliably has become a national priority. Technology adoption in mining has historically lagged other industries, creating a competitive disadvantage.

"The primary reason the U.S. keeps falling behind on critical minerals is our technology and software adoption lag. Delivering autonomy in mining and refining isn't about automating one task at a time; it's about building a system where every part of the operation is connected, observable, and continuously improving," stated Turner Caldwell, CEO of Mariana Minerals.

Turner Caldwell, CEO, Mariana Minerals

How to Integrate Advanced Robotics Into Industrial Operations

• Start with integration, not isolation: Rather than deploying robots as standalone tools, embed them into existing software systems so data flows seamlessly across the entire operation and informs decision-making at every level.
• Focus on continuous monitoring over periodic inspection: Autonomous systems that collect data continuously generate far more value than robots that conduct inspections on a schedule, enabling predictive maintenance and faster anomaly detection.
• Pair specialized hardware with general-purpose AI reasoning: Use robots designed for specific environments (like Spot for mining) combined with foundation models that can adapt to new tasks without requiring retraining on each specific use case.
• Prioritize data normalization and contextualization: Raw sensor data becomes actionable only when it's normalized and combined with other operational data streams, allowing the system to distinguish between normal variation and genuine problems.

What Other Robots Are Using This New AI Model?

Gemini Robotics-ER 1.6 is not exclusive to Spot. Boston Dynamics is also deploying the model on Atlas, its all-electric humanoid robot that replaced the earlier hydraulic version in mid-2024 . Apptronik's Apollo humanoid, developed with Mercedes-Benz and backed by Google, is using the model for warehouse and general-purpose labor tasks. Agility Robotics' Digit bipedal robot, which is already operating on warehouse floors at Amazon and GXO, also runs on the Gemini Robotics stack for task planning and success detection .

The model's improvements extend beyond instrument reading. On Google DeepMind's ASIMOV safety benchmark, which evaluates how well systems identify safety hazards in real-world scenarios, ER 1.6 scored 6 percentage points higher on text-based safety identification and 10 percentage points higher on video-based safety identification compared to the previous generation . For industrial deployments where liability and worker safety are paramount, these improvements are significant.

What Does This Signal About the Future of Industrial Robotics?

The Copper One deployment demonstrates that the gap between robotics research and real-world industrial application is narrowing. For years, robots have excelled in controlled laboratory settings but struggled with the variability and complexity of actual work environments. Spot reading gauges in a dynamic mining operation with 93% accuracy suggests that foundation models trained on diverse real-world data can generalize to new industrial contexts without extensive task-specific engineering .

Boston Dynamics notes that this represents a shift in how developers approach robotics applications. Rather than writing extensive custom code for each task, developers can now use natural language prompts to guide AI reasoning models, which then orchestrate robot actions through existing APIs . This approach acts as a force multiplier, allowing smaller teams to build more sophisticated autonomous systems faster than traditional programming methods would permit.

The copper mining deployment is the second in a series of technology initiatives Mariana Minerals has announced as part of its broader strategy to build a fully autonomous mining and refining operation . As more industrial companies adopt similar integrated approaches, combining specialized robots with advanced AI reasoning and unified software platforms, the economics of automation in traditionally labor-intensive industries begin to shift. The window for deploying these systems at scale appears to be opening now.