Turner
Caldwell

On a windy Monday in April 2026, Governor Spencer Cox stood at the edge of a copper mine in southeastern Utah, 40 miles from Moab, watching a truck drive itself. No driver. No one directing it. Just a Pronto autonomous haulage system, wired into something called MarianaOS, doing what humans have done at mines for a century - only faster, cheaper, and continuously.

The person who put that truck there is Turner Caldwell. He's the Co-Founder and CEO of Mariana Minerals - a company that arrived at the mine site with software engineers, ML researchers, and chemical engineers from BASF and ExxonMobil, rather than the usual mining veterans. Caldwell spent nine years at Tesla, ascending from factory floor design to leading the entire metals and minerals team, deploying billions of dollars across lithium and battery facilities globally. Then he walked away to build something harder.

What he built - Copper One in San Juan County, Utah - is now the world's first mine running autonomous systems across all three operational domains simultaneously: mining, refining, and capital project execution. All of it coordinated by MarianaOS, Mariana's proprietary platform. Not a demo. A live mine, shipping copper cathode.

Autonomous haul trucks dispatched by MineOS. Sandvik AutoMine drilling without an operator. Boston Dynamics' Spot robots patrolling underground for inspection. Real-time sensor feeds feeding PlantOS, which adjusts refinery chemistry using machine learning. The result: 50% lower mining costs, 30% lower refining costs, compared to conventional operations.

The path that got Caldwell here started not in Silicon Valley but in Australia - or at least, the idea of it. "The day that I graduated from college," he said on the a16z podcast, "the urge that I had was to just move to Australia and try to find a job in a mine." A Stanford mechanical engineering degree redirected him toward Tesla instead, and for nine years he moved steadily upstream through the supply chain - from factory design to battery cells to cathode materials to refining to mining itself.

Mariana Minerals is the continuation of that upstream journey, and it's a deliberate one. Traditional mining companies operate, in Caldwell's reading, the way automotive and aerospace did before their own disruption moments: entrenched teams with little incentive to question how they've always worked. "As new generations come in," he says, "they're going to do new things and challenge the status quo."

Backed by Andreessen Horowitz, Breakthrough Energy Ventures, and Khosla Ventures - $85 million in Series A funding closed in July 2025 - Mariana is building toward 10 mineral projects in 10 years. The company currently has three sites under development: Copper One (copper, Utah), a Texas produced-water lithium extraction facility with Select Water Solutions, and a third stealth project. The target is to have refined product in customers' hands before the end of 2026.

It's worth pausing on that number: 10 mines in 10 years. In an industry where the average mine takes more than 15 years just to permit and build a single project, Caldwell is proposing to run a 10x faster playbook. His bet is that software compresses the timeline - that MarianaOS is the unfair advantage that makes an impossible schedule a normal one.

The conventional mining industry has a 15-year problem. From first drill to first shipment, the average mine takes 15 years to permit, design, fund, build, and operate. Caldwell's answer to this isn't lobbying for faster permitting (though that helps) - it's eliminating every source of delay that software can touch.

CapitalProjectOS handles infrastructure planning and execution, replacing the fragmented spreadsheets, contractor gaps, and coordination failures that balloon costs and timelines. MineOS gives operators a real-time view of every machine at the mine face. PlantOS continuously optimizes refinery chemistry. All three update simultaneously, and all three are informed by the data that only Mariana, as operator rather than software vendor, actually generates.

That last point matters more than it sounds. Caldwell has pointed out that the training data advantage belongs to whoever owns the operation. Licensing software to mining incumbents doesn't give you the feedback loops. Operating the mine does. Mariana chose to be the operator, not the vendor.

The automation-and-labor question comes up everywhere Mariana talks about its approach. Caldwell's position: autonomous systems will expand the talent pool, not shrink it, because they allow additional mines to operate despite workforce constraints. More mines, same or fewer workers per mine, net more jobs in the sector. It's an argument that's easier to make when you're the one operating the mine and the robots simultaneously.

The broader context is a U.S. that has, by Caldwell's telling, grown comfortable outsourcing the hard stuff. "The U.S. has gotten really comfortable with operating in a way where we do some things here, but we do a lot of things elsewhere." Critical minerals - copper for EV motors and grid wiring, lithium for batteries, nickel for cathodes - are exactly the things being done elsewhere. Mariana is the bet that they don't have to be.

There's a generational argument buried in all of this too. Traditional mining companies are not structureless; they're staffed by experienced people who have run operations for decades. The issue, in Caldwell's framing, is incentive. Those teams have optimized for what works, not for what could work. New entrants - Mariana-types - have no legacy system to protect. "As new generations come in," he says, "they're going to do new things and challenge the status quo."