At Dellecod, we spend a lot of time thinking about the invisible backbone of the modern world. Not the glamorous high towers of AI interfaces or the slick end-user hardware we all tap, wear, or carry — but the bedrock beneath them. Literally. Copper. Lithium. Nickel. Manganese. These critical minerals make the future possible, and yet, their journey from the ground to the grid often feels stuck in the past.
A few of us watched a talk recently that left us quiet for a while. It wasn’t polished or promotional. It was Turner, a former Tesla engineer, laying out the real story behind critical minerals — how we get them, why we’re bottlenecked, and how we might begin to change that. In tech circles, it’s easy to forget that software is still made of hardware. And hardware is made of minerals.
The supply chains for these essential materials have always felt like someone else’s problem. Mining is slow, messy, heavily regulated, and rarely seen as the frontier of software innovation. But Turner made an uncomfortable point: we can’t electrify, decarbonize, or even continue our data obsession without rethinking the way the planet delivers raw material.
Mariana — the company he’s building — wants to do just that. Vertically integrated, tech-enabled, and rooted in operational discipline, it’s aiming to turn mining and refining into a nimble, automated, intelligence-driven industry. When we heard him explain things like dynamic refining systems, AI-powered plant operations, or the need to swap static engineering assumptions for software-defined infrastructure, it resonated. Strongly. It’s not unlike what we try to do in our own field: abstract the rigid, automate the brittle, and make large systems feel more responsive to change.
And the urgency is very real. Demand for lithium is set to quadruple in a decade. Nickel is becoming the quiet pillar of battery and alloy technology. Copper demand is exploding thanks to AI data centers, EVs, and transmission lines. But meanwhile, U.S.-based mining projects still take years just to get permitted. Western projects can’t find labor, and equipment supply chains lag orders of magnitude behind China’s. The dynamic is stark: China builds a refinery in six months and employs 13,000 people to do it. The West struggles to assemble a team of 1,000 over several years.
Why does this matter for software folks? Because industrial systems — and the minerals that feed them — are quickly becoming the limiting factor of what we can build. Machine learning might define the edge of intelligence, but without aluminum and nickel, nothing ever ships out of the lab. And if we care about resilience, sovereignty, and sustainability, then waiting for someone else (or some other country) to figure it out doesn’t cut it.
Turner’s approach takes inspiration from both Tesla and DeepMind — not in their surfaces, but in their architecture. Mariana is building what amounts to plant operating systems: reinforcement learning agents that optimize chemical refinement. Construction platforms that shrink multi-year build cycles into quarters. Flexible circuits that sense and adapt to material properties instead of forcing constraint upstream. With a team of 200, they aim to do what traditionally needed 10,000.
You don’t build a better supply chain with just sentiment and slogans. You build it with tools that work under variability. You build it with systems that degrade gracefully, that can sense, plan, and act in real time. Autonomy, in the physical world, is about far more than cars. It’s about projects that scale, materials that comply, and people who don’t burn out.
It’s easy to think of energy as the constraint. But minerals are in many ways the more foundational one. They define what’s even possible. Cheap, abundant, tailored access to critical materials will shape which software gets built, where it runs, and how sustainable it is to maintain. It's one thing to have transformers running in nitrogen-cooled server halls — it's another to ask what physical world enabled that transformer to ever exist.
So we’ve been thinking about this more deeply. The inflexible parts of global infrastructure are ripe for reinvention. And not through one magic leap, but through disciplined loops of re-thinking. Mariana’s story isn’t really about mining, when you step back. It’s about believing that unsexy, hard, heavily regulated domains can still transform — if you bring the right synthesis of software realism, operational humility, and speed.
As engineers and builders, we should get more comfortable playing on this level. Sometimes the next frontier isn’t virtual. Sometimes it’s several million years below the surface. And if that world can be mapped, modeled, and modernized, then maybe everything else we care about — batteries, chips, networks, climate — gets just a bit easier to build.