This week, I read about SpaceX going public. I don't follow financial markets closely, but one idea in the article stayed with me.
The argument was this: the investors who will actually gain from this gold rush are probably not the ones buying SpaceX stock. They're the ones behind the scenes, the companies building test equipment, chips, and ground stations. The article called them the shovel sellers. In a gold rush, the person selling the shovel doesn't have to bet on any particular claim. They just need the digging to continue.
I've been sitting with that image.
A word I keep bumping into
The article described Starlink wrapping the planet "like a neural network feeding the hungriest AI systems with synaptic real-time data." I noticed the borrowing. Satellite infrastructure described as synapses. A logistics network described as a brain.
That kind of language is rarely accidental. When we reach for cognitive metaphors to describe physical systems, we're usually trying to say something about the level of complexity involved, or about the aspiration. The machine that can think. The network that can feel.
But I wonder what the borrowing hides. Neural networks in the brain don't just transmit. They learn. They change. They encode experience. The satellite version does none of those things. It routes data. That's not less impressive, but it is something different. And the difference matters for what we expect from it.
What the infrastructure doesn't carry
The article's main claim was that whoever controls the infrastructure controls the market. That's probably true. It is also the part I want to think against for a moment.
In learning, the parallel would be: whoever controls the delivery of content controls what gets learned. And for a long time, that felt like a reasonable description of how education worked. Curriculum, textbook, teacher, institution. The infrastructure shaped the outcome.
AI changes the delivery side completely. Content is now frictionless. Any question answered, any explanation generated, any skill broken into steps on demand. The infrastructure of access has never been more available.
But access is not the same as learning. Something else is doing the work, or not doing it. The part that evaluates, connects, doubts, sits with difficulty, and builds a mental model robust enough to survive the next question. That part is not in the delivery system. It never was.
Access is not the same as learning. The part that makes knowledge usable is not in the delivery system. It never was.
The question behind the question
The shovel sellers in learning are the competencies that make knowledge usable: the ability to direct one's own attention, to stay with something uncomfortable long enough for it to become familiar, to notice when understanding is shallow rather than satisfied.
These don't live in any infrastructure. They don't scale. They don't route. They have to be built the slow way, and they stay with the person who builds them.
The article ends on a certain confidence: SpaceX isn't just raising money, it's activating an ecosystem. That's probably right. What I keep wondering is whether that ecosystem is also asking what kind of thinking it produces. Not just what it transmits.
May 26, 2026