Selling shovels in a gold rush

In California, thousands dug. Most found nothing. But a few companies made a killing. Levi Strauss sold trousers. Samuel Brannan sold shovels. They didn't need to find gold to get rich, they only needed other people to believe in becoming prospectors.

The pattern repeats because human nature repeats. Cryptocurrency: most speculators lost, but Nvidia sold GPUs and Coinbase took a cut of every trade. Now, large language models: a thousand startups race to build the defining application. Most will fail. But the ones selling compute, tooling, and infrastructure profit from every attempt regardless of outcome.

I find this framing useful not as a business strategy, but as a way of thinking about where we should position ourselves when it comes to designing solutions for real problems.

In healthcare, the gold rush equivalent is the shiny pilot. Everyone wants to launch the new model of care — the innovative program, the thing you can announce. And sometimes those work. But what I've seen more often is that the programs that succeed at scale aren't the most creative ones. They're the ones that had measurement infrastructure underneath them. That's the shovel: Someone had already figured out how to track utilisation patterns, how to evaluate cost-effectiveness, how to show a funder that the thing actually worked.

Without this layer, even good ideas die quietly when the pilot funding runs out and no one can prove the impact.

I think about this with software too. The current wave of AI tooling is full of people racing to build the cleverest application. But the part most of them skip is trust infrastructure: How do you verify that an AI-generated output is actually faithful to what a user provided? How do you keep data local when the default assumption is that everything goes to a server?

Most challenges in life aren't glamorous problems that need gold. Many of them are just in need of shovels. The discipline isn't predicting which gold strike will pay off. It's noticing – and then building – what every prospector needs, regardless of whether they find anything.