> ## Content Index
> Fetch the complete content index at: https://www.elliotcsmith.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Selling Picks, Building Systems
- URL: https://www.elliotcsmith.com/selling-picks-building-systems/
- Published: 2026-11-03T22:12:00.000Z
- Updated: 2026-11-03T22:12:00.000Z
- Author: Elliot Smith
- Tags: product, building, AI

There's the old saying that the best thing to do in a gold rush is to sell picks and shovels. It's probably true to some extent. If you can sell the tool everyone needs then there's a fairly stable way to make some money.

What I've been thinking about lately is whether there's another angle on this line of thinking. Selling picks and shovels is good business but you're still out of work when the gold runs out. This is where the analogy starts to wear a bit thin but if you can find a way to manufacture gold rushes then you're in business for the long term.

To avoid stretching the analogy too much, I was originally thinking about this in terms of design and design systems. I've built enough products that I have what I'd call a passable eye for design but I'd never call myself a designer.

Despite that, products will usually benefit from a decent set of rules and components which prevent every page getting its own twist on how things fit together. AI development has probably made this a little more important because most models will follow rules if they're there but happily make things up if they're not.

The thought I had was that while I wouldn't claim to be a designer, nor feel overly comfortable selling ready made design systems (the picks in this example), I feel like I have a pretty good idea on how to setup a design-system-system.

That sounds a little obtuse I'll admit. What I mean by that is I think I could write down the steps and build the tools and feedback loops needed to iterate on a design system. That would still need someone or something in the loop to do the actual iterations but I think with enough time or tokens a decent enough system would come out the other side.

(I will caveat here to say I have only an N=1 confirmation that this is true but the point here is not so much about design systems)

Taking this idea a bit further I think there's a need to shift our collective thinking now that the 'do the work' part of so many desk jobs is able to be highly automated. Lots of jobs are going to shift from 'do it' to 'make sure its done right'. Building the systems that create the right environment for the output to happen.

If you're in software you've probably experienced this already. There's a lot of time now being spent on harnesses, CI and testing, rules and skills. All to ensure that the code generated by agents isn't total garbage. Each new model lifts the base level of quality but also helps to reveal the things that are most in need to rules and control.

Anyone who's worked in a team environment or on a product beyond the first dozen prompts has probably started to feel the need for guard rails.

While the term 'systems thinking' is fairly overused and overloaded I think it's the closest current term we have for what this new work looks like. There's another part of me that looks at it as working on the first derivative. Picks are up stream of mining for gold but they're still in the linear flow of finding gold.

The kind of work we need to do is about moving one step into the more abstract. For a lot of people that is probably going to initially feel annoying and unproductive. Working on the system has a slow start but ensures that output has a much longer lifetime.

It's certainly a new set of skills but I am fairly confident that anyone willing to learn can move from 'builder' to 'system builder' with enough time and attention.

For now, this is mostly a collection of observations and personal experience. If you've read this far and have first hand experience with this I would love to hear from you. There's comments if you're subscribed or my email is easy enough to work out.