Evolution by experiment
First posted on LinkedIn.
One of the highlights of my week is spending time in my friend’s office. Learning from his team how they have been using tools like Claude, and how they are building their own harness.
What stood out was how they approached similar problems.
In ShopBack, our teams have been experimenting in their own repos on building skills, connectors, and context.
No centralised committee. Our hypothesis is that this transformation will be one that is evolutionary — where there is competition of ideas, cross-pollination, and through natural selection, the dominant approach emerges.
Seeing another team up close sharpens this. You start to see patterns that repeat.
This creates speed, but also duplication. Multiple teams working on similar problems, with slightly different approaches.
The role of “central” isn’t to decide what to build.
It is to:
- define interfaces so things can plug into each other
- make work visible so strong ideas stand out
- make it easy to copy so useful patterns spread
Spending time outside your own company becomes part of the system. It increases variation and improves selection.
Over time, a default stack should emerge — not by design, but by adoption.