I’ve been thinking about a pattern I keep running into at the very start of AI development.
Before the “real” coding begins, I’m usually deep in experimentation—trying to nail a landing page or a specific look and feel. Or I’m doing heavy marketing and product research and suddenly have a whole bunch of materials.
The problem is: where do I store this stuff? I want a place to keep it all together, but I really don’t want to pollute the main part of the repo. I’ve been weighing two paths:
The Branch Strategy: Keeping it in a separate branch within the main project.
The Prefix Strategy: Creating a completely separate repo with a prefix like
design-project-name
I’m leaning toward the separate repo. There’s something about keeping the core codebase “pure” while letting the preliminary chaos live elsewhere. But I wonder if that’s just creating more silos.
How are you all handling the “pre-work” phase?
Spot on! A common approach is to explore in spike branches or a scratch repo, capture the key takeaways in a short design note, then build the “real thing” cleanly. That keeps the repo tidy and intentional while still preserving the learning, which is especially valuable in team settings, for continuity, and perhaps for the very act of knowledge transfer.