A recipe is a list of instructions with the reasons taken out.

For years I cooked by feel, the way most people do. I’d swap things, cut the sugar, use whatever was in the cupboard, and it mostly worked, and when it didn’t I learned by doing it again. Nothing wrong with that.

What’s changed is the conversation I now have before I change anything. I’ll take a recipe and work through it with an AI, ingredient by ingredient, asking what each one is actually for. Not what it is (that I know), but what work it does and how and why. Take sugar, it does more than sweeten. It holds moisture, it browns, it softens the crumb. So when I want less of it, I already know what I’m taking out and what I need to put back to compensate.

And I love the cascade. Say I want to leaven a dough with sourdough instead of baking powder. That changes the water in the dough, which changes the flour, which changes the timing. One swap quietly moves an assumption that the rest of the recipe was resting on, and the whole thing has to rebalance.

Which is, funnily enough, the same thing I do in my actual work. Find what each part is doing, surface the assumption it’s resting on, and then you can start tweaking on purpose rather than by luck.

I’ve been tweaking recipes for years without that conversation, and it was fine. Now I make better choices faster, and I understand my own cooking more, which bit of flavour is doing which job and how exactly. It’s become one of my favourite ways to use AI, and one of the most fun.

✦