The biggest Al transformation I made this year happened in my kitchen.

First, I taught it how we actually like to eat.

I didn’t want an AI meal planner that simply handed me five random recipes every week. I wanted meals my family would genuinely look forward to eating, and I wanted the recommendations to reflect how we actually cook at home.

My criteria are pretty simple. I prefer fresh ingredients, mostly homemade food, lots of flavor, reasonable cooking time, and enough variety that dinner does not start feeling repetitive. One of my favorite food series is Salt Fat Acid Heat, and I borrowed some of that philosophy for the system. I wanted recipes built around the fundamentals that make food taste good, rather than relying on heavily processed ingredients or complicated techniques.

Over time, I gave the system feedback about what worked and what didn’t. Meals we loved became part of the profile it could draw from again. Ingredients we consistently avoided became less likely to appear. I don’t like peas, for example, so the system learned that zucchini is usually a better fit for our family in dishes where peas might traditionally be used.

That learning also happens while I’m cooking. I might ask whether mushrooms would work in a dish, what kind of acid would balance a sauce, whether something would brown better in the oven, or what I can substitute when I’m missing an ingredient. Those small choices gradually add more context about the flavors, ingredients, and cooking methods I prefer.

Some of the early experiments included French onion chicken and rice, pollo con arroz with zucchini, fresh one-pot spaghetti, beef stroganoff, steak bites, and homemade queso. What mattered was not building a giant recipe database. It was creating enough context for the system to understand how we like to eat and use that knowledge when planning the next week.

Five meals aren’t five separate shopping lists.

This was one of the changes that made the whole workflow much more useful. Traditional meal planning tends to treat every recipe independently. You choose five meals, build five ingredient lists, combine them, and head to the store.

My workflow looks across the entire week instead.

If three recipes use onions, it calculates how many onions I actually need across all three. If two meals use zucchini, it accounts for the total quantity. If I already have garlic, rice, olive oil, chicken broth, or fresh herbs at home, those amounts are deducted before anything makes it onto the shopping list.

The inventory piece also helps me use food before it goes bad. If I already have fresh ingredients that need to be used soon, the workflow can take those into account when choosing the week’s meals. And if a recipe calls for only a small amount of something that is usually sold in a much larger package, the system can look for another sensible way to use the remaining ingredient later in the week.

I don’t want five meals that all taste the same simply because they share ingredients. The goal is to find useful overlap where it makes sense, while still giving us variety. In practice, that means buying closer to what we will actually use, using more of what is already in the kitchen, and throwing much less food away.

Then I tackled the part I hated most: shopping.

Planning the meals helped, but shopping was still full of unnecessary decisions.

A traditional shopping list might tell me I need tomato paste. That still leaves me standing in front of a shelf with twenty kinds of tomato paste, deciding which size, brand, or version to buy. Multiply that by an entire grocery list and I was still spending a surprising amount of mental energy on decisions I did not really care about.

So I connected the workflow to the online inventory at my local supermarket. Once the meals and ingredient quantities are finalized, the system can check what is actually in stock and help narrow the shopping list down to specific products and package sizes.

I can also change the purchasing criteria depending on what matters that week. Sometimes I want the highest-quality ingredients. For an ingredient I rarely use, I may want the smallest available package so I am not throwing half of it away later. If I am trying to keep the grocery bill down, I can make price the priority instead.

The underlying workflow stays the same. I only change the instruction.

That also means I can find out before I leave home that something is unavailable and adjust the plan or select a substitute in advance. By the time I am actually shopping, I am no longer evaluating every option on every shelf. Most of the low-value decisions have already been handled, while the choices I actually care about are still mine.