> ## Content Index > Fetch the complete content index at: https://www.darrenma.eu/llms.txt > Use this file to discover other available public pages before exploring further. # AI did not teach me to cook. It removed the friction - URL: https://www.darrenma.eu/ai-did-not-teach-me-to-cook-it-removed-the-friction/ - Published: 2026-08-07T07:29:52.000Z - Updated: 2026-08-07T07:29:52.000Z - Description: How I use AI to adapt recipes, build weekly shopping lists, and remove enough planning friction to cook for my family three times a week. - Author: Darren Ma - Tags: Cooking, AI, Habits *How I use AI to adapt recipes, plan the shopping, and make cooking three times a week sustainable.* ## Too much inspiration, not enough dinner When I started cooking more seriously, I did what I usually do: I collected too much information. I had cookbooks and would occasionally find something I wanted to make. Online there were even more options: YouTube videos, recipe websites, PDFs, and snippets saved from chats. The willingness was there, but it was often followed by the same sequence: > I don't have that. I don't have this. What even is that, and where am I supposed to buy bok choi? Then I would close the book and cook nothing. Recipes vary a lot. Some included ingredients I did not like or could not easily identify in Italy. Others were designed for six people when I needed three or four. None of these were big problems. They were still enough to make me close the book and give up for the night. For me, the planning was hard. The cooking was easy. AI helped remove enough of the planning friction that I could get to the part I actually wanted to do, cooking! I do sometimes use AI to generate recipes too. But the useful part for building the habit was taking recipes I already wanted to cook, making them fit my kitchen, and making sure I had everything ready. ## What I ask AI to do I keep my recipes as Markdown notes in an Obsidian vault. Because I work with the vault in VS Code, assistants such as Codex or Claude can read the source recipe and the examples already stored beside it. I might start with a PDF, a recipe URL, or something copied from a video. I then ask the assistant to: - turn it into a clean recipe note - adjust the number of portions - convert everything to metric - remove sections I do not need - remove ingredients I dislike - suggest locally available alternatives - keep the instructions short and easy to reference while cooking This is not complicated or clever. It is mostly a conversation about making somebody else's recipe fit my kitchen. ## A real example One recipe that caught my eye was HelloFresh's [Sriracha Pork Chops with Bánh Mì Lunch](https://www.hellofresh.com/recipes/sriracha-pork-chop-for-dinner-5b106d44ae08b539993371f2?ref=darrenma.eu). The original was designed as a cook-once, eat-twice recipe. You make pork chops for dinner, then use the leftovers in bánh mì sandwiches the next day. I only wanted the dinner. I asked the assistant to remove the lunch section, convert the imperial measurements to metric, and remove the cilantro because I do not like it. I also needed help making sense of the pork cut. A “pork chop” is not always labelled in the same obvious way when I am shopping in Italy, so I could check whether the cut available to me would work and whether the cooking method needed to change. The adapted recipe specified `680 g` of pork. I bought the closest package I could find, about `700 g`, cooked all of it, and the amount worked well. The result was no longer a PDF I had to reinterpret each time. It was a familiar recipe note in English, using metric measurements, with only the meal I intended to cook. ## The shopping list was the real win Adapting individual recipes helped, especially while I was building the library. The bigger improvement was what happened after I had several recipes ready. I could select a few meals that might suit the family that week and ask AI to create one shopping list for all of them. The list is grouped into sections such as produce, meat and seafood, dairy, grains, and pantry ingredients. Shared ingredients are combined instead of appearing several times. I can check off what is already at home, then shop for the rest. When the week arrives and I decide to cook one of those meals, the ingredients are there. I open the recipe on my phone, glance through the method, and start. I wrote separately about [how I sync that Obsidian vault between Bazzite and iPhone](https://www.darrenma.eu/i-only-wanted-to-sync-a-few-recipes/). That is the friction that was worth removing. Inspiration no longer collapses at the first missing ingredient. ## I use AI less now I have been cooking roughly three times a week for about six months. At the beginning, while I was building the recipe library, I used AI for this process practically every day. I needed more help converting recipes, understanding ingredients, and getting everything into a format I liked. Now I mostly use it for the weekly shopping list. Some recipes are familiar enough that I already know what to do, although I still usually glance at the saved note. The heavy assistance became lighter as my own experience grew. I think that is a good sign. I needed a lot of help while I built the library and became more comfortable cooking. Now that the habit is established, I do more of it myself. ## Getting consistent output AI has not made any meaningful mistakes in my ingredients or shopping lists so far. The problem I did have was inconsistency. One recipe would arrive as plain text. Another would include emojis. Sometimes the shopping list used checkboxes, sometimes it did not. The information was correct, but the growing library looked like it had been assembled by several different people who had never spoken to one another. I fixed that by creating examples in the format I wanted. I have one recipe that acts as my recipe template and one shopping list that acts as the shopping-list template. Instead of explaining every formatting decision again, I point the assistant at the relevant example and ask it to follow that structure. My permanent rules are simple: - always use English - always use metric measurements - follow the referenced template - keep everything easy to read while cooking or shopping - use checkboxes in shopping lists - do not use emojis I do not actually copy and paste one magic prompt. I normally ask in natural language because every recipe needs something slightly different. The consistency comes from showing the assistant what good output looks like. ## The prompts, if I had to write them down These are cleaned-up versions of how I ask naturally, not prompts I have been copying word for word. You do not need a complicated template. Mine is just quick facts, metric ingredients, a few useful notes, and numbered steps. The shopping list links the selected recipes, combines quantities, groups items by section, and uses checkboxes. ### Adapt a recipe ```text Read the attached recipe, PDF, or URL and create a recipe note using [recipe template] as the exact structure and style. Adapt it for [number] people. Use English and metric measurements. Remove [unwanted section or ingredient]. Replace ingredients that are difficult to buy in Italy with obvious local alternatives, but tell me when a substitution could materially change the dish. Keep the method easy to scan while cooking. Use no emojis. Save the finished recipe at [vault path]. ``` ### Build the weekly shopping list ```text Create one shopping list for these recipes: - [recipe] - [recipe] - [recipe] Use [shopping-list template] as the exact format. Deduplicate shared ingredients, calculate the combined quantities, group everything by shopping section, and keep likely pantry basics separate. Use English, metric measurements, checkboxes, and no emojis. ``` I have used the same idea for larger meals too. If several dishes need to be ready at a specific time, I can provide the recipes and ask for a low-stress preparation timeline that works backwards from serving time. It accounts for prep, cooking, resting, shared appliances, and anything that can be done earlier. What matters is giving it the recipes, showing it the format I use, and being clear about what I need back. ## What changed I enjoy cooking for my family. It takes some pressure off everyone else and gives me a practical way to contribute at home. I also enjoy improving at it. Cooking is a useful skill, and one I want to keep developing. AI did not create that motivation and it obviously did not do the cooking for me. It helped with the part that used to make me give up before I started. Six months later, I am still cooking about three times a week and I need less help than I did at the beginning. That is a pretty good result for me. ## References - [HelloFresh: Sriracha Pork Chops with Bánh Mì Lunch](https://www.hellofresh.com/recipes/sriracha-pork-chop-for-dinner-5b106d44ae08b539993371f2?ref=darrenma.eu)