As you might have learned from reading my blogpost series “The Dainosaur”, I use AI extensively in my day to day work as solution architect. But recently I found a very different way to put it to work. One with much higher stakes than writing code.
The reason I started looking for a new way to use AI to my advantage, was a health issue. A couple of weeks ago I was diagnosed with type 2 diabetes. Obviously this was quite a shock and I realized I had to make some serious changes in my life to stay healthy.
I’m not keen on going to the gym, but I have started to exercise a bit more. Taking the stairs more often or walking to the supermarket for groceries. But that’s not the only change I had to make. I also had to make drastic changes in my diet.
I cut out unhealthy snacks, switched to coffee without sugar, and swapped Cola Zero for water. But I also wanted to start eating healthier diners. And although I’m not a picky eater, there are some ingredients I do not like. Although a bit strange for a Dutch guy, cheese is one of the things I do not like to eat (or smell for that matter). And many recipes include cheese. Also there are some vegetables and herbs I do not like. Anyway, I wanted to eat meals that are healthy for a diabetic but taste well. And then I thought: “that’s a perfect job for AI”. So what I did was, I fired up the Claude CLI and started working on generating some recipes.
My approach
The first step was to ask Claude to generate a list of ingredients that can be purchased in a standard Dutch supermarket and store it in a file (ingredients.md). The result was a vast list of ingredients, categorized into groups (meat, fish, vegetables, fruits, herbs, spices, etcetera). After that, I removed all the ingredients I do not like from the list.
The second step was writing instructions in a CLAUDE.md file. I described what my goals are and which constraints I wanted Claude to adhere to: only use ingredients on the ingredients list, generate recipes that are healthy for a person with type 2 diabetes, generate recipes that can be cooked in a kitchen with the equipment in my kitchen, etcetera. After that I asked Claude to generate recipes. And the results were great.
The recipes
Claude generated 30 recipes with lots of variation and all suitable in a diet for a diabetic. It also created an overview with a table containing all the recipes, with for each recipe the main protein, amount of carbohydrates, preparation time and some comments. Below you see the table:
| Recipe | Main protein | Carbohydrates | Preparation time | Notes |
|---|---|---|---|---|
| Pan-fried chicken breast with roasted vegetables | Chicken | Low | 35 min | Oven + pan |
| Pangasius with garlic butter and green beans | Fish | None | 20 min | Quick |
| Chicken thigh curry with white rice | Chicken | Moderate (white rice, small portion) | 30 min | Filling |
| Chickpea soup with smoked sausage | Sausage / legume | Low | 30 min | Savory, budget-friendly |
| Shrimp stir-fry with pak choi and rice noodles | Shrimp | Low–moderate | 15 min | Quick, Asian |
| Ground lamb with white beans and tomato | Lamb | Low | 30 min | Mediterranean |
| Omelet with mushrooms, ham and chives | Egg | None | 10 min | Quick, budget-friendly |
| Tuna salad with arugula, egg and green beans | Fish / egg | None | 15 min | No hot preparation |
| Beef stew with mushrooms | Beef | None | 2 hours | Slow-cooked |
| Pan-fried sea bass with red cabbage salad | Fish | None | 20 min | Elegant |
| Chicken soup with vegetables | Chicken | Low (little noodles) | 30 min | Classic |
| Ground meat with cauliflower mash | Ground meat | Very low | 25 min | Comfort food |
| Trout with sesame crust and pak choi | Fish | None | 15 min | Quick, Asian |
| Brown bean soup with bacon and leek | Bacon / legume | Low | 30 min | Dutch, budget-friendly |
| Mackerel with waxy potato and green peas | Fish | Moderate (potato limited) | 25 min | Classic, budget-friendly |
| Chicken with bell pepper, leek and green peas | Chicken | None / low | 20 min | One-pan meal |
| Endive mash with bacon and egg | Egg / bacon | Moderate (potato limited) | 25 min | Dutch classic |
| Broccoli stir-fry with beef and soy sauce | Beef | None / low | 15 min | Quick, Asian |
| Haddock with mustard sauce and endive | Fish | None | 15 min | Quick, refined |
| Lamb ragout with chickpeas and cumin | Lamb | Low | 1.5 hours | North African, hearty |
| Entrecôte with rosemary butter and broccoli | Beef | None | 20 min | Premium, quick |
| Turkey breast with tarragon sauce and green beans | Turkey | None | 25 min | French, refined |
| Oven-baked halibut with dill-yogurt sauce and bimi | Fish | None | 25 min | Quick, elegant |
| Pork chop with apple mustard and braised white cabbage | Pork | None | 30 min | Farmhouse dish, comfort food |
| Black bean soup with spiced chicken | Chicken / legume | Low | 30 min | Latin American |
| Tilapia with chermoula and roasted bell pepper | Fish | None | 30 min | North African |
| Shrimp scrambled eggs with bean sprouts and spring onion | Egg / shrimp | None | 15 min | Quick, Asian |
| Steak with red onion sauce and mushrooms | Beef | None | 25 min | French bistro |
| Kidney bean soup with smoked sausage and bell pepper | Sausage / legume | Low | 25 min | Savory, filling |
| Air fryer spareribs with white cabbage salad | Pork | None | 35 min | Air fryer, comfort food |
Each recipe contains all the information I need: ingredients, instructions on how to cook the recipe (including the time it takes) and an explanation why the recipe is suitable for a diabetic. This is awesome and very convenient. I have printed each recipe I have cooked, put it in a plastic sleeve and put it in the recipes binder that we already had in our kitchen.
To give you an idea of what types of recipes Claude has generated, here are some examples:
Pangasius with Garlic Butter and Green Beans
Healthy | Savory | Diabetes-friendly | Quick (20 min)
Pangasius is a mild, tender white fish full of protein and virtually free of carbohydrates. Green beans are low-GI vegetables that keep blood sugar stable. The garlic butter makes this dish flavorful and satisfying. In the air fryer, the fish is ready in 10 minutes with a lightly crisp exterior.
Quantities (for 1 person)
Multiply all quantities by the number of people.
| Ingredient | Quantity |
|---|---|
| Pangasius fillet | 180 g |
| Green beans (fresh or frozen) | 175 g |
| Garlic | 2 cloves |
| Butter | 15 g |
| Olive oil | 1 tbsp |
| Parsley (fresh or dried) | 1 tbsp, finely chopped |
| Apple cider vinegar | ½ tsp (optional, as acid) |
| Black pepper | to taste |
| Salt | to taste |
Cooking instructions
Preparation (5 min)
- Take the pangasius out of the fridge 15 minutes in advance.
- Pat the fish dry with kitchen paper. Season with salt and pepper.
- Finely chop the garlic.
Green beans (10 min)
- Bring a pan of salted water to the boil. Cook the green beans for 7–8 minutes until tender-crisp.
- Drain and keep warm.
Pan-frying the pangasius — skillet (8 min)
- Heat the olive oil in a skillet over medium-high heat.
- Place the pangasius in the pan. Fry for 2–3 minutes, turning carefully (the fillet is delicate).
- Add the butter and garlic. Fry for another 2–3 minutes, spooning the butter over the fish regularly.
- Sprinkle with parsley (and apple cider vinegar if desired).
Pangasius — air fryer variant (12 min)
6a. Coat the pangasius with olive oil, salt and pepper. Preheat the air fryer to 185 °C (365 °F). 7a. Air-fry the pangasius for 8–10 minutes — pangasius is thinner than cod and cooks faster. 8a. Meanwhile, melt the butter with garlic in a small pan over low heat (2 minutes). Stir in the parsley. 9a. Drizzle the fish with the garlic butter on the plate.
Serving
- Spoon the green beans onto the plate, place the pangasius beside them and spoon the garlic butter over the top.
Diabetes tips
- Pangasius has virtually no carbohydrates — ideal for type 2 diabetes.
- Green beans have a low glycemic index and are rich in fiber.
- Avoid store-bought fish butter (often contains sugar); make it yourself as described above.
- Pangasius is a thin fillet — watch the cooking time closely, as overcooking makes the fish dry and flaky.
Ground Meat with Cauliflower Mash
Healthy | Savory | Diabetes-friendly | Comfort food
Cauliflower mash is a smart replacement for mashed potatoes: just as creamy, but with a fraction of the carbohydrates. The ground meat adds a hearty, savory flavor. This is a complete Dutch comfort dish that fits a diabetes-friendly diet.
Quantities (for 1 person)
Multiply all quantities by the number of people.
| Ingredient | Quantity |
|---|---|
| Ground meat (beef or mixed) | 150 g |
| Cauliflower | 300 g (in florets) |
| Onion (yellow) | ½ piece |
| Garlic | 1 clove |
| Butter | 10 g |
| Crème fraîche | 1½ tbsp |
| Nutmeg | a pinch |
| Tomato paste | 1 tbsp |
| Worcestershire sauce | 1 tsp |
| Paprika powder (smoked) | ½ tsp |
| Olive oil | 1 tbsp |
| Black pepper | to taste |
| Salt | to taste |
Cooking instructions
Mashing the cauliflower (15 min)
- Cook the cauliflower florets in salted water for 12–15 minutes until very soft. Drain thoroughly — leftover water makes the mash watery.
- Mash finely with butter and crème fraîche. Season with nutmeg, salt and pepper. Keep warm.
Ground meat (12 min)
- Heat the olive oil in a skillet over medium-high heat. Sauté the onion for 3 minutes. Add the garlic, cook for 1 minute.
- Add the ground meat. Fry until crumbly and browned (5 minutes) — break up clumps with a spatula.
- Stir in the tomato paste, Worcestershire sauce and paprika powder. Cook for 2 minutes more.
- Add 3 tbsp water if the mixture becomes too dry. Stir and let reduce for 2 minutes.
- Taste and adjust seasoning.
Serving
- Spoon the cauliflower mash onto the plate, make a well and spoon the ground meat into it.
Diabetes tips
- Cauliflower has a GI of about 15 versus about 78 for potatoes — a huge difference.
- This dish contains hardly any carbohydrates and is therefore excellent for type 2 diabetes.
- Crème fraîche in a small amount adds creaminess without significant carbohydrates.
Spareribs from the Air Fryer with White Cabbage Slaw
Low-carb | High-protein | Savory | Comfort food | Diabetes-friendly
Crispy, deeply spiced spareribs from the air fryer in under 30 minutes — without the sweet sauces of a takeaway restaurant. A dry rub of smoked paprika, cumin and garlic delivers all the flavor. Fresh white cabbage slaw cuts through the richness and brings balance.
Quantities (for 1 person)
Multiply all quantities by the number of people.
| Ingredient | Quantity |
|---|---|
| Spareribs | 300–350 g |
| Smoked paprika | 2 tsp |
| Cumin (ground) | 1 tsp |
| Garlic powder | 1 tsp |
| Chili powder | ½ tsp |
| Black pepper | ½ tsp |
| Salt | 1 tsp |
| Cabbage slaw | |
| White cabbage | 200 g |
| Mayonnaise | 1 tbsp |
| Greek yogurt | 1 tbsp |
| Coarse mustard | 1 tsp |
| Apple cider vinegar | 1 tbsp |
| Salt, black pepper | to taste |
Cooking instructions
Preparation (10 min)
- Mix the smoked paprika, cumin, garlic powder, chili powder, pepper and salt into a dry rub.
- Pat the spareribs dry and rub them generously with the mixture. Let rest for 5–10 minutes.
- Shave or slice the white cabbage into thin strips. Mix the mayonnaise, Greek yogurt, mustard and apple cider vinegar. Toss through the cabbage. Season to taste. Refrigerate.
Air fryer (25 min)
- Preheat the air fryer to 180°C (355°F).
- Place the spareribs in the basket (curve them to fit if needed). Cook for 20 minutes at 180°C (355°F) — turn halfway through.
- Raise the temperature to 200°C (390°F) and cook for another 5 minutes for a crispy, deeply browned exterior.
Serving
- Plate the spareribs. Serve the cabbage slaw alongside, cold from the fridge.
Diabetes tips
- Spareribs without a sweet sauce are low-carb — the dry rub delivers all the flavor without sugar or honey.
- Greek yogurt in the slaw replaces part of the mayonnaise for more protein and less fat.
- White cabbage is particularly high in fiber and contains almost no carbohydrates — a perfect low-carb side that’s also gut-friendly.
The results
I have cooked several of the recipes during the past 4 weeks (some multiple times) and I have really enjoyed them. They were really tasteful and the instructions were spot on. My son has decided to support me and eat the healthy meals with me. And so far, he has also thoroughly enjoyed the meals. My wife already has a different diet for medical reasons (eating is somewhat complicated in our household). So I always cook for my son and me, which is easy because all quantities in the recipes can easily be multiplied by the number of people (one of the constraints I gave Claude by the way).
This is all great, but the impact on my health is even more impressive. I have lost almost 4 KG over the past 4 weeks. I also regularly measure my glucose level and this has gradually dropped several points: from around 15 to around 6 mmol/L. In my opinion this proves that the recipes work for me.
I hope to keep my diabetes under control without medication by keeping up my new lifestyle. And because I really enjoy eating the meals based on these new recipes, I’m convinced I will be able to do so. And if I would need more variations in the future, I can always ask chef Claude to generate some more recipes.
Also get your own personal AI chef
Since I have used Claude to generate the recipes, I have shared the instructions with several people. They have updated the instructions for their situation. One person used it for generating recipes that take several food allergies into account. Another person created recipes that help with the preparations for running a marathon. They all really enjoy the meals they have created based on these recipes.
In order to make sharing this way of generating recipes easier, I have created a simple generic recipe-generator skill. You can download it from my GitHub repo: https://github.com/EdwinVW/chef-claude. Follow the instructions in the README file to get started. The repo and skill is in Dutch. But obviously you can use AI to create a version in another language. This is open source, so feel free to create a PR if you have any improvements.
I really hope many people will benefit from the skill and can use it for generating healthy recipes for their specific situation. If you use the skill, I would appreciate it if you share your experiences by leaving a comment on this issue in the GitHub repo or on the LinkedIn post in which I announced this blogpost. Thanks in advance!