Cooking with an Algorithm: I Let an AI Plan My Weekly Meal Prep
My wife has a habit that’s equal parts adorable and terrifying: she organizes things with military precision.
Our spice cabinet looks like it belongs in a culinary museum. Every spice is labeled, categorized, and stored in matching glass containers. Turmeric. Cumin. Coriander. Paprika. Garam masala. Star anise. Cardamom (both green and black, because apparently they’re different). She’s even labeled the mixes—her homemade blends for fish curry, chicken korma, and beef masala.
One day, I took a picture of this magnificent spice collection and did something ridiculous: I asked an AI to suggest recipes based on what we had.
I expected generic results. “Make curry.” “Add turmeric to chicken.”
Instead, the AI gave me a week’s worth of meal ideas that were absurdly good. Dishes I’d never heard of. Flavor combinations that shouldn’t work but did. Nutritional profiles broken down to the gram.
This is the story of what happened when we let an algorithm plan our meals—and whether AI actually understands flavor, or just knows how to fake it really well.
The Spice Cabinet That Started It All
Let me set the scene.
My wife spent an entire weekend organizing our spices. She bought matching jars, printed labels, and sorted everything by cuisine type and frequency of use. The result was a work of art.
I stared at it and thought: This is too beautiful to use. What if I mess up the organization?
She stared at it and thought: Now we can cook anything.
Then I had an idea. I pulled out my phone, took a photo of the labeled spices, and uploaded it to ChatGPT.
“Here’s what I have,” I typed. “Give me a week of meal ideas.”
I expected it to suggest basic recipes—maybe some chicken curry, a stir-fry, a pasta dish. Instead, it gave me this:
AI’s Weekly Meal Plan (Based on Our Spices)
Monday: Bengali Fish Curry with Mustard and Turmeric
- Pan-fried fish in a mustard-turmeric sauce with green chilies and tomatoes
- Serve with steamed rice
- Nutritional Profile: ~450 cal, 35g protein, 12g fat, 40g carbs
Tuesday: Moroccan Chickpea Stew with Cumin and Paprika
- Slow-cooked chickpeas with tomatoes, paprika, cumin, coriander
- Add spinach in the last 10 minutes
- Nutritional Profile: ~380 cal, 15g protein, 8g fat, 60g carbs
Wednesday: Korean-Inspired Chicken with Gochugaru and Star Anise
- Marinated chicken thighs with soy sauce, garlic, star anise, and gochugaru (Korean chili flakes)
- Roast until crispy
- Nutritional Profile: ~520 cal, 40g protein, 18g fat, 35g carbs
Thursday: Indian Beef Masala with Homemade Spice Mix
- Slow-cooked beef in onion-tomato gravy with your labeled “beef masala” mix
- Finish with fresh coriander and a squeeze of lime
- Nutritional Profile: ~580 cal, 45g protein, 22g fat, 38g carbs
Friday: Thai Basil Shrimp with Cardamom-Infused Rice
- Stir-fried shrimp with Thai basil, garlic, and a hint of cardamom
- Serve over cardamom-scented jasmine rice
- Nutritional Profile: ~410 cal, 30g protein, 10g fat, 48g carbs
Saturday: Middle Eastern Lentil Soup with Cumin and Coriander
- Red lentils cooked with cumin, coriander, garlic, and lemon
- Top with crispy fried onions
- Nutritional Profile: ~320 cal, 18g protein, 6g fat, 52g carbs
Sunday: Japanese Curry with Garam Masala Twist
- Slow-cooked vegetables and chicken in a Japanese curry base
- Add a pinch of garam masala for depth
- Nutritional Profile: ~490 cal, 28g protein, 14g fat, 58g carbs
I stared at the screen.
“This… this is insane,” I said.
My wife leaned over my shoulder. “Korean chicken with star anise? That’s not a thing.”
“Yeah, but what if it is now?”
The First Test: Bengali Fish Curry That Tasted Like Home
We decided to try the AI’s Monday recipe: Bengali Fish Curry with Mustard and Turmeric.
This dish is sacred in Bengali cuisine. My wife grew up eating it. Her mother makes it with a specific ratio of mustard oil to turmeric, a careful balance of heat and tang.
If the AI got this wrong, the experiment was over.
What the AI Suggested:
- Marinate fish in turmeric and salt
- Make a paste of mustard seeds, green chilies, and turmeric
- Fry the fish lightly, set aside
- Cook the mustard paste in mustard oil with tomatoes and water
- Add the fish back in, simmer for 10 minutes
What We Did:
We followed the instructions exactly.
My wife was skeptical. “This is too simple. My mom’s recipe has more steps.”
But when we tasted it?
It was perfect.
Not identical to her mother’s version—but close enough that she paused mid-bite and said, “How does it know?”
The flavor balance was right. The mustard wasn’t too bitter. The turmeric didn’t overpower. The fish was tender.
“It tastes like home,” she said quietly.
And that’s when I realized: this AI might actually understand flavor.
The Science of AI and Flavor Profiles
Here’s the thing: AI doesn’t taste anything. It’s not sitting in a kitchen with a spoon, adjusting seasoning.
But it has access to something humans don’t: millions of recipes and their underlying patterns.
How AI “Understands” Flavor
-
Pattern Recognition Across Cuisines
- AI models are trained on vast recipe databases (think AllRecipes, Serious Eats, food blogs)
- They learn which spices are commonly paired: cumin + coriander in Indian food, star anise + soy sauce in Chinese cuisine
- They recognize ratios—not just ingredients, but proportions
-
Chemical Flavor Compounds
- Some AI models (like IBM’s Chef Watson) are trained on flavor compound databases
- They know that cumin contains aldehydes that pair well with the sulfur compounds in onions
- They can predict which spices will complement or clash based on molecular structures
-
Cultural Context
- AI doesn’t just suggest “chicken + turmeric”—it suggests Bengali fish curry because it recognizes the cultural context of mustard oil, turmeric, and fish
- It knows that star anise works in Korean cooking because it’s used in gochujang and doenjang stews
Source: Research from Nature Scientific Reports (2011) on food pairing theory shows that Western cuisines tend to pair ingredients with shared flavor compounds, while Asian cuisines often pair ingredients with contrasting compounds. AI models trained on these patterns can generate culturally authentic recipes.
The Week of AI Cooking: What Worked (and What Didn’t)
We committed to the full week. Here’s how it went:
Monday: Bengali Fish Curry ✅
Verdict: Near-perfect. My wife was impressed. Surprise Factor: The AI nailed the mustard-to-turmeric ratio.
Tuesday: Moroccan Chickpea Stew ✅
Verdict: Delicious, but we added more paprika than suggested. Surprise Factor: The AI recommended adding spinach at the end, which we wouldn’t have thought of. It worked.
Wednesday: Korean Chicken with Star Anise ⚠️
Verdict: Good, but weird. Surprise Factor: Star anise in Korean chicken is not traditional. But it added an unexpected depth that we liked. My wife said, “This is fusion food pretending to be authentic.”
Thursday: Indian Beef Masala ✅
Verdict: Better than my usual version. Surprise Factor: The AI told us to finish with lime juice. Game changer.
Friday: Thai Basil Shrimp with Cardamom Rice ✅
Verdict: Outstanding. The cardamom rice was fragrant and subtle. Surprise Factor: We were skeptical about cardamom in rice, but it elevated the dish.
Saturday: Middle Eastern Lentil Soup ✅
Verdict: Comforting and simple. Surprise Factor: The crispy fried onions on top were essential, and we almost skipped them.
Sunday: Japanese Curry with Garam Masala ⚠️
Verdict: Interesting, but not traditional. Surprise Factor: Garam masala added warmth, but purists would hate this. We loved it.
Does AI Understand Flavor, or Just Data?
Here’s the question that haunted me all week:
Does the AI actually understand flavor, or is it just really good at pattern matching?
The Case for “It Understands Flavor”
- It suggested dishes that worked—not just on paper, but in taste
- It balanced flavors: sweet, salty, sour, bitter, umami
- It adjusted for cooking techniques (e.g., frying fish before adding to curry, adding spinach at the end of stews)
- It provided nutritional breakdowns that were accurate
The Case for “It’s Just Data”
- It doesn’t taste anything—it’s predicting based on patterns
- When it suggested “Korean chicken with star anise,” it was blending cuisines in a way that’s statistically plausible but not culturally authentic
- It can’t adjust for personal preference the way a human chef can
- It doesn’t know if we’re out of an ingredient or if something went wrong
My Verdict:
AI understands flavor in the way a composer understands music without hearing it.
It knows the structure of flavor—which notes harmonize, which clash, which create tension and release. But it doesn’t experience the taste.
A human chef tastes and adjusts. AI predicts and hopes.
And honestly? That’s enough.
The Nutritional Breakdown: Does AI Care About Health?
One of the surprising things about the AI’s meal plan was the nutritional detail.
Every dish came with:
- Calorie count
- Protein, fat, carb breakdown
- Suggested portion sizes
Monday’s Fish Curry:
- Calories: 450
- Protein: 35g
- Fat: 12g
- Carbs: 40g
When we checked the actual nutrition (using MyFitnessPal), it was within 10% of the AI’s estimate.
This is where AI excels.
Humans are bad at estimating calories. We eyeball portions, forget about cooking oil, undercount carbs. AI has access to USDA nutrition databases and can calculate macros with precision.
Source: A 2018 study in Nutrients found that AI-based nutrition tracking apps are significantly more accurate than manual logging, with error rates under 15% compared to 30-50% for human estimates.
For someone tracking macros, managing diabetes, or just trying to eat balanced meals, this is huge.
The Weird Part: AI’s Creativity
The most unsettling thing about this experiment?
The AI suggested dishes I’d never heard of, and they tasted amazing.
Take the Thai Basil Shrimp with Cardamom-Infused Rice. That’s not a traditional Thai dish. Thai cuisine uses basil, garlic, fish sauce—but cardamom in rice?
We tried it anyway. The cardamom was subtle, floral, and paired perfectly with the savory shrimp.
My wife said, “This is what happens when an algorithm reads a thousand cookbooks and starts improvising.”
It’s not authentic. But it’s good.
And that raises a question: If AI can create new dishes that taste great, is it being creative?
What Is Culinary Creativity?
Traditional chefs would say creativity is about:
- Understanding ingredients on an intuitive level
- Improvising based on taste
- Drawing from personal experience and culture
AI does none of that. But it does:
- Combine flavors in statistically unexpected ways
- Suggest pairings that humans might not consider
- Generate novel dishes that are culinarily sound
Is that creativity? Or just sophisticated remixing?
I don’t know. But I ate the cardamom rice, and it was delicious.
The Verdict: Should You Let AI Plan Your Meals?
After a week of AI-generated cooking, here’s what I learned:
What AI Is Great At:
✅ Suggesting dishes based on available ingredients ✅ Providing accurate nutritional breakdowns ✅ Balancing flavors across a week (variety without repetition) ✅ Offering culturally diverse recipes ✅ Saving time on meal planning
What AI Is Bad At:
❌ Adjusting for personal taste (it doesn’t know I hate cilantro) ❌ Accounting for ingredient substitutions ❌ Understanding texture preferences (my wife likes crispy fish, I like it soft) ❌ Knowing when to break the rules (sometimes you need comfort food, not balanced macros)
My Recommendation:
Use AI as a sous chef, not the head chef.
Let it suggest ideas, plan macros, and inspire new dishes. But taste as you go. Adjust. Make it yours.
Because the best meals aren’t just about data—they’re about intuition, love, and the willingness to add extra garlic even when the recipe says not to.
The Spice Cabinet Lives On
My wife’s spice cabinet is still immaculate.
But now, when I look at it, I don’t just see turmeric and cumin and star anise.
I see possibilities.
I see an AI that can turn a photo of spices into a week of meals.
I see a future where cooking isn’t about memorizing recipes—it’s about collaboration between human intuition and algorithmic precision.
And honestly?
I’m okay with that.
As long as the food tastes good, and my wife keeps labeling the spices, I’ll keep asking the AI for ideas.
Because at the end of the day, whether it’s a human or an algorithm suggesting the recipe, what matters is this:
We cooked together. We ate together. And the meal was delicious.
And that’s all that really counts.