Most men my age try to prove they’re still young by running a marathon, buying a sports car, or growing an aggressively confident beard.
I, on the other hand, decided to do something far more dangerous:
I let artificial intelligence cook every single meal for me for an entire month.
Breakfast, lunch, dinner—every day, for 30 days.
No improvising, no cheating, no “this looks weird, so I’ll adjust the seasoning.”
Just blind trust in ChatGPT, Claude, and Google Gemini.
Why?
Because at 55 I wanted to know whether AI could actually improve my eating habits, save me time, reduce my stress, and maybe even make me feel healthier.
Or whether it would burn my kitchen down while trying to turn yogurt into soup.
Here is everything that happened—every fail, every triumph, every meal that should be illegal.
DAY 1: A TOTALLY NORMAL BREAKFAST (HOPE RETURNS)
I eased into the experiment with a gentle prompt:
“Give me a simple, healthy breakfast suitable for a 55-year-old man.”
The AI suggested:
- Greek yogurt
- Berries
- Honey
- Walnuts
A safe, uncontroversial start.
I felt optimistic.
“AI cooking is going to be easy,” I told myself.
This was the last moment of peace I would experience for a while.
DAY 2: THE FIRST RED FLAG (A DISASTER SOUP)
For lunch, Claude gave me a recipe called “Rustic Chickpea Stew with Depth.”
Depth, as it turns out, meant:
- Too much paprika
- A terrifying amount of cumin
- A suggestion to “add a splash of vinegar for brightness,” which tasted like battery fluid
I ate it—because rules are rules—and immediately questioned my life decisions.
DAY 3–5: LEARNING THAT AI CAN’T COOK… YET CAN CREATE MAGIC
These days were chaotic, but they taught me the first major lesson of the experiment:
AI has no intuitive understanding of food—but it does understand patterns.
Once I gave it specific instructions:
“Only 6 ingredients, must be under 25 minutes, must not set my mouth on fire,”
the recipes improved dramatically.
One night it created a lemon-garlic salmon with sautéed spinach that tasted like something from a restaurant where waiters wear aprons that cost more than my shoes.
Another day, Gemini gave me a chicken bowl with quinoa, cherry tomatoes, and tahini dressing that instantly entered my weekly rotation.
I started thinking:
Is AI secretly the best personal chef I’ve ever had?
Of course, the universe punished me immediately.
DAY 6: THE AI-PUDDING INCIDENT
For dessert, ChatGPT offered:
“One-pot creamy vanilla pudding.”
What it forgot to mention is that “one-pot” does not mean you can cook everything at once like some medieval witch.
Its instructions:
- Heat milk
- Add cornstarch
- Add sugar
- Add vanilla
- Stir continuously
- Then add uncooked pasta “for a comforting texture”
PASTA.
In pudding.
The result was the most confusing food I’ve ever eaten.
Part dessert, part school cafeteria trauma.
DAY 7: THE FIRST BREAKTHROUGH – PERSONALIZED MEAL PLANS
After some strategic prompting, I discovered AI’s greatest strength:
nutrition planning.
I gave it my data:
- 55 years old
- moderately active
- mild acid reflux
- no spicy dinners
- preference: Mediterranean-style meals
- budget: $20/day
- skill level: “I know how to use a knife but I wouldn’t call myself an artist”
Suddenly, everything clicked.
AI created perfect weekly meal plans:
balanced, realistic, delicious.
And for the first time since the experiment began, my stomach felt safe.
DAY 8–12: THE GOLDEN AGE OF AI COOKING
These days were shockingly smooth.
Highlights included:
1. Honey-Mustard Chicken with Roasted Carrots (Gemini)
This was scandalously good.
2. Spinach-Feta Omelette (ChatGPT)
Simple. Elegant. Foolproof.
3. Mild Turkey Chili (Claude)
Healthy, comforting, and did not burn my esophagus.
4. Greek “Power Lunch Bowl” (ChatGPT)
Protein-rich, low-sodium, perfect for a 55-year-old.
5. Mediterranean Lentil Soup (Gemini)
So balanced I considered framing the recipe.
I was ready to declare AI the future of cooking.
But then… the experiment tried to murder my steak.
DAY 13: THE SOUS-VIDE TRAGEDY (OR: HOW I MADE BEEF JERKY)
This was the moment my X/Twitter post was born.
I asked Claude for a “perfect medium-rare steak using sous-vide.”
Claude recommended:
- 4 hours
- 140°F (60°C)
- In a pot of water
- Kept warm in the oven
Here’s the problem:
The oven dries water.
Which dries meat.
Which turns steak into a prehistoric artifact.
After four hours, I removed the bag.
Steam escaped.
The steak was not steak.
It was the most expensive jerky ever produced in my kitchen.
DAY 14–16: THE “AI DOESN’T UNDERSTAND PHYSICS” ERA
More fails followed:
Fail #1: Pancakes with 3 bananas and no flour
Result: A sweet omelette with abandonment issues.
Fail #2: “Healthy” tuna salad with hot mustard AND vinegar
Result: Chemical warfare.
Fail #3: A smoothie containing spinach, mint, yogurt, and cumin
No comment.
This phase taught me another truth:
AI is brilliant at ingredients, terrible at textures, and completely incompetent at heat.
DAY 17–22: MASTERING THE ART OF PROMPTING
By this point, the experiment was no longer about the recipes—it was about communication.
AI cooking only works when you give it boundaries, like:
- “Do NOT use more than one acidic ingredient.”
- “Avoid unusual texture combinations.”
- “Use moderate salt.”
- “No chili after 5 p.m.” (my acid reflux thanks me)
- “Limit prep to 20 minutes.”
- “Pretend the cook is 55 and gets tired of chopping.”
Once I set these rules, everything improved instantly.
AI created the best dish of the month:
Baked Halloumi with Zucchini, Bell Peppers, and Lemon-Garlic Yogurt
High protein.
Low effort.
No stress.
Delicious.
This meal alone justified the entire experiment.
DAY 23–26: AI AS A HEALTH COACH
One day, on a whim, I typed:
“Optimize today’s meals for inflammation reduction.”
And AI delivered:
- Ginger-lemon tea
- Salmon with turmeric
- Blueberries
- Spinach salad
- Almonds
I felt lighter.
Clearer.
Less bloated.
Then I tested:
“Optimize for heart health.” → olive oil, walnuts, whole grains.
“Optimize for sleep quality.” → chamomile, banana, warm oats.
“Optimize for energy.” → eggs, avocado, citrus.
AI wasn’t just cooking—it was coaching.
I realized something important:
AI won’t replace chefs, but it may replace dietitians.
DAY 27–30: THE FINAL STRETCH & THE BIGGEST SURPRISE
These last days were my favorites.
Why? Because at this point AI had learned me.
My preferences.
My limitations.
My cooking laziness.
The meals were consistently:
- balanced
- tasty
- realistic
- affordable
- gentle on my 55-year-old digestive system
The final four dinners were shockingly good:
1. Crispy Tofu Stir-Fry (ChatGPT)
I didn’t even know I liked tofu.
2. Mediterranean Chicken with Olives (Gemini)
Elegant and easy.
3. Light Shrimp Pasta (Claude)
Restaurant quality.
4. Greek Yogurt Parfait with Almonds (ChatGPT)
Perfect ending to the experiment.
WHAT AI CAN DO WELL IN THE KITCHEN
After 30 days, here’s my verdict:
Strengths
✔ Meal planning
✔ Balanced nutrition
✔ Healthy recipes
✔ Grocery lists
✔ Adapting to age and health needs
✔ Creativity within boundaries
✔ Reducing food waste
✔ Making me try new things
Weaknesses
✖ Heat management
✖ Sous-vide temperatures
✖ Texture combinations
✖ Salt control
✖ Excessive confidence
✖ Dangerous curiosity (“Add mustard to your smoothie!”)
AI is a brilliant assistant—
but a terrible autonomous chef.
THE 5 AI RECIPES THAT ACTUALLY WORKED (EVERY TIME)
1. Lemon-Garlic Salmon with Sautéed Spinach
Simple, healthy, delicious.
2. Baked Halloumi with Roasted Vegetables
A monthly favorite.
3. Mild Turkey Chili
Perfect for digestion.
4. Chicken-Quinoa Power Bowl
Balanced macros.
5. Greek Yogurt Breakfast Parfait
Shockingly satisfying.
FINAL CONCLUSION: WOULD I DO IT AGAIN?
Absolutely—
with some rules.
AI won’t replace human intuition, but it will:
- save you time
- improve your nutrition
- reduce stress
- diversify your meals
- keep you accountable
And for a 55-year-old like me, that matters more than culinary perfection.
So yes—AI can cook for you.
Just… don’t let it touch your steak.