The night AI helped me make dinner without burning down the house

I had two frozen chicken breasts and a small collection of vegetables in the refrigerator that needed to be used: half a Japanese eggplant, half a zucchini, and a squash.

I am not the cook in our house. Jason is the person who can look at ingredients and see dinner. I look at the same ingredients and see several independent opportunities to ruin dinner at once. My culinary standard for the evening was therefore modest: make something edible, use the vegetables, and leave the kitchen attached to the rest of the house.

So I told AI what I had, including the fact that the chicken was frozen, and asked what I could make. It gave me several options. Honey mustard sounded difficult to hate and safely within my range of competence, so I chose that and told it which two honey-mustard bottles were actually in the refrigerator.

Then the AI confidently told me to cook the chicken under pressure for 12 minutes.

That did not sound right. We have a cooking chart on the refrigerator, and I had watched Jason cook the same kind of chicken before. Both pieces of household evidence said seven minutes, not 12. So I stopped. I did not assume the machine knew better because it had produced a nicely formatted answer. I told it what the chart said and what Jason normally did.

The plan changed: seven minutes at high pressure, then a five-minute natural release. Also, two chicken breasts cook at the same time. I mention that because I was new enough to this process that it was a real question.

At that point I stopped treating AI like the authority and started treating it like help. The first answer was wrong. If I had reduced my role to following directions, I could have marched straight into dry-chicken country. I knew something about my own kitchen that the machine did not, and the plan got better when I said so.

The chicken cooked. I took it out, saved some of the liquid, and moved on to the vegetables. Then I noticed the instructions had skipped a small but operationally significant detail: cutting them. I asked, and yes—apparently vegetables still require cutting, even in the age of artificial intelligence.

I cut them, sautéed them, mixed the sauce, and kept moving. There was no cinematic transformation in which I discovered my lifelong culinary calling. There was just the very satisfying moment when all the separate uncertain steps turned into an actual plate of food.

My final review was: “Pretty good actually.” I would use a little more sauce next time.

It was a small result, and I like it for exactly that reason. Two frozen chicken breasts and some vegetables became dinner. The AI was wrong once and incomplete once. I caught both, kept cooking, and ended up with food I would willingly make again.