Plan a week of meals for a family of four under $300 with ChatGPT

Plan four family dinners from fridge and pantry photos and save about $150 with Gemini

Amanda Caswell, a mom of three writing for Tom's Guide, kept buying food the family already had. Caswell photographed the fridge, freezer and pantry, uploaded the photos to Gemini (using the Nano Banana 2 image model) and asked for meals made from what was there, plus the cheapest items to buy. Gemini spotted lean ground sirloin, frozen broccoli, chicken, black beans and spaghetti, and built a 4-day plan: beef and bean pasta, a cheesy chicken and broccoli bake, an everything frittata and burrito bowls. Caswell estimates about $150 in savings compared with a usual weekly spend.

Plan a week of family dinners for $3.06 a serving from a supermarket sale catalog with ChatGPT

Personal finance writer Nicole Pedersen-McKinnon wanted a seven-night dinner plan for a family of four using only ingredients on sale, that the kids would actually eat. Asking ChatGPT and Claude to read Woolworths' online specials pages didn't work; neither could do it accurately. So Pedersen-McKinnon downloaded the current Woolworths sales catalog, uploaded it to ChatGPT and, after a few goes at the wording, asked for recipes of up to eight ingredients, before-and-after prices, the value of leftover ingredients and a shopping list. The week's shop came to $106.85, a saving of $38.85, and the net cost was $3.06 a serving once leftovers were counted. Not every dish was a hit: one child declared 'Salmon is gross.'

Turn leftover holiday ham into a gluten-free fried rice with ChatGPT

After Christmas, Shasta Kearns Moore's household still had a lot of ham. Moore asked ChatGPT for healthy ideas without too many vegetables, since the people at home aren't big fans. The first five ideas didn't fit, because one eater is gluten-free and the family prefers Asian cooking, so Moore told it that. ChatGPT came back with new suggestions and, on request, a recipe for a simple fried rice dish that was 'a big hit.' Moore says it took seconds and saved 'time, money and a headache.'

Generate 30 diabetes-friendly recipes from your own ingredient list with Claude

After a type 2 diabetes diagnosis, Edwin van Wijk had the Claude CLI build a categorized list of ingredients available in a standard Dutch supermarket, removed everything disliked (cheese included), then wrote goals and constraints in a CLAUDE.md file. Claude generated 30 varied recipes plus a table showing each one's main protein, carbohydrates and why it suits a diabetic. Each recipe van Wijk cooks gets printed for the recipe binder in the kitchen. Alongside exercising more and other diet changes, van Wijk reports losing almost 4 kg in four weeks and a glucose drop from around 15 to around 6 mmol/L.

Adjust a 1,200-calorie diabetes meal plan day to day with ChatGPT

Ankush Joshi has lived with type 2 diabetes since 2012 and used ChatGPT to build a 1,200-calorie, high-protein, high-fiber plan of four meals a day. When a day goes off plan, Joshi asks for quick changes, such as moving calories toward dinner or adding 20 g of protein without adding calories. After 2 laddoos and a small chakli at a festival, ChatGPT estimated the extra calories and suggested lighter swaps and a 45-minute walk. Joshi says it is less accurate for homemade, restaurant and festival food, and the rule stays: 'NEVER change medicines based on AI.'

Work out what recipe a mystery grocery order was for with Copilot

Anthony Blake's online grocery order arrived with a set of ingredients that looked as if they'd been added automatically from a recipe page, maybe via an Amazon Echo. Nobody knew what the ingredients were meant for, and searching Google didn't help. So Blake asked Copilot to check the web for a recipe using that ingredient list. Its answer: an Asian coleslaw, which Blake says made perfect sense given the household's tastes.

Check suspicious texts and emails for scams with Norton in ChatGPT

Tom's Guide writer Amanda Caswell added the Norton scam-detection app inside ChatGPT and tested it on real messages by uploading screenshots. The first test, a text about a part-time job, came back 'Unknown,' with only general advice to be careful with unsolicited job offers. A real email claiming a payment had been declined, which looked believable because Caswell had just updated a credit card, was flagged within seconds with 'high confidence' as phishing aimed at payment details. A vague money-making email from someone named Mark was flagged with high confidence as a 'gambling scam,' and a real Lollapalooza lineup email that had landed in spam was judged legitimate, with a note to stay cautious.

Shortlist used Volvos and prep dealer questions with ChatGPT

Sara Clarke, replacing a car bought in 2010, used ChatGPT to search dealerships within a few hours of the Washington, DC, area for Volvo XC60 and V60 models, narrowing by year, mileage, trim, features and color. Its results were similar to Clarke's own searches, though it occasionally surfaced cars that were older or more expensive than requested. It then drafted questions for dealers, such as asking for the CPO inspection checklist and a Carfax or AutoCheck report, whether dealer add-ons were required and the out-the-door price with all taxes and fees. Clarke used those alongside consumer guides to collect quotes by email, phone and text. Its financing-office advice was useful, but it kept recommending steps Clarke had already done, such as asking for all-in pricing. Clarke said it didn't guarantee a better price but kept the process organized and less stressful.

Build a custom ChatGPT reading tutor for a dyslexic child

Arlyn Gajilan's 11-year-old son, Tobey, who is dyslexic, came home saying, 'Everyone keeps telling me I'm smart, but I'm just so slow.' Gajilan built a custom GPT in ChatGPT to help Tobey with reading fluency, spelling and organizing written work. Gajilan fed it Tobey's report cards, neuropsych evaluation, hobbies and competitive streak so lessons were gamified, set it up to encourage progress even when Tobey struggled, and after each session asked it, 'Why this choice?' or 'What is he learning?' Within weeks, Tobey was gaining confidence. Gajilan later turned the idea into a paid tutoring platform, while saying human tutoring is 'absolutely 100% the gold standard.'