AI · Product · 2026
Guestimate works out exactly how much food and drink you need for a get-together, whether it's a barbecue, a birthday or a weekend at a country house, and turns it into a shopping list priced with real products from Mercadona, Spain's biggest supermarket. Tell it what you're hosting and for how many, and it handles the maths, the variety and the budget, so nobody runs out and nothing goes to waste. Built in two days.
Try it
Pick an event, set how many guests, and watch the basket fill up in real time.
What it is
A planner for anyone hosting a social event who doesn't want to buy too much or too little. You choose the type of gathering, whether it's a barbecue, a birthday, a dinner with friends or a weekend at a country house, then say how many people are coming and any dietary restrictions, and Guestimate calculates the right quantities, picks concrete products, and prices the whole basket with real Mercadona items. Instead of a static list at the end, a live basket fills up as you choose, showing the running total. For trickier asks there's a built-in AI assistant: type a special request in plain language ("add something vegan", "more snacks for the kids") and it adjusts the list, matching each addition to a real product with its real price.
How I built it
Built in two days as a solo project to take an idea end to end, from data to a deployed product. The hard part wasn't the interface, it was the data: I wrote a scraper that pulls thousands of real Mercadona products with live prices into a Supabase database, and a portion engine that converts "12 people at a barbecue" into specific quantities, with automatic variety so a meat order becomes chorizo, pork belly, chicken and chops rather than one big lump. A weekly GitHub Action keeps prices fresh on its own. The special-request feature runs a language model that returns items to add or remove, which I then pair against the real catalogue so each suggestion inherits a real price and category. A feedback loop closes it: when people report they had too much or too little, the quantities for that event type quietly recalibrate.
Built with