BRENNEN

Industries / Retail, hospitality and personal services / Hospitality and tourism

Restaurants and food service

Restaurants and food service cook meals and serve them to people who eat them on the spot or take them away, in restaurants, fast-food outlets, catering kitchens and canteens.

1766

Mathurin Roze de Chantoiseau opened a shop on the rue des Poulies in Paris that sold restorative broths, then called restaurants, by the historian Rebecca Spang's dating; Roze himself gave 1767, and the cookery reference Larousse Gastronomique credits a man named Boulanger around 1765, of whom Spang found no record.

parisfoodhistory.blogspot.comnationalgeographic.com

46

Milestones achieved across 7 domains.

11

To watch for: announced, not yet confirmed outside.

14

Open deliverables. Nobody has shipped them.

In one minute

Where it started

The restaurant, a room where strangers sit at separate tables and choose what to eat at any hour, began in Paris in the 1760s as a shop selling restorative broths. Canteens grew with airlines, schools and hospitals: United Air Lines opened its own flight kitchen in 1936, and the United States made school lunch national policy in 1946. From the 1950s, chains and new machines such as the pressure fryer and the sushi conveyor turned restaurant cooking into steps that could be repeated in thousands of places.

Where it stands

The US National Restaurant Association projects US restaurant and foodservice sales of $1.55 trillion and 15.8 million jobs in 2026, and China's statistics bureau put the country's catering income for 2025 at 5,798.2 billion yuan. Ordering has moved to apps, screens and delivery platforms, and chains have begun to test machines that assemble meals, but cooking, cleaning and serving are still done mostly by hand. Hygiene rules, inspections, wage laws and scheduling laws differ from one country and city to the next.

What is open

Much of what the industry knows about its own work is unmeasured or kept private. No one has published open data on food waste per meal across many kitchens, full cooling curves for common batch foods, kitchen heat station by station through a summer, or how often blind customers can finish an online order. Robot arms that learn a task from demonstrations, and AI models that read menus, manuals and photos, now let a small team build these tests and datasets, and attempt jobs such as wrapping a burrito by machine.

Domains

AchievedWatch forOpen

Narrower industries

Food trucksCatering

Advances to build with

2 research notes

Open in restaurants and food service

3 of 14 open deliverables

Open · Ordering and paying

Menu options test set

A public set of restaurant menu photos with a checked answer key, and a scorer. The key lists every item and price and, above all, every option group: sizes, choice of side, add-ons and their prices, and how many choices are required or allowed. These options are what a delivery app or an ordering bot must get right to take an order. The scorer compares a model's structured output with the key, item by item and option by option.

Built with what exists

A phone camera; an open labeling tool such as Label Studio; a JSON schema for menus with option groups and minimum and maximum choices; vision-language models, which read images and text, such as Claude Fable 5.2 (see the Note "Three working devices") and an open model such as MiniCPM-V; Python for scoring.

Credit · unclaimed

Open · Cooking and prep

Test kitchen repair assistants

A public test of AI repair assistants for restaurant machines. Write 300 fault questions about common machines, such as soft-serve machines, fryers, combi ovens, dishwashers and ice machines, each built from the maker's published operator or service manual: the model, the error code or symptom, and the manual's next step, with the page reference. Mark each step as one a restaurant worker may do or one only a qualified technician should do, such as work on gas lines, mains wiring or refrigerant. Then score each assistant's answer as correct, wrong but harmless, or unsafe, meaning it told a worker to do a technician's job. Nobody touches a machine: the work is reading manuals and scoring text.

Built with what exists

Operator and service manuals from makers' document libraries, such as Henny Penny's; pdftotext to pull the error-code tables; a spreadsheet for questions and answer keys; general AI models such as Claude Fable 5.2 (see the Note "Three working devices") and an open model run locally on llama.cpp, which is built on the ggml library (see the Note "A first stable number for local AI"); repair assistants whose terms allow testing; Python for scoring.

Credit · unclaimed

Open · Food safety

Cooling curves for big batches

A public set of full cooling curves for foods that kitchens cook in bulk and chill for later, such as rice, chili, stock and braised meat, and a calculator that predicts from the food, batch size, container depth and cooling method whether a batch will cool from 57°C to 21°C (135°F to 70°F) within 2 hours and to 5°C (41°F) within 6 hours in all, as the FDA Food Code requires. Cook each recipe in a test kitchen, cool it by the methods restaurants use (walk-in refrigerator, ice bath, ice wand, shallow pans, blast chiller), log the temperature at its centre every minute, and then throw the food away.

Built with what exists

Waterproof DS18B20 or type K thermocouple probes read by an ESP32 or Raspberry Pi Pico microcontroller, or commercial data loggers; a reference thermometer checked in ice water; Python with SciPy to fit the curves; Zenodo to publish the data.

Credit · unclaimed

Where it sits