Cookwala / Why
Why an open standard for cooking, and why now
Three things are true at once. Machines are starting to cook. Food is thrown away while people go hungry. And one of the most intimate datasets in any home is about to be collected by appliances. Cookwala is one small, open answer to all three.
The problem, in four parts
1. Machines are learning to move; nobody wrote down how to cook
Home and kitchen robots are shipping or taking orders. Each maker writes its own closed recipes, mostly from a few cuisines, and decides alone what "simmer" means and when chicken is safe. There is no shared, checkable definition of "done" or "never".
2. Unsafe food and unsafe kitchens
Unsafe food causes about 600 million illnesses and 420,000 deaths a year measured · WHO. Cooking machines add new ways to be wrong: hot oil estimated by a clock, a recipe that says 240 °C, an agent that obeys text it read.
3. Food is thrown away while people go hungry
About 14 % of food is lost between harvest and retail measured · FAO; about 1.05 billion tonnes were wasted in 2022 measured · UNEP; about 733 million people faced hunger in 2023 measured · SOFI. Food banks rescue what they can with phone calls and spreadsheets that differ by site.
4. One of the most intimate datasets a home can produce
A robot that cooks well knows a household's schedules, layout, children, health, religion and budget. No rule says what it may do with them, or who else may know.
The answer: three things a machine can check
What to make. A recipe as typed steps with a shared vocabulary of operations, each with a physical envelope: the physical band an operation must stay in, for example simmering is water at 85 to 96 °C: simmer is water at 85 to 96 °C; deep frying is oil at 160 to 190 °C.
When it's done. Every heat step ends on something a device can measure: a temperature, a food-state cue, a time window, with a defined path when it times out.
What must never happen. Safety limits enforced on the device itself. No recipe, agent, message, extension or operating mode can raise them; a maker changes a default only through a documented safety case and a signed firmware release. A device that cannot verify a step refuses before it heats anything. No oil thermometer means no deep frying.
Around that core: signed records anyone can check; agents that act only under a mandate: a signed permission from a person that says what an AI agent may do for them, with limits and an expiry and treat all text as data; a profile for food banks that carries no personal data; a profile that keeps a household's facts at home.
Why open, why now
Standards that arrive after the products are written by the products' makers. The window to write the rules for cooking machines in the open, before they are in millions of homes, is the next few years. An open standard lets a small maker, a food bank and a university lab use the same format as the largest company, and lets regulators point at one thing.
Cookwala is royalty-free, model-neutral and device-neutral, governed in public today and headed for a neutral foundation. It started from one family's Egyptian recipes on fifi.cooking.
It started with a family cookbook
Cookwala began with a family archive of Egyptian home cooking, the recipes of Dr. Fatma Alkawokgy (1943 to 2026), published at fifi.cooking in 25 languages. They were written for people. "Fry until golden." "Add water to cover." "Let it simmer until it is ready." Anyone who grew up in that kitchen knows exactly what those lines mean. A machine does not, and neither does a stranger. When one of those dishes was described so that an oven, a robot or a food bank could check it before cooking, it turned out that nobody had written down what "simmer" is, when chicken is safe, or when a step must be refused. Cookwala is the attempt to write that down, in the open, so that the dishes outlive the machines and the machines refuse rather than guess.
What it sits beside, and what it is not
None of these is a machine-checked contract between a recipe and a device. Cookwala is only that, and uses the others where they fit.
| Schema.org Recipe | Describes a recipe for search engines: ingredients, times, nutrition. No operations, no end conditions, nothing a device verifies. A Cookwala recipe can be published with Schema.org fields for discovery. |
|---|---|
| LeRobot (Hugging Face) | A dataset format for what a robot did: observations and actions. It does not say what a recipe is or when a step is unsafe. Cookwala exports consented execution logs to it. |
| HACCP | A management system for people: hazards, critical control points, records. Cookwala encodes control points so a machine can check them; it does not replace a kitchen's HACCP plan. |
| Matter (CSA) | Device control: set an oven to a temperature, read a probe. Cookwala says what the oven should be asked to do and when to refuse; a mapping table lives in bindings/matter.json. |
| ROS 2 | Robot middleware: how a robot moves and talks to itself. Cookwala sits above motion and ships ROS 2 interfaces for the task layer. |
Three goals, told honestly
Help end hunger
Hunger is driven by poverty, conflict, climate shocks, prices and policy. Cookwala's part is real and partial: surplus reaching a plate safely, programs that can compare their numbers, and later, farms and stores that plan to need. We never claim Cookwala ends hunger.
Make people healthier
Food-safety limits a device must meet and record; nutrition rules a program can apply and a dietitian can review. General guidance only; nothing medical.
Put robots to work for people
Robots that cook every cuisine safely, learn with consent and credit, serve people who cannot cook for themselves, and give human time back, with a labour voice in governance.
Three horizons, without hype
Now
- Safer devices that refuse rather than guess
- Recipes that work across machines, with the cook's name on them
- Food banks rescuing more, safely, with phones and spreadsheets
- Agents that can't be tricked by text in a recipe
- Open tools you can run today
Next
- Home and commercial robots that cook every cuisine
- A consented dataset that keeps cultural recipes alive
- New roles: recipe engineers, food-robot technicians, certifiers, rule-pack reviewers
- Less waste; cheaper, healthier meals
- A standard regulators point to
Later
- Cooking as infrastructure: nutritious food available to anyone, anywhere, including disaster zones and people who can't cook for themselves
- Less waste, lower emissions
- Human time returned to people
- Machines that inherit the world's culinary heritage instead of erasing it
Each horizon is an assumed direction, not a date, a forecast or a count. What exists today is listed on the roadmap with a status per item.
What we got wrong so far
The first one-liner ("the world's first and largest robot cooking recipes index") overstated what existed. The first schemas accepted unknown fields. The first simulators used a strawman baseline. We wrote critiques of our own idea from six viewpoints (a red team, a technical reviewer, a demanding adopter, an AI lab, a relief agency, a health agency) and published them with responses; no outside review has been received yet. Critiques and responses · The backstory and the gap list