What AI Doing to Craft Is What Ultra-Processing Did to Food

There is a headline circulating this week that has nothing to do with nutrition and everything to do with how I think about health. It concerns India's textile traditions — the block-printers, the weavers, the embroiderers whose knowledge lives in their hands — and how AI tools originally designed to preserve those crafts are now able to reproduce them digitally, at scale, without the craftsperson in the loop at all. My reading is that this is not primarily a story about intellectual property. It is a story about what happens when we extract the output of a living practice and sever it from the process that made it meaningful.

I kept thinking about food.

For roughly two decades, nutritional epidemiology has been building a case against ultra-processed foods — not because any single ingredient in them is poison, but because the processing itself does something to the matrix of the food that whole or minimally processed versions do not. The NOVA classification, now widely used in public health research, groups foods not by nutrients but by the degree and purpose of processing. And the consistent finding across large cohort studies — some tracking hundreds of thousands of people over a decade or more — is that higher ultra-processed food intake is associated with elevated risks of cardiovascular disease, type 2 diabetes, depression, and all-cause mortality, with effect sizes that are clinically meaningful even after adjusting for overall dietary quality. We are not talking about small signals in noisy data. We are talking about effects in the range of 10 to 50 percent increased risk depending on the outcome and the population studied.

The mechanism is genuinely not fully understood, which I will say plainly rather than dress it up. Candidates include disrupted gut microbiome diversity, hyperpalatability overriding satiety signals, displacement of fibre and phytonutrients, and additives whose long-term effects at cumulative exposure levels we are still characterising. 'We don't know the exact pathway' is not a reason to dismiss the association. It is a reason to sit with appropriate uncertainty while the picture sharpens.

Here is where the textile story and the food story meet. The craftsperson's knowledge is inseparable from the act of making. The weaver's body learns tension, rhythm, the resistance of a particular cotton in a particular season. That knowledge cannot be fully extracted into a digital file without something being lost — not sentimentally, but functionally. The pattern survives; the adaptive intelligence that produced it does not transfer.

Cooking is the same. What whole-food cooking transfers to the eater is not just nutrients on a label. It is fibre structures that slow glucose absorption, water content that supports satiety, bacterial diversity from fermented or lightly processed ingredients, and — this is the part that sounds soft but has emerging mechanistic support — the sensory anticipation of a meal that takes time, which itself appears to prime digestive enzyme activity and appetite-regulating hormones before the first bite. When we extract the 'food' from the act of preparing and eating it in recognisable form, we do not simply get convenient nutrition. We get something with a different physiological fingerprint.

For readers observing Islamic practice, this lands with particular texture. The Sunnah traditions around eating — sitting, eating slowly, stopping before fullness, beginning with intention — are not wellness theatre. They describe a relationship with food that modern appetite research keeps independently rediscovering. Eating rate, for instance, directly affects ghrelin and GLP-1 signalling; slower eating consistently produces earlier satiety at lower caloric intake across trials. The tradition arrived at the practice by a different route, but the physiology it produces is the same physiology the randomised trials are measuring.

For readers who reach for secular framing: slow, recognisable, hand-adjacent food is not a class signifier or a luxury protocol. It is, to the extent the evidence can be read, closer to how human digestive systems developed across evolutionary time. The processed exception became the statistical norm only in the last fifty years. The body has not caught up.

So what is the practical thing you can do today — not the ideal protocol, the today thing?

One meal this week, cook something you can recognise at every stage. Not a twelve-step project. A handful of lentils, an onion, water, whatever spicing your culture reaches for. The point is not perfection or virtue. The point is giving your gut something it can interpret without a processing intermediary. Notice whether you feel full differently. That is your personal trial, n of 1, no controls, but genuinely useful data about your own system.

The AI textile story will resolve in courts and policy rooms I am not qualified to opine on. But its underlying question — what is lost when we digitise a living practice and remove the practitioner — is one that nutrition science is answering, slowly and imperfectly, for food. The answer appears to be: more than we initially assumed.

We do not need to romanticise hand-weaving or hand-cooking to notice that the extraction model has costs. Noticing is enough to start.