There is a problem at the heart of nutrition that the industry has, for decades, refused to name. It is this: we are still operating on a model that treats the human body as though it were a boiler. Food enters, calories emerge, and the arithmetic, one is told, should settle the matter. In 1950, this was forgivable. In 2025, when we have sequenced the human genome, started mapping the microbial ecosystems in the gut, and built sensors capable of tracking blood glucose in real time, it is something closer to embarrassing. And yet the official guidance remains, more or less, “eat your vegetables and watch your portions”. One you could find displayed on a poster in a waiting room.

What I want to propose here is not a refinement of this model but a replacement. Food is not fuel. Food is code. It is executable biological information: molecular instructions that interact with microbial communities, modulate immune responses, alter gene expression, and reshape metabolic pathways over time. When we eat, we are not adding energy to a furnace. We are running a programme. And like any programme, the outcome depends entirely on the operating system that interprets it.

That operating system is your microbiome.

This, I believe, is the piece most people are still missing. The oral and gut microbiomes are not supporting players in the drama of digestion. They are the operating system itself. It is in the mouth where the code first makes contact with its interpreter: the oral microbiome initiates carbohydrate breakdown, modulates the nitric oxide pathways that regulate blood pressure, and seeds the microbial populations further downstream. The gut microbiome then executes the deeper programme, metabolising what you have eaten into short-chain fatty acids, neurotransmitters, inflammatory mediators, and signalling molecules that reach your immune system, your brain, your endocrine pathways. Together, these communities do not assist digestion. They govern it. Two people eat the same banana. One spikes past 180 mg/dL. The other barely registers. Same input, radically different output. The Weizmann Institute demonstrated this in 2015 across 800 participants, and the microbiome was among the strongest predictive variables. The food did not change. The operating system did.

Once we accept this framing, everything about how we approach nutrition must change. If the microbiome is the OS, then food is the code that runs upon it. And code, unlike fuel, can be designed. It can be versioned, adapted, made conditional. One does not write the same programme for every machine, and one certainly does not write it once and expect it to remain valid as the machine changes. Yet that is precisely what a diet plan is: a static programme deployed on a system that updates itself daily.

And so we must ask the uncomfortable question. If food is code, then, what is junk food? It is not simply “unhealthy.” It is corrupted code. Ultra-processed food is engineered with extraordinary precision, but not to communicate with your biology. It is engineered to override it. The exact ratios of sugar, salt, and fat are calibrated to bypass satiety signalling, to trigger dopaminergic reward loops, to keep you eating past the point where your body asked you to stop. It degrades the operating system itself: we now have robust evidence that ultra-processed diets reduce microbial diversity, weaken gut barrier function, and promote chronic low-grade inflammation. In the language of software, this is not a bad programme. It is a programme that corrupts the OS on which it runs. And we are feeding it to entire populations, beginning in childhood.

To be fair to the field, it has not been entirely standing still. Personalised nutrition has made genuine progress. Nutrigenomics has demonstrated that genetic variation affects how individuals metabolise specific nutrients: polymorphisms in MTHFR alter folate metabolism, variants in FTO influence appetite regulation, CYP1A2 determines whether one is a fast or slow caffeine metaboliser. This is real science, and it matters. Companies working in this space have moved well beyond simple questionnaires, combining genetic data with blood biomarkers, dietary logging, and sometimes wearable data to construct meaningfully differentiated nutritional profiles. That represents a legitimate advance over the one-size-fits-all guidelines of the last century.

But here is the limitation, and it is a structural one. Nutrigenomics tells you about your hardware. It tells you what your genome predisposes you to, how your enzymes are wired, where your metabolic architecture has features worth knowing about. What it does not tell you is what your operating system is doing right now. Your genome does not change from week to week. Your microbiome does. It shifts with your diet, your sleep, your stress, your medication, your environment, your age. A nutrigenomic profile gives you a fixed map of a system that is, in reality, constantly in motion. It is necessary. It is not sufficient. To truly personalise nutrition, one needs both: the hardware specification and a live reading of the OS.

It is this convergence that makes contextual nutrition possible, and I want to be precise about what this term means, because it represents a step beyond what even the best current personalised nutrition can offer. Contextual nutrition treats your food strategy as a live query against your full biological state: your genome, yes, but also your microbial composition this week, your inflammatory load, your sleep debt, your hormonal phase, the antibiotic course you finished last month that quietly restructured your OS. The tools to support this are arriving with remarkable speed. Continuous blood glucose monitors provide real-time metabolic feedback. Wearable biosensors are beginning to capture immune markers. Microbiome sequencing is maturing from taxonomy toward functional metabolomics. Together, these form something without precedent: a live debugger for the biological programmes we run three times a day.

Where should this land first? In the consumer market? In the hospital? And here I wish to be direct, because someone should be. It is, to my mind, amazing that modern medicine can perform targeted gene therapy and robotic surgery, and then feed the recovering patient a tray of corrupted code selected by a catering contract awarded on price. If food is code, hospital food is legacy software: nobody wrote it with intent, nobody maintains it, and nobody measures what it does. Meanwhile, the clinical evidence is already substantial. Specific amino acid profiles modulate tumour metabolism. Short-chain fatty acids from microbial fermentation support gut-barrier integrity during chemotherapy. Timed protein intake accelerates post-surgical tissue repair. The code exists. It is simply not being deployed.

Within a decade, the best clinical institutions will prescribe food matched to biomarkers, microbial status, inflammatory trajectory, and recovery phase. The same logic extends to every institution that feeds people at scale: schools, care homes, workplace canteens. In Europe, roughly a third of meals are consumed in institutional settings. These are environments in which one could reprogramme population-level metabolic health without requiring anyone to change their personal behaviour. But only if one treats institutional food as executable code designed for biological outcomes, not as a line item in a procurement budget.

The food-as-code framing also rewrites, in ways not yet fully appreciated, the relationship between agriculture and health. Soil has a microbiome. Plants have a microbiome. Your gut has a microbiome. This is one stack, not three separate systems. The microbial diversity of the soil determines the phytochemical profile of the crop, which determines the biological instructions that food carries into the human OS. Montgomery and colleagues showed in PeerJ in 2022 that crops from regenerative farms carried measurably higher concentrations of anti-inflammatory phytochemicals, vitamins, and minerals compared to the same crops grown conventionally on neighbouring land. Regenerative agriculture is not a feel-good narrative about farming. It is, quite literally, an upgrade to the source code.

Regulation, it must be said, is completely unprepared. Current food law asks whether a product is safe and whether the label is honest. These are fine questions. They are also wholly insufficient. When a company engineers a prebiotic blend that demonstrably shifts microbiome composition toward reduced systemic inflammation, that product is neither a drug nor a conventional food. It is a biological programme. And we have, at present, no regulatory category for biological programmes. By 2030, we will either build frameworks capable of evaluating food as functional code, or we will have a grey market of unsubstantiated “gut health” claims with no meaningful oversight. I know which outcome the industry’s lobbying budgets favour.

I wish to finish with an observation that should be uncomfortable. The industrial food system already writes code for your body. It has been doing so for decades. It simply optimises for the wrong outputs: shelf life, margin, and the dopamine hit that ensures you buy again tomorrow. That is code too. It is “hostile” code. The question has never been whether food will be treated as a programme. It already is. The question is who writes it, what it optimises for, and whether anyone is paying attention to the operating system on which it runs.

Et voilà. Nous pensions parler de nourriture.