The 700 Ingredients Weren't a Database — They Were a Language for Real Food.

Someone once asked me whether the seven hundred ingredients inside Foodna were a database.

I know it's a kind question, but every time I hear it, something in me wants to correct it. Database is too quiet a word — too much like a warehouse. Shelves, IDs, entries, things stacked in to wait until they're pulled out. That isn't what I made. What I made was closer to an obsession: the seven hundred ingredients weren't a database. They were a language for real food.

Because when an ingredient enters Foodna, it isn't logged as a name. It's understood as a set of signals the body reads.

Let me start with why an ordinary nutrition database wasn't enough — not enough that I had to go and rebuild the whole thing myself.

Most of the databases out there stop at three things: calories, macros, a handful of major vitamins and minerals. A century ago that was a magnificent leap. But the way it sees food is coarse enough to make me restless. It'll tell you how much protein a thing has, but not whether that protein's amino-acid profile is complete, or usable by the body. It'll tell you the "carbs," without distinguishing the kind that spikes your glucose from the kind wrapped in fiber and resistant starch, released slowly. And it's blind to the fact that the same ingredient — raw, cooked, chopped, fermented, paired with fat or with acid — is a completely different sentence to the body.

It records the name. What I wanted Foodna to read was the grammar — what an ingredient, in the real context in which it's actually eaten, is saying to the body.

Back to that carrot (I mentioned it on another page; here I'll come at it differently). An honest system, seeing a carrot, can't just write "beta-carotene." It has to ask: is this preformed vitamin A, or a carotenoid the body still has to convert? Was there any fat alongside, so those fat-soluble compounds actually get caught? Was it chopped, loosened by heat — or gnawed whole and raw? The same handful of questions I'll chase after a tomato (lycopene tends to absorb better cooked, with oil present), after red cabbage (chop it and wait, so the glucosinolates get their chance), after a single egg (the choline and lutein in that yolk — please don't throw them out in the name of "health").

And within this language, the dimension that fascinates me most — and is hardest — is the fifth: nutrition density.

Say "nutrition density" and most people think: lots of vitamins, lots of minerals. What I care about is another layer — the rare, high-value active compounds that quietly fall off ordinary labels and yet genuinely matter. They're counted so rarely that a system reading only the label behaves as if they don't exist. And a large part of my stubbornness in building Foodna was simply refusing to let them be lost by the system.

Let me give you a few of the ones I watch most closely — but first, let me say this plainly: I'm not slapping superfood labels on ingredients. I'm keeping the genuinely valuable signals in real food from being lost by the system. Those two things are worlds apart. The first is marketing. The second is honesty.

A tin of sardines in olive oil — the label gives you protein and fat. But the precious part is the long-chain EPA and DHA, the marine omega-3 studied for its links to inflammation balance and brain function; the iodine and selenium that ride along; the complete protein; the calcium in the little bones you eat whole. The label mentions almost none of it. A glass of cold kefir straight from the fridge (never heated — heat it and you lose the heat-sensitive live cultures) is worth more than its protein and calcium: it carries bioactive peptides formed in fermentation, and kefiran, a polysaccharide that's drawn research interest — none of which you'll find a slot for in an ordinary dairy entry. The purple of red cabbage is anthocyanins; inside a pomegranate seed are the ellagitannins (punicalagins) under study; in green tea, catechins and EGCG; and what's truly valuable in broccoli isn't something it already contains — it's its potential: the precursor to sulforaphane, which the body only gets if you chop, wait, and don't incinerate it.

And the ones even easier to overlook: the resistant starch that rises in rice or potato once it's cooled, the beta-glucans in oats, the choline in an egg yolk, the well-absorbed heme iron and B12 and zinc in red meat, the taurine and creatine and carnosine in seafood. The names sound technical, but every one of them is something real, everyday food already carries. All I want Foodna to do is not miss them.

And at every one of these, I force myself into the most restrained phrasing possible. Not cures cancer. Not detoxes. Not heals. But studied for, a signal for, may support. Because the moment you cross that line, what you're making stops being food intelligence and becomes wellness talk — the exact thing I least want Foodna to become.

Choosing which seven hundred was its own hard fight.

The rule I set myself was: cover real, everyday food — not a pile of superfoods. So the list holds eggs, sardines, beans, red cabbage, kefir, carrots, apricots, tomatoes, all kinds of seafood, all kinds of ferments — Mediterranean, Asian, the things that actually appear on ordinary tables. Because a system that only knows goji berries and chia seeds is useless to real life.

Then the part that drained me most: every single one, cross-checked against papers, reviews, databases. Which signal has solid evidence behind it, which is just a popular claim; which mechanism holds, which has been oversold. I also had to find or generate a photograph worthy of each ingredient. And I tested a fair number of models. It was in the testing that I ran straight into the thing that mattered most, and was hardest to handle —

Models have a deep-set tendency: they'll make everything sound healthy and reasonable. Ask one what's good about any food and it'll hand you a fluent, positive, airtight-sounding paragraph. Everything is "rich in antioxidants." Everything "supports immunity." That indiscriminate goodwill is precisely the danger. Because if everything matters, then nothing does.

So the hardest judgment landed back on the human after all: the hard part wasn't collecting the ingredients. It was deciding what earns each one the right to be seen. Which signal is genuinely important — important enough to count inside those five dimensions — and which is just a pretty line from the wellness circuit, pleasant to hear and hollow underneath. That cut, the model can't make. Only I can. What it takes isn't compute. It's judgment — knowing whether a piece of evidence is hard enough, whether a mechanism deserves to be taken seriously.

Finally, there's one dimension I never intended to concede from the start: the visual.

I insisted on giving every ingredient a photograph with real-food color and editorial dignity — not a backend table, not a clinical dashboard. A tomato red as if just off the vine, olive oil green enough to glow, an apricot still wearing the fine down on its skin. This sounds like aesthetic fussiness. It's actually the logic that closes the whole loop.

Because the reason Foodna exists is to make people want real food again. And on a screen, real food keeps losing to the over-styled. So: if Foodna is going to make people want real food again, the ingredient itself has to have visual dignity. A tomato photographed with real care does something entirely different to a person's desire than a row that reads tomato, vitamin C, lycopene lying in a table. The first makes you want to eat it. The second only lets you know about it.

So no — the seven hundred ingredients really aren't a database. A database only remembers. What I wanted was to make each real food readable again, understandable, worth wanting — through a language I built, piece by piece, to understand every sentence they say to the body.

A warehouse stores things away. A language makes them able to be understood again. What I've been doing, all along, was the second one.