The Patent, Explained

Written by Dr Jatin Joshi, co-founder and Chief Medical and Scientific Officer of Personally, and a researcher at the University of Oxford's Centre for Evidence-Based Medicine. I have a financial stake in this company, and you should weigh what follows with that in mind.

"Adaptive" is one of those words the supplement industry has half worn out. It gets printed on plenty of products that are not adaptive in any meaningful sense, a single fixed formula with a personalised label on the front. So your scepticism is the right starting point. The honest question is not whether we use the word, it is whether there is anything real underneath it, and how you would tell.

Let me answer that by showing you the actual problem we set out to solve, then the method we built to solve it, and only at the end the patent, which is not the headline here. The patent is just the artifact that says the method was real enough and new enough to be recognised as one.

Making a formula that genuinely adapts to a person is difficult for reasons that are easy to underestimate until you try.

The first is that ingredients do not sit in neat isolation. You cannot simply take everything that might help you and add it together, because nutrients interact, some compete for absorption, and several have sensible upper limits you should not cross. A real formula is a balancing problem, not a shopping list, and the more ingredients you are working with, the more the balance matters.

The second is that you cannot finely adjust a capsule. Pills and pre-made blends come in fixed doses, so a conventional product can only ever offer you the doses that already exist on a shelf somewhere. If the right amount for you sits between two pill sizes, or if it is an unusual number rather than a round one, a capsule cannot give it to you. You end up rounded to whatever was manufactured in bulk, which is not the same as the amount that actually suits you.

The third is that you are not the same from one season to the next. Your diet shifts, your bloods shift, the demands on your body shift, and a medication you start or stop can change the picture again. A fixed formula is, at best, right for the version of you that filled in the form on the day you bought it. The moment your circumstances move, it is slightly wrong, and it has no way of knowing.

And the fourth, quietly the hardest, is doing all of this consistently and defensibly for a lot of people at once. Any competent nutritionist can think carefully about one person over an afternoon. Building a system that applies the same careful reasoning to thousands of different people, every cycle, without cutting corners or drifting into guesswork, is a genuinely different kind of problem.

There are two halves to how we handle this, and they matter equally.

The first is the engine that decides what should go into your formula. It takes a defined set of inputs about you, what you tell us in the questionnaire, your blood markers, and, if you use a wearable, signals from that, and it applies a set of rules to work out your formula. The important word is rules. This is not a generative artificial intelligence improvising a blend, and it is not a black box that produces a different answer depending on its mood. It is deterministic, which is a plain way of saying that the same inputs always produce the same considered formula, for reasons that can be written down and followed. That property is doing a lot of quiet work, and I will come back to why it matters.

Those inputs are exactly that, inputs. The engine uses them to personalise your nutrition, not to diagnose you, monitor you, or stand in for anything your doctor does. It reads your data to decide what to put in a powder, nothing more than that.

The second half is that we make the formula from scratch for you rather than assembling it from pills. Because it is blended to order, each ingredient can be set to the precise amount the engine calculated, an exact, often unround number, rather than snapped to the nearest capsule that happens to exist. Our library runs to dozens of evidence-based ingredients, and a real formula typically draws on a large share of them, each one weight-calibrated individually. We do this in our own facility in Boulder rather than handing it to a third party, which is deliberate: when the whole promise rests on making an exact formula to order, the making is not a step you want to lose sight of. That is the part people underestimate. The clever decision about what you need is worth very little if you then cannot actually make it, and making it is a manufacturing problem as much as a scientific one.

Put the two halves together and you have something that can genuinely move with you. When your inputs change, the formula the engine produces can change with them. When they do not, it may not, and that is honest: I am not going to tell you your formula is rewritten every single month for the sake of it. Each cycle it is reassessed against your current data, and you are told whether it changed, why, and what was adjusted. Sometimes the most truthful answer is that nothing needed to.

Here is the turn on your original question, is this real or is it marketing. A fixed formula with a personalised label is marketing. A formula that is genuinely recomputed from your own data, through a defined process, is adaptation. The line between the two is whether the output actually depends on the inputs in a way you could trace.

Ours does, and because the process is deterministic rather than improvised, it can be examined. You can ask why a particular ingredient is in your formula at a particular dose and there is a real answer, a chain of reasoning from your data to that decision, not a shrug about what the algorithm felt like. That is unusual in this industry, and it is the thing I would point to if you asked me to prove the word means something.

Now the patent, late and on purpose. We hold two granted US patents covering the method of building personalised formulations this way, with a third in development. I have deliberately not led with them, because a patent number is not a reason to trust a product and I would be uncomfortable using it as one.

What a patent actually tells you is narrow and worth being precise about. To grant one, the patent office has to be satisfied that the method is new and not an obvious combination of things already known. So the patents are external confirmation that the problem was real and the solution was not already lying around, which is a fair thing to want to know when someone claims to have built something novel. What a patent is not is proof that the thing works. Those are two different questions, and anyone who waves a patent at you as though it settled the second one is hoping you will not notice the difference. I would rather point it out myself.

I will also be straight about scale, since a careful reader gets there anyway. The more finely the inputs are read and the larger the ingredient set in play, the more your formula diverges from the next person's. Early on, with a smaller set of combinations, that divergence is more modest than it becomes as the system grows. The method is the same either way. Its resolution simply increases as it scales, and I would rather tell you that than imply a degree of one-to-one uniqueness we are still building toward.

The word "adaptive" is worth exactly as much as the machinery behind it, and usually that machinery is not there. What we have tried to build is the machinery: a deterministic engine that reasons from your data through rules you could follow, and a bespoke manufacturing process that can actually produce the doses it calculates. The patents are the artifact left behind when a method like that is recognised as new. They are the evidence, not the argument.

If you want to see what that produces for your particular data, the quiz is where it starts. It is a way of finding out what the system would actually build for you and why, not a checkout, and you are free to read the reasoning and walk away.