I've had my genome read twice. The first time was 23andMe: a chip test, which reads a few hundred thousand positions chosen in advance, roughly 0.02% of the text. The second was sequencing.com: true whole-genome sequencing, all six billion letters. One gave me a small amount of data, the other gave me thousands of times more.
Neither taught me anything about myself.
A wall of predispositions
What sequencing.com shows you is a list of genes you might have a predisposition for: conditions that could, in some statistical sense, be in your future. I sat with our clinical geneticist, Pilar Guatibonza Moreno, MD, and went through my results with her. Her professional assessment of that list was blunt, and I'll soften it only slightly: that list is AI-generated output with no person behind it.
What a marker tells you
Here's why that matters. I didn't see anything alarming in my own results. But some people will, and a predisposition marker stripped of context reads like a diagnosis. People get stuck on something might happen to me. That anxiety can do more damage to their health than the marker ever will.
Genetics has a name for why that fear is usually misplaced: penetrance. Penetrance is the proportion of individuals with a particular variant who express the associated phenotype or condition. If everyone carrying a variant got the disease, it would be fully penetrant. Almost nothing is. Very few genotypes are completely penetrant[1]. Most disease-associated variants show what geneticists call incomplete penetrance: plenty of people carry them, live full lives, and never develop the condition at all[2].
The numbers make the point better than I can. Some BRCA2 variants, among the most feared findings in consumer genetics, have penetrance as low as 47% in some populations[1]. That's a coin flip, not a sentence. Some pathogenic variants sit below 5%: from twenty individuals carrying the pathogenic variant, nineteen are never affected[3]. A well-known Parkinson's-linked variant elevates risk, but in most heterozygous individuals the disease never manifests[4]. This is also why healthy people routinely carry variants that look alarming on paper, and why a condition can skip through an unaffected parent to appear in a grandchild[5].
For a variant that isn't fully penetrant, the genome alone cannot tell you who will develop the condition and who won't.
And penetrance is the part a dashboard can't handle: a population statistic, one that can only be estimated from large population studies[1]. Whether it means anything for you depends on your family history, your environment, your other variants, your age. Context is exactly what a report generated by a pipeline doesn't have, and exactly what a geneticist who knows your family history does.
So a platform that hands you a list of predispositions with no person attached is handing you probabilities dressed up as predictions.
One honest exception, because rigor matters: a small number of variants really are fully penetrant. Huntington's disease above a certain repeat length is the textbook case[6]. Even so, neither 23andMe nor sequencing.com can diagnose it: that takes professional-grade whole-genome analysis or a dedicated CAG repeat assay. Those findings are serious, and they're precisely the ones that should never arrive via dashboard. They deserve a conversation with a professional. Which is rather the point.
The disease model
Every platform in this space runs on the same logic: find out what diseases you might get, so you can change something before they arrive. The entire product is organized around what could go wrong with you.
We think that's backwards. Your genome is the most personal document you'll ever own, and most of what it has to say is about how your body works rather than about disease. How you respond to foods. Where your effort goes furthest. What your family history means for your goals rather than your fears.
A person, not a platform
So at darkMatter we're trying something different, and I'm the first test subject.
Our Clinical Geneticist started not with my genome but with me: my family history in depth, my health goals, what I want to change. Only then does she go to the genome, looking for what's useful for my day-to-day life instead of what's frightening. Nutrition, food response, training, habits. Findings in service of goals.
I won't share what she found — that's the point, actually. It's mine. But I'll say the experience is nothing like scrolling a dashboard.
People talk to their therapists weekly and think nothing of it. Almost nobody has a relationship with their geneticist.
We think that's temporary. The genome is a lifelong document, science keeps rereading it, and the questions you bring to it change as your life does. That deserves a conversation, not a login.
This isn't a product you can buy from us yet. It's what we're building, and I'm writing this from inside the experiment. If you want to hear when it's real, the newsletter below is where we'll say it first.