Your DNA Is Finished, but the Product Cannot Be

I uploaded my Ancestry DNA file to Genomelink earlier this year.

The upload itself was strangely anticlimactic. A file moved from one company to another. The DNA inside it did not become newer. My biological sample had already been collected, converted into rows of positions and letters, and interpreted once by Ancestry. Genomelink did not need another tube of saliva. It needed the same old file and another reason for me to look at it.

That is the unusual thing about consumer DNA products. The underlying material is finished before the product relationship begins.

My genotype is not going to receive a software update tomorrow. But the interface around it can keep updating forever. A new trait can appear. A reference population can become more detailed. A Viking sample can be added. A region can split into smaller regions. A distant relative can upload their own file. The same biological object can return as a different answer every few months.

Genomelink has spent nearly a decade building a company around that gap between fixed data and unfinished meaning. Its newest product, YourRoots, makes the gap much larger.

The first idea was not another DNA report

Genomelink began in 2016 as a weekend project between Tomohiro Takano, Yuta Matsuda and Kensuke Numakura, who were already working in genomics in Japan. Their original vision was not simply to sell people another ethnicity estimate. They wanted to build a “DNA App Store,” a platform where a person could store genetic data once and connect it to many different applications.

The company described this as a personal DNA cloud. The user would retain ownership of the raw data while allowing selected services or researchers to access it. It was a very platform-era idea: DNA as infrastructure, apps as the interpretation layer, the user as the supposed owner of the underlying asset.

The founders moved toward the United States with almost no traction. Takano later described arriving in the Bay Area without a stable visa, meaningful funding or a product that people were already using. The team was preparing to return to Japan when Berkeley SkyDeck accepted them. They later passed through StartX and Y Combinator, and the small company survived long enough to keep narrowing the idea into products people would actually buy.

There is something almost too neat about the founding story. They began with the largest possible idea, a universal platform for consumer genomics, and then had to discover what a person would do inside it after uploading the file.

Infrastructure is not useful merely because it exists. An app store still needs apps.

A fixed file creates a retention problem

A DNA test is naturally intense and naturally brief. I order the kit, wait for it, produce the sample, wait again, and then receive the revelation. For a few hours or days I click through the regions, traits and possible relatives. I send screenshots to family members. I compare the percentages with what I thought I knew.

Then the main event is over.

This is a difficult shape for a recurring consumer business. The customer has already completed the action with the greatest emotional charge. The file is already on the server. The company cannot ask me to become genetically different next month.

It can, however, interpret me again.

Genomelink’s earlier product did this through an expanding library of traits and reports. The same raw file could be used to say something about nutrition, fitness, personality, cognitive tendencies, ancestry or similarity to ancient samples. Some of these interpretations may genuinely improve as research and reference datasets change. The fact that the DNA is fixed does not mean that scientific knowledge around it is fixed.

But the same structure also creates an almost unlimited packaging opportunity. A fixed data object can be divided into hundreds of questions. Every question can become a report. Every new database can become a new version of the self.

The product does not need my DNA to change. It needs the list of things I might want to know about myself to remain unfinished.

YourRoots makes identity into an ongoing task

I also have a large family tree on Geni. From far away it looks complete. There are names, dates, marriages, children, villages and lines that extend backward for centuries. When I zoom in, it becomes a field of uneven evidence. One branch is dense. Another stops with a person whose parents are unknown. A place name changed language three times. Two records disagree on a birth year. Someone copied a family connection from someone else, who may have copied it from somewhere else again.

A family tree is much better recurring software than a DNA report because it can never really be finished.

That is what YourRoots adds to Genomelink. Launched as a separate genealogy brand in 2025, it lets people import a GEDCOM family-tree file, place ancestors on an interactive global map, trace migration paths, compare trees with DNA matches and ask AI to investigate the missing edges. Its AI Ancestor Finder looks for parents of “end-of-line” ancestors. AI Deep Research studies one person and proposes records and next steps. Auto Research Mode can send new hints daily or weekly so that the tree keeps growing while the user is somewhere else.

The language is revealing. The DNA upload is an event. The family tree is a process.

There is always another dead end to investigate, another surname to cluster, another migration path to draw and another possible relative to contact. The product no longer has to keep inventing isolated traits around the same file. It can connect the file to a living network of records, people and unresolved questions.

YourRoots is therefore not just a genealogy feature attached to Genomelink. It is a possible answer to the company’s deepest product problem. It turns a finite biological dataset into an unfinished world.

A map makes uncertainty look settled

I understand why I want this product. I have spent years moving between Estonia, Taiwan and other places, repeatedly trying to explain how location changes a person. A map that can turn my family tree into migration paths is almost custom-made for the kind of story I already want to tell.

It would also make the tree feel more coherent than it really is.

Maps have a way of converting uncertain text into confident geography. A database may contain an old village name, a parish, a county, a spelling error or nothing more specific than a country. The interface resolves it into a pin. A line joins one generation to the next. The movement becomes visible. The family appears to have travelled through history with the same clarity that I travel through Google Maps.

YourRoots itself acknowledges that locations can be inaccurate, especially when several towns share a name, historical names no longer exist or the source text is unclear. Users can adjust pins. The AI features also return sources for verification rather than pretending every suggestion is proven.

Still, the emotional effect of the interface arrives before the verification work.

A possible parent in an ordinary search result is a hypothesis. A possible parent inserted beside the missing edge of my family tree is almost a person returning. A speculative location in a paragraph is uncertain. A pin on the map feels found. A generated migration line looks less like a guess than a route.

In most AI products, a hallucinated detail is annoying. In genealogy, it can become inherited memory.

The danger is not only that the system might be wrong. It is that the visual language of trees, lines, generations and maps can compress several different levels of confidence into one family story. The user may technically be invited to verify the source, while emotionally receiving the connection as an answer.

Ownership becomes the onboarding story for another platform

Genomelink’s original promise was unusually attractive because it framed the user as the owner. Instead of leaving my genetic data trapped inside Ancestry or 23andMe, I could download it, upload it elsewhere and choose what to connect it to. The DNA App Store would give the file mobility.

I did exactly that.

But the value of Genomelink and YourRoots grows when the company can aggregate more of the things around the file: DNA uploads from different testing companies, family trees, historical records, ancient genomes, user corrections, matches and research behaviour. The more complete the platform becomes, the less it feels like a neutral safety-deposit box and the more it becomes another central place where identity is assembled.

This is not necessarily a contradiction. User ownership and platform aggregation can reinforce each other. A portable file makes it easier for a new platform to persuade people to upload. Cross-platform matching becomes useful precisely because Ancestry, 23andMe and MyHeritage are separate silos. Genomelink can say, correctly, that it is giving the user more control while also building value from the combined network of everyone who exercises that control in the same place.

Ownership is not the opposite of aggregation here. It is the onboarding mechanism.

The person arrives because the data is theirs. The platform becomes powerful because the person leaves a copy there.

The self is an unusually renewable product

I began looking at Genomelink because I wanted to know how large the company was. It is tiny, around fifteen people according to Y Combinator. That makes the breadth of the ambition more interesting. This is not a giant laboratory company sequencing millions of new samples. It is a small software company finding more interfaces through which existing samples can become meaningful again.

The sequence of products makes sense when viewed this way.

First, the DNA file produces traits. Then it produces ancestry reports. Then it produces ancient comparisons and cross-platform matches. Then the family tree turns the person into a network. AI searches the unfinished parts. The map turns the network into movement. Daily and weekly hints create a reason to return.

Each step moves slightly further away from what the raw DNA file directly contains and closer to what a person hopes the file can explain.

Who am I?

Where did I come from?

Who am I connected to?

Which people are missing?

What happened between them?

What story do all these fragments form when placed on one screen?

The first question can be answered with a report. The later questions can support a platform for years.

I can export my Geni tree and upload it to YourRoots. I probably will, partly because I genuinely want to see what the map finds and partly because I know that I am vulnerable to exactly this type of interface. I like maps. I like incomplete systems. I like the feeling that one more connection could reorganize everything I thought I knew.

But I want to remember what the product is doing while I use it.

My DNA was already finished before I opened Genomelink. The company’s work is to make sure the meaning around it never is.

Connected reading

• The Algorithm Thinks Your Past Is Your Personality
• Every Interface Contains a Business Model
• The Product Is No Longer the Product
• The Interface Is Now Part of the Answer
• The Interface as an Archaeological Site

AI authorship note

This essay began with Kris Haamer asking how much money Genomelink makes, how large the company is, how it started and what its new YourRoots product is. Kris had previously uploaded his Ancestry DNA data to Genomelink and maintains a large family tree on Geni. AI supported company research, product analysis, source verification, drafting, editing and implementation. Kris developed the Raw Drafts angle around a fixed DNA file becoming an endless product and approved the article for publication.