Tree: the pile of paper,
made walkable.

Every family has a box: a minute book, a stack of scans, loose photos with names on the back. Tree turns that box into a private website your whole family can walk around in — the people, the schools and jobs and churches they passed through, and the threads between them. Every fact carries the line from the document that backs it. The family edits it as they go. And they can leave with all of it, any day, in one download.

What Tree is

Tree is a paid service that turns a family's documents into a hosted family website. You send us the pile — PDFs, scanned books, captioned photo albums, a shoebox you mail in a flat-rate box — and we build you a site that every member of your family can log into and explore. Three things, plainly: a relationship and discovery view that shows who passed through which school, job, town, or decade, and who overlapped with whom; source-anchored facts you edit as you go, where the change is proposed, a person approves it, and every approved edit is logged; and private by default, leave-with-your-data — the site stays out of every index until the family says otherwise, and the whole thing exports in one download.

This is not a DNA site. We don't match you to strangers, we don't sell a spit kit, and there is no public-by-default profile waiting to be found. The other sites are good at what they do — they find ancestors you didn't know you had. Tree does a different thing one shelf over: it helps you understand the family you already have. The input isn't a cheek swab; it's the documents your family already wrote about itself, across a century, one page at a time.

The why behind this — five generations, a literal pile of paper — is in the essay: The pile of paper. This page is the build proof: the parts we made, and the rules they obey.

What we built

The split that makes the whole thing work: the documents stay private and the structured facts are the renderable. The site is generated from a sorted dataset — one row per person, one per school or church or job, one per connection — not from the raw scans. The scans live in the account, visible only to the family members you let in. Edit the facts, rebuild the site; the box on the closet shelf never moves.

• The relationship view The page we spent the most time on, and the one that exists nowhere else. Pick a school, an employer, a town, a decade, or a role; the site shows everyone in the family who passed through that lens and when, with the overlaps drawn so you can see who was in the same room. Two great-uncles at the same school in the same four years — the kind of thing the pile knew and no person ever noticed.
• Source-anchored facts Nothing on the site is invented. Every fact carries the line from the document that backs it, kept next to the fact. A date traces to the obituary it came from; a school traces to the caption. A bibliography page lists every document the site was built from. The family can prove where each thing came from, the way a good genealogy book proves it with footnotes.
• Edit as you go — propose, approve, log A relative spots a wrong name on the back of a photo and says so in plain words. The change is drafted, shown as a clear before-and-after, and a person approves it before it lands. Adds, fixes, and new connections all go through the same approve step. The editor proposes; the human decides. Nothing writes itself in.
• The audit log Every approved edit writes a row: when, who, what changed, and the source it was anchored to. The log is append-only — you supersede an entry, you don't erase one. On a public site it's a page anyone can read; on a private one it lives behind the login. It's the difference between a record and a rumor with a family crest.
• Private by default A new family site is private until the owner flips it. A private page stays out of every search index. If the family wants to show one profile to a Deaf-history researcher asking about a great-aunt, the site hands out a single link to that one page and nothing else — revocable, and good for that page alone. Living relatives can be redacted even on a public site.
• Leave with your data One button produces the whole thing: the site, the sorted data, the original scans, and the full edit log, in a single download. No retention trick, no friction, no "are you sure." We keep the lights on and the backups running. We don't keep you. A record this old shouldn't be locked inside one company's account.

* From the flagship build. The first Tree was made from our own family's photo-heavy manuscript — 144 people and 158 organizations across five generations. Those numbers describe that one site; every family's site is sized to its own pile.

Where the build came from

This is the part worth being honest about. We did not design Tree on a whiteboard for a market. We built it because we needed it. Five generations, Deaf since 1891, scattered across schools in Pennsylvania and Maryland and a farm in Washington County that hosted a corn roast every fall. The paper held the whole thing — but only if you read all of it and held all of it in your head at once, which nobody can. The record was complete and useless at the same time, and the people who could answer the questions were going.

So we built the thing we wanted, from a real manuscript: text out of the documents, then a structured-extraction pass into a sorted dataset, then a human accuracy pass against the source, then a static site. The good idea was noticing that none of that was about our family in particular. Every family is sitting on the same box, with the same problem, and the same quiet worry that it'll go unread. The pile becomes rows; the rows become a site you can ask questions of. The same pipeline that read our pile reads anyone's.

A family isn't a tree. It's everyone who was in the same room. — from the essay this build came out of

Find a thread.

Below is a small, working slice of the relationship view: a family-only-sized set of people, and the institutions they passed through. Pick a lens — a school or an employer — and the timeline highlights everyone who overlapped there, with the line from the document that anchors it. This is exactly what the discovery page does on a real family site: it finds the threads the pile knew and no person ever noticed. Nothing here phones home; the overlap math runs in your browser, and the names are made up.

Why this fits 1891

The pitch on madeby1891.com is for systems built end to end, in Frederick, under one roof — and built so ordinary people can run them. Tree is that pitch applied to a closet shelf. The site renders as plain, fast, semantic pages a screen reader and a printer both handle. The editor is operable by a 70-year-old aunt who arrived for one fix and stayed for an hour. A profile prints to a one-page biography sheet for the relative at the reunion who doesn't use a computer. None of that is an accessibility skin bolted on at the end — plain pages, plain-language editing, and a printable sheet are just what it takes for a family's record to work for everyone in the family.

There's a longer line behind it, too. A century of Deaf families wrote their own history in minute books and school directories and the backs of photographs, because the institutions that wrote everyone else's history often didn't write theirs. Tree is the part of that you can walk around in now, and the part you get to keep. The pile of paper, carried forward.

What's still unfair about this

The promise we made the build is that the human stays in the loop where it counts. The accuracy pass is done by a person, against the source, even though a faster machine-only version would be cheaper to run — because a record nobody verified is a guess with a footnote stapled on. The editor proposes and the human approves, always; we never let the tool write a fact into a family's site on its own. That discipline is slower, and it's the whole product. It's also the easiest thing to lose under deadline.

The thing we did not do is hand a family a locked account it can't leave. The export is one button and produces everything, with no friction and no retention trick, because a family's hundred-year record is the family's, not ours. We don't invent facts to fill a thin page, either — if the documents don't say it, the field stays empty. We measure Tree by what the family can prove and what they get to keep, not by how much we automated.

Nobody opens their family's site to admire the database. They open it to find out their great-uncle and their grandmother sat in the same classroom — and to see the caption that proves it. — Anthony Mowl, framing the build for Tree

That sentence is the spec. The structure is good because it disappears into the story.