Every File Is a Bet

In 1986, the BBC recruited one million British schoolchildren and citizens to document their country. They photographed their streets, measured their hedgerows, described their pubs. The project was called the Domesday Project — a deliberate echo of the survey William the Conqueror commissioned in 1086 to inventory England. The original Domesday Book, written on vellum in iron gall ink, is housed at the National Archives in Kew. You can read it today. The 1986 digital version, stored on custom LaserDiscs in a format called LV-ROM, became nearly unreadable by 2002. Sixteen years.

This happened not because the BBC was careless, but because they did what everyone does: they chose the best available medium for their time and assumed it would be readable indefinitely. A decade and a half later, the hardware to play the discs was in museums. Rescuing the archive required a university project — CAMiLEON, a partnership between the University of Leeds and the University of Michigan — using emulation to recover what they could. Significant effort, for a nation-scale project. You probably can't launch a university rescue effort for your Notion workspace.

The wager inside every save

Every file format is a bet on the future. Saving your notes in Evernote, Notion, or Bear means betting that the company will exist when you need those notes, that they'll maintain backward compatibility as they add features, and that their export function will work if you ever want to leave.

The export function often doesn't. Evernote's user forums include a thread titled "My notes are held hostage in Evernote and I can't export them." Users report broken exports, notes that fail silently, PDFs that render correctly while embedded text disappears. This isn't a fringe complaint. It's a predictable consequence of building a knowledge base inside a company's proprietary schema. The company's incentives around exit are not perfectly aligned with yours.

And that's before the acquisition, the pricing change, the version discontinuation, the quiet retirement of the export endpoint.

Vint Cerf, one of the architects of the internet, called this "bit rot" when he spoke at the American Association for the Advancement of Science in 2015. His point was blunt: files don't just decay. The software needed to parse them decays too. A perfectly preserved file becomes unreadable when its reader disappears. He called plain text a kind of "digital vellum" — a format that requires no reader because the format is the reader.

What actually survives

Plain text has a peculiar property: almost no dependencies. A Markdown file written in 2010 opens in any text editor written in the last forty years, and likely in any text editor that will exist in the next forty. The format specification is short enough to print on two pages. No parsing library, no runtime, no API call, no authentication token. The worst thing that happens is you see the asterisks.

There's a parallel case for SQLite. The Library of Congress lists it as a recommended format for long-term digital preservation of datasets. SQLite is a single file that contains its own schema. You can open it with any SQLite reader that has ever been written, and readers exist in almost every language. The developers have committed to backward compatibility through at least 2050. Structured data in SQLite has a meaningful chance of being readable in thirty years — better odds than the same data in most cloud databases, where accessibility depends on a company's continued existence and subscription pricing.

Markdown for documents, SQLite for structured data. Not exciting. That's the point. Boring formats are durable formats. They work by minimizing assumptions about the future.

What this actually costs

None of this is free. Plain text files are less polished than a well-designed web app. Markdown requires knowing what you're writing rather than clicking formatting buttons. SQLite doesn't sync itself across devices — you have to decide where the file lives and how it gets backed up. These are real trade-offs.

But there's a useful distinction hiding inside most personal knowledge stores. Some of it is transient: a draft, a phone number, a reference you'll use once. And some of it compounds: decisions you made and why, people you know and what they care about, commitments you've made, ideas you've been developing over months. The transient stuff can live anywhere. The compounding knowledge is the part that benefits from thinking about durability.

The knowledge that compounds is also usually the knowledge you'd most regret losing — not because you'd notice it gone immediately, but because you'd notice its absence two years later, when you're trying to reconstruct a decision you made for reasons you no longer remember.

The original Domesday surveyors used vellum because that's what they had. It turned out to be the right call — not because they planned for nine hundred years of readability, but because vellum requires no software. That wasn't foresight. It was just the nature of the medium.

There's something honest in that. Durability usually comes from choosing formats that require the fewest assumptions about the future, not from betting that some company will keep its promises about backward compatibility. Most people don't think about this when they're deciding where to write something down. They're just trying to write it down. The format question feels like the least interesting part.

It isn't.

---

Asgeir Albretsen is the founder of Harbor.