The hard drive on the shelf holds a wedding video, three years of family photos, a graduate thesis, and the first app someone in the family ever built. Pull it down in 2026 and a surprising amount of it won't open the way it did when it was saved. Some files throw an error. Others open to the wrong fonts, missing layers, or a soundtrack stripped of its vocals.
The code project won't even start. A dependency it needed in 2015 has moved, renamed itself, or disappeared. The storage held up fine; the contents stopped being readable. That's the digital keepsake problem, and most of what people believe about it is wrong.
Myth: If the File Is There, It Still Works
Storage and readability are two different problems. A JPEG from 2008 can sit on a drive in perfect condition and still render incorrectly in modern viewers, because the color profile it referenced is no longer handled the same way. A Word document from the late 1990s can open in current software and silently lose its formatting, its embedded objects, or its footnotes. The Duke University Libraries preservation guide notes that older word-processing documents often resist opening in modern software, and the fix is usually format migration rather than better backups.
When you look at an old file, separate two questions. Can the bytes be read off the medium? And can the format still be interpreted correctly by software you can actually run? A yes to the first and a no to the second is where most keepsakes die.
Myth: Research Code Is Durable Because It Was Peer Reviewed
A paper is a frozen artifact. The code attached to it is a living dependency graph that keeps moving after publication, and it tends to stop running long before the paper stops being cited. A piece from Neliti on the ten-year decay of research code walks through why: code that only ran on the author's laptop, environments nobody wrote down, dependencies that weren't pinned, repository URLs that go to 404s, and library upgrades that change results without warning.
The habits that keep research software alive for a decade are unglamorous. Lockfiles that pin exact versions. A container image that captures the environment. A tagged release deposited somewhere with a DOI, so the citation points to a specific state of the code and not a moving branch.
Myth: The Cloud Is Taking Care of This for You
Uploading photos to a service is not the same as preserving them. The service decides which formats it stores, which it transcodes, and which it downsamples without telling you. It also decides when to retire a feature, change a sharing model, or shut down entirely. A cloud copy is a convenience layer, not an archive. Keep a local copy of the originals, in a widely supported format, on media you can still read in ten years.
Myth: One Copy on a Good Drive Is Enough
Optical discs and consumer hard drives age worse than their marketing suggested. An arXiv analysis of long-term digital storage found that CD-ROMs sold as reliable for decades were often only usable for a handful of years in practice, and that bit rot, media faults, and hardware obsolescence all compound over time.
A workable minimum for anything you'd miss:
- Three copies. The original plus two backups, so no single failure is catastrophic.
- Two formats of media. An internal drive and a separate external one, so a single technology failing doesn't take everything with it.
- One offsite copy. A drive at a relative's house, or an encrypted cloud copy, so a fire or theft isn't the end.
- A yearly check. Open a sample of files, verify checksums, and replace unreadable drives before the next one goes.
Myth: Any Format Is Fine If It Opens Today
Formats are not interchangeable when the horizon is a decade. Some are documented, standardized, and widely implemented. Others depend on one vendor's goodwill. PDF/A, TIFF, FLAC, and plain text survive software churn better than a proprietary project file that only one application has ever understood.
When you're saving something you want to keep, make a second copy in an archival-grade format alongside the working file. The working copy stays editable. The archival copy stays readable.
Habits That Carry a Decade
Preservation is less a project than a slow routine. A few practices do most of the work:
- Save originals, not just exports. The raw photo, the project file, the source code. You can regenerate an export from the original; you can't go the other way.
- Write down the environment. For code, a lockfile and a container. For creative work, a short note naming the software version and any fonts or plugins the file depends on.
- Pin the link. When something needs to be citable or recoverable, give it a stable identifier instead of a URL that can move.
- Migrate before you have to. When a format starts to feel old, convert a copy forward while software that reads it still exists.
None of this makes the files immortal. It means the version of you ten years from now gets to open them, read them, and recognize what they were, instead of staring at an error and trying to remember what was on that drive.
