Why Document Everything?

Memory is not an information system. Unaided recall fails in ways that have been measured repeatedly over the last century, and saving information without structure does not repair those failures.

How memory fails

Working memory holds only about four items at once ^[1]. Without active rehearsal, those items are lost within roughly 15 to 30 seconds, displaced mainly by new input rather than by time alone ^[2].

Long-term memory decays as well, along a well-characterized curve. Ebbinghaus' 1885 measurements, replicated under controlled conditions in 2015, show that most of the benefit of a single study session is lost within the first 24 hours, with the rate of loss slowing but continuing over days and weeks. A small residue persists for a month or more without rehearsal ^[3].

Why saving alone is not enough

Retrieval is the point of keeping information, and retrieval fails often. In a naturalistic study of 275 participants accessing 860 of their own shared files, users failed to re-find 13% of files organized in personally created folders, and 22% of files accessed through default or group folders ^[4]. The information was on disk in every case. It could not be produced on demand.

A tab left open, a note typed into a chat, a file called untitled.txt on the desktop: each is a form of saving that reliably fails this re-finding test. A folder of such files is operationally indistinguishable from an empty one.

What documenting adds

Documentation is deliberate externalization: an act that pairs information with enough context for it to be located and understood later. A 2025 meta-analysis of the cognitive-offloading literature confirms that externalizing memory to notes, reminders, and written records consistently improves performance on memory-dependent tasks, with a larger benefit when offloading is required rather than left to choice ^[5]. A complementary review finds the same effect for offloaded intentions: setting an external reminder outperforms unaided prospective memory across the lifespan ^[6].

Paper and digital

Paper works as a medium for documentation, and for centuries it was the only one. The often-repeated 2014 finding that handwritten notes produce better conceptual learning than typed notes has not, however, held up under direct replication. Morehead, Dunlosky and Rawson's 2019 replication, which also ran an internal meta-analysis of similar studies, found only a small and statistically nonsignificant advantage for longhand ^[7]. On the narrower question of medium for encoding, paper has at most a weak edge over typing, and that edge shrinks further when notes are later reviewed.

The case for digital strengthens as soon as the object to be documented is not prose. Paper holds handwriting and simple images; it cannot hold a video, an audio file, source code, a dataset, or long-form text one did not type oneself. Digital records hold all of these. They can also be searched as full text, replicated across devices and cloud services, and migrated to new formats as old ones age. A substantial paper archive also occupies physical space that digital storage does not, and its contents cannot be indexed, linked, or copied without re-digitizing them ^[8].

Those affordances have failure modes: file formats drift, services are discontinued, and a file without a backup is one drive failure from gone. The rest of this series is largely the engineering around those failure modes, which is why digital is the medium it covers.

In practice

Documenting everything does not mean recording every thought. It means treating anything worth revisiting as belonging in a file rather than in memory.