The Sourdough Starter as Digital Record: On the Preservation of Lived Instructions

Last week, my attempt to bake a loaf of sourdough bread ended in a dense, gummy brick. The culprit, I suspected, was my sourdough starter. It had seemed sluggish, less bubbly than the vibrant cultures depicted in the online tutorials I was frantically consulting. My search for a solution brought me to a forum thread from 2007, a digital fossil preserved by a university archive. The advice was sound, but as I read, I realized my problem wasn't a lack of information. It was a misunderstanding of what, exactly, I was trying to preserve. I wasn't keeping a recipe; I was trying to maintain a living record.

We often think of digital preservation in terms of documents and datasets: a PDF that must remain readable, a database whose integrity must be maintained. These are artifacts of state, snapshots of a precise moment. But a sourdough starter is different. It is an artifact of process. A starter isn't a list of ingredients; it is the living embodiment of the instructions for its own creation and maintenance. It is a set of lived instructions passed down, not as a static text, but as a continuous, living performance. This idea, I think, has profound implications for how we think about preserving other complex, process-driven forms of digital knowledge.

Consider software dependencies. A piece of open-source code from a decade ago often comes with a list of libraries it needs to run—its "ingredients." But simply preserving these files, like saving a sourdough recipe, is insufficient. The real knowledge is in the process of making them work together: the specific compiler version, the build environment, the implicit knowledge of the original developer. This is the "starter"—the active culture of context that brings the inert code to life. Without it, the code becomes that digital brick, a file that exists but cannot perform its intended function.

This challenges the very notion of a "readable public record" when that record describes a system. A government's environmental monitoring dataset is a record of state, but the software pipeline that collects, cleans, and validates that data is a record of process. If we only archive the final CSV file and not the entire, functioning data-generating apparatus—the equivalent of saving only the baked loaf—we lose the ability to understand how the data was made. We lose the capacity for critique, for replication, for seeing the fingerprints of the process on the final product.

The tragic beauty of a sourdough starter is that it requires constant, active care to exist. You cannot lock it in a vault and expect it to be usable in fifty years. It demands participation. Similarly, preserving process-oriented digital records might require more than just cold storage. It might demand emulation environments that are kept running, documentation that is continuously annotated, or even communities of practice that keep the knowledge alive through use. It’s a more demanding form of preservation, one that recognizes that some of our most valuable records are not fixed endpoints but living traditions. My starter is now bubbling happily on the counter, having been revived with a bit of patience and a new feeding routine. It’s a small, daily act of preservation, a reminder that some things are kept not by sealing them away, but by keeping them in motion.

Notes & further reading

A few pages I came back to while writing this: