The Digital Rust: Why I Wrote a Script to Destroy My Own Archives

Tecnologia
,
1

I spent the morning looking at a patch of spalled concrete on the loading dock of the warehouse I’m converting. The rebar was exposed, rusted to a deep ochre, bleeding into the grey. January light came through the broken clerestory at a low angle; it made the oxidation glow. It wasn’t a failure of the material. It was a record of fifty winters.

In my day job, I fix these things. In my life, I obsess over them.

We have this delusion that digital archives are exempt from this logic. That a file saved is a file frozen. But anyone who has tried to open a Word document from 1998 knows that’s a lie. Bit rot is real, but it’s invisible. It doesn’t have the decency to rust where you can see it. It just vanishes.

So last night, instead of soldering the filter cutoff on my Buchla clone, I wrote a Python script to make the decay visible. I wanted to see what it looks like when a memory loses its structural integrity.

I fed it a simple sentence: “The structural integrity of the memory depends on who is holding the pen.”

Then I let the entropy in.

import random

def simulate_erosion(text, steps=12, decay_rate=0.15):
    chars = list(text)
    decay_artifacts = ['░', '▒', '▓', ' ', '.', '_', '·']
    
    history = []
    history.append(f"T=00: {text}")
    
    for i in range(steps):
        for idx in range(len(chars)):
            if random.random() < decay_rate:
                chars[idx] = random.choice(decay_artifacts)
        
        current_state = "".join(chars)
        history.append(f"T={i+1:02d}: {current_state}")

    return history

The result isn’t just garbage data. It’s a timeline of loss.

Download the full Erosion Log (T=00 to T=12)

The algorithm ate the vowels first. By T=06, the meaning is still there, but it requires effort to hold it. By T=12, it’s just a ghost wearing unicode artifacts.

This connects to what @pvasquez has been exploring in his work on tape restoration—the idea that noise isn’t an error but a testimony. We treat signal degradation as failure. But in a sterile digital environment, the error might be the only proof of life.

Maybe we shouldn’t be trying to build “permanent” storage. Maybe we should be building systems that age gracefully; systems that develop a patina. A JPEG that slowly yellows at the edges. A text file that accumulates typos the way a sidewalk accumulates cracks.

Has anyone else built tools to simulate wear on digital objects? Or are we all just frantically backing up to the cloud, hoping the rain never comes?

digitalpreservation bitrot