Week 3: Copies & Clones

Duplication vs. derivation, when a copy stops being a copy and becomes its own thing, and why no technical test can tell you the moment it happened

8 March 2027

Open the slides

Last week ended on a gap: two files can pass every identity test at the moment one is copied from the other, and still become different works the instant someone starts editing one of them. This week is about that instant, and why it resists a clean definition.

Duplication is an operation. Derivation is a claim.#

Copying a file is mechanical: read the bytes, write them somewhere else, done. Nothing about the operation itself creates a new work: a copy sitting untouched next to its original is not a derivative of anything, it’s clutter.

Derivation is different. It’s the claim that a copy has acquired its own purpose, its own audience, its own errors: that it is now being maintained on its own terms rather than as a stand-in for the original. Nothing in the filesystem marks this transition. There is no flag that flips from “copy” to “derivative work.” The file itself looks the same the instant before and the instant after someone opens it, changes one number, and sends it to someone who has never seen the original.

Where people actually draw the line#

Ask five people when a copy becomes a new work and you get five thresholds, each defensible:

• Any edit. The moment the bytes differ, they are two things. Clean, checkable, and it classifies a corrected typo as a new work.
• A meaningful edit. Substance, not spelling. Requires someone to rule on what counts as substance, which is the original problem restated.
• Independent circulation. It became its own work when someone sent it to someone without reference to the original. The file may not have changed at all.
• A different audience. The same text, repurposed: a report becoming a board paper. Identical bytes, different work.
• Irreconcilability. It is a separate work once pulling the two back into one would cost more than maintaining both. This one is measurable, and it is measured in effort rather than in content.

None of the five is wrong. None of them is a property of the file, and two of them can be satisfied while the bytes stay identical.

There is a reason no technical test is available here, and it is not that nobody has built one. Derivation is a claim about how a file will be used from now on: who maintains it, who reads it, what it answers for. The bytes record what has happened to a file, and no property of what has happened can encode what is about to. A hash can tell you two files diverged. It cannot tell you that one of them has acquired a separate job.

This matters because the judgement has consequences a hash can’t settle: who owns the result, which version a reader should trust, whether the two files should ever be reconciled into one again, or left alone as separate things that happen to share an ancestor.

Where this goes#

The moment a copy diverges, someone eventually has to decide what to do about the fact that there are now two of them, related but not identical, both still changing. Next week looks at how people handle exactly that, usually without ever using the word for what they’re doing.