The original that never was#
Weeks 4-11 treated the surviving evidence as given. It never was — someone spent a budget deciding what would survive. This week makes that decision visible.
An aura is not a file format#
Benjamin (1936): even a perfect copy is missing the original’s “presence in time and space” — its aura.
Read literally against a JPEG, that sounds like nostalgia. Read correctly: a reproduction can carry an object’s content while dropping everything about how it came to exist.
Week 11 showed this exactly — same image data, two dropped facts.
Why source-tier costs money#
The image family’s entire source file: 0.0109 GB — a few cents a year to store.
Auditing, describing and making provenance accessible is a per-item cost. It doesn’t shrink because the file is small.
$120/year across three families cannot buy full audited provenance for everything.
Toy Story 2, 1998#
A routine command deleted ~90% of the film’s working files. The tape backups everyone trusted had been silently failing for a month.
The film survived because Galyn Susman had a personal copy on a home computer — reconciled file-by-file against the salvage.
Two lessons from an accident#
An untested backup is not evidence of preservation — only an unverified claim of it. Same distinction as week 11’s record vs. signature.
A budget states what an archive commits to keep. It says nothing about what survives anyway, by luck, outside the plan.
Two proposals, same $120#
Deep image, thin everything else — full source-tier for the image family. $104.005 total, naive value 38.
Start preserving real audio — source-tier for 4 of 5 audio items, the archive’s first real audio bytes. $118.010 total, naive value 40.
Neither dominates. Image polishes an already-known-incomplete record; audio creates coverage where none exists.
The response refuses to average it away#
A memo has to state which consideration it weights higher — not collapse both questions into one number.
One rate change flips it#
Source-tier labor: $12 → $20/item (a per-item audit, per week 11).
“Deep image” → $120.005 — tips over budget.
“Start preserving real audio” → $105.003 — falls further under it.
The same $8/item change moves them in opposite directions: they spend on families of very different sizes.
Before Wednesday#
Read both texts. Open proposals.csv and sensitivity.csv. Recompute
one row yourself before the workshop.
Workshop: The original that never was
Text walkthrough#
For a reader who cannot use the deck view: Walter Benjamin’s 1936 essay argues a perfect copy still lacks the original’s “presence in time and space,” its aura — read literally against a JPEG that sounds like nostalgia, but Benjamin’s real point is that a reproduction can carry an object’s content while dropping everything about how it came to exist, exactly what week 11 showed when a manifest carried the same image data while dropping two facts an audit could have used. Preserving a family at source tier is expensive not because storing bytes costs money — the entire image family’s source file is 0.0109 GB, a few cents a year — but because auditing and describing each item’s provenance is a per-item cost that does not shrink with file size, so a $120-a-year budget across three families cannot buy full audited provenance for everything. Caleb Stephens’s account of the 1998 near-loss of Toy Story 2, where a routine command deleted roughly 90% of the film’s working files and the assumed tape backups turned out to have been silently failing for a month, shows this tradeoff going wrong by accident: the film survived only because Galyn Susman happened to have kept a personal copy, which teaches two things — an untested backup is only an unverified claim of preservation, the same distinction week 11 drew between a record and a signature, and a budget states what an archive commits to keep, saying nothing about what survives anyway by luck. This week’s pack gives two proposals for the same $120 budget: “Deep image, thin everything else” brings the image family to full source-tier preservation for $104.005 with naive value 38, while “Start preserving real audio” brings four of five audio items to source tier — the archive’s first real audio bytes — for $118.010 with naive value 40; neither dominates, because spending on image polishes an already incomplete record while spending on audio creates coverage where none exists, and the response text refuses to resolve that by averaging the two considerations into one number. When source-tier labor rises from $12 to $20 per item to account for a week-11-style audit, “Deep image” rises to $120.005 and tips over budget while “Start preserving real audio” falls to $105.003 and stays comfortably under it — the same $8-per-item change moves the two proposals in opposite directions because they spend their source-tier dollars on families of very different sizes, and checking that required rerunning the arithmetic rather than re-reading the proposal.