The Stress-Test Table in Practice: The Archive She Almost Sold Twice
The first offer arrived in Sofia’s inbox in January, when the grant she’d counted on had just fallen through. An AI company wanted to license her archive — twenty years of documentary photographs of Portugal’s coastal fishing communities, from Sesimbra to Viana do Castelo — as training data. The number attached would have funded a year of fieldwork. She read the contract twice, hovered over the reply, and asked for a week.
She almost signed. What stopped her was not certainty but its absence — and Chapter 6 of the book has a tool built precisely for the moment when immediate rewards lean on long-held principles: the Stress-Test Table. Four questions, answered in writing for each value at stake, because — as the book puts it — writing makes abstract trade-offs concrete and harder to rationalize.
Her first table, for the value she could name fastest:
Value
The trust of the people in the pictures. The fishermen and their families let me in for twenty years on an unwritten promise: your faces serve your story. Not “your faces serve whatever a model learns to generate.”
Temptation
A funded year. And honestly — because the tool demands honestly — relief. No grant applications for twelve months. The money is real and my rent is real.
Cost of Integrity
Also real, so face it: the archive goes on earning nothing. I keep scrambling. And the bitter footnote — models will train on somebody’s fishing villages regardless; my refusal protects a principle, not the world.
Payoff of Fidelity
I remain the photographer who can walk into any house in those villages and be poured a coffee. The archive keeps its standing as witness — worth more over decades than any license. And I keep the version of myself that made the promise.
Then, because the tool insists on testing every value under pressure and not only the comfortable ones, she made a second table for a value that argued with the first: sustaining the work itself. Its temptation column was the mirror of the other’s cost column, and reading the two side by side she could see the actual shape of her dilemma — not principle versus greed, which is how it had felt at two in the morning, but promise versus endurance, which is harder and more honest.
She declined in February. In May the company came back with a higher number — hence almost twice — and this time the decision took an afternoon instead of a week, because the tables already existed. She reread them, changed nothing, and replied. What the second offer did accomplish was to push her toward the door the tables had quietly pointed at: if the archive had that much value to a machine, it had at least that much to institutions that keep promises for a living. A maritime museum in Lisbon is now negotiating for a curated acquisition — with the subjects named, credited, and consulted, which is to say, on the archive’s own terms.
The book advises keeping completed tables and rereading them, because patterns in how you navigate pressure are worth more than any single decision. Sofia’s file of tables is four entries long now. The pattern she reports: the temptation column is always shortest and always written last — and the honesty of that column is what decides whether the whole table works.
Sofia is a composite — drawn from conversations with photographers, writers, and artists fielding AI licensing offers for their life’s work, with details changed. The two-in-the-morning arithmetic is general.
Try it yourself
The next time an immediate reward leans on a principle, write the four rows — Value (named specifically), Temptation (honestly, including the relief), Cost of Integrity (the real one, not the noble one), Payoff of Fidelity (long-term, concretely) — for each value at stake, including the ones that argue with each other. Keep the finished tables. The second time pressure comes, you’ll already know your own handwriting.
Comments