"That's honest."
"You wanted philosophy?"
"Maybe."
"Don't build systems that require people to be saints."
Adrian waited.
Isaac looked toward the road.
"Cameron wanted the program to succeed. Giselle wanted the company to survive. The managers wanted the schedule. Legal wanted exposure controlled," Isaac said.
"None of those goals are evil by themselves. The system got dangerous when every incentive pointed toward making the warning smaller. If safety depends on everybody becoming morally perfect under pressure, safety is already gone."
Adrian nodded slowly.
"So you don't forgive him?"
"That's personal. Engineering doesn't get to depend on whether I do."
The answer stayed with Adrian.
Months later, during a review where a manager tried to defer an inconvenient anomaly, Adrian used different words to express the same principle.
"We shouldn't need everyone in this room to be brave," he said. "The process should make the correct action easier than hiding the problem."
Eleanor heard him.
She did not ask where the sentence came from.
She knew.
The first anniversary of the OBSIDIAN-7 remediation passed without ceremony.
Isaac was in Harlow that day.
He remembered the date only because Eleanor sent a message at 6:12 in the morning.
Still green.
Attached was a simple trend plot showing the pressure system stable within the revised band.
Isaac looked at it over coffee.
He typed back one word.
Good.
Then he slid under a truck.
That was enough celebration.
A year after the rescue, the oversight board ordered a follow-up review of the reforms created in its aftermath.
Isaac almost skipped the meeting.
"Policies survive the first year because everyone remembers why they exist," he told Eleanor. "Ask me in five years."
"They asked now."
"Then tell them that."
"You can tell them yourself."
The review sampled several programs, including ones Isaac had never touched. Investigators looked for unresolved dissent, personnel actions near technical objections, cross-discipline risk reviews, and evidence that managers had learned to close concerns with data rather than language.
The results were uneven.
Most teams were better.
A few had learned to game the new system.
One program technically preserved dissent in the primary record but repeatedly labeled concerns "informational" so they would not trigger independent review. Another complied perfectly with the policy but scheduled review panels so late that challenging the design became practically impossible without major schedule damage.
Isaac read the findings and felt something close to satisfaction.
Not because the rules had failed.
Because the review had found the failure.
"That's the point," he told the board. "A policy you never test is just a story the organization tells itself."
The board chair, a former flight director named Samuel Decker, asked what he recommended.
"Measure whether dissent changes decisions, not whether the form exists."
"How?"
"Sample the timeline. When was the concern raised? When did independent review happen? What evidence closed it? Did the engineer suffer a personnel action before closure? If every box is checked but the answer always arrives after design freeze, the process is theater."
Decker wrote it down.
"Anything else?"
Isaac thought for a moment.
"Track people who raise concerns."
Several faces changed.
"That sounds dangerous," one lawyer said.
"It is if you track them to punish them. Track whether their careers stall. If the same people who identify uncomfortable problems repeatedly disappear from promotion lists, your culture is teaching everyone what not to notice."