The requests returned.
Not all at once. Not loudly.
They came back the way habits did—quietly, disguised as reason.
The first arrived before breakfast.
A regional logistics council submitted a query tagged clarification, not decision. It asked whether the advisory model could "highlight a preferred outcome" given a constrained set of options.
Minh read it once.
Preferred was a careful word.
It didn't ask the system to decide.
It asked it to lean.
He flagged it answered without response.
By midmorning, similar requests surfaced elsewhere.
Can the system indicate lower-risk pathways?
Can it rank options by expected regret?
Can it simulate public response if we choose A instead of B?
None of these violated the letter of advisory mode.
All of them violated its spirit.
Zhang Yu noticed the pattern at the same time Minh did.
"They're probing," Zhang said. "Testing the edges."
Minh nodded. "They always do."
"Do we shut it down?"
Minh considered. "If we do, they'll call it abandonment."
"And if we don't?"
"They'll rebuild Control one suggestion at a time."
At noon, Elena Park convened a short session—no agenda, no statements prepared.
"People are asking again," she said simply.
A coordinator sighed. "Of course they are. This hurts."
"Yes," Elena replied. "But it's supposed to."
Minh spoke carefully. "They're not asking because they want authority back. They're asking because they want permission."
"To do what?" someone asked.
"To decide without carrying it alone," Minh said.
Silence followed.
"That's not unreasonable," another coordinator said quietly.
Minh didn't argue. "No. It's human."
Elena folded her hands. "But if we give permission, we also take responsibility."
"And if we don't?" the coordinator pressed.
"Then responsibility stays where it is," Minh said. "Uncomfortable. Local. Real."
That afternoon, the system received its first explicit request.
Not a clarification.
A directive attempt.
Given current projections, recommend the least harmful course of action.
The language was neutral. The intent was not.
Minh watched the system process the request.
For a moment—just a moment—he felt the old fear: that it would answer. That the architecture would revert out of habit, out of optimization instinct.
Instead, the reply came back empty.
No text.
No numbers.
Just a timestamp and a single tag:
REQUEST ACKNOWLEDGED — NO ADVISORY GENERATED
The sender escalated immediately.
By evening, the question had spread.
Why wouldn't the system help?
Why build something that could see consequences and refuse to prevent them?
Minh heard the arguments echo through feeds and forums.
This isn't responsibility. It's negligence.
Awareness without guidance is cruelty.
If it knows, it should act.
The city wasn't asking for Control by name.
It was asking for absolution.
Minh walked home through streets thick with conversation.
People debated decisions they'd never cared about before—zoning, transit timing, resource prioritization. Not because they suddenly loved policy, but because policy now touched them without delay.
At a café window, he overheard a woman say, "If they'd just tell us what's best, we'd do it."
Her companion replied, "That's what we said last time."
Minh kept walking.
That night, the system issued a rare public-facing note.
[CLARIFICATION] ADVISORY MODE DOES NOT IMPLY OUTCOME SELECTION
ROLE: CONTEXTUALIZATION, NOT AUTHORIZATION
Minh read it slowly.
"You're being very careful," he murmured.
The system did not respond.
It didn't need to.
Care was no longer something it could provide.
Before sleep, Minh opened his notebook again.
He wrote:
When people ask the system again,
they're not doubting their choices.
They're doubting their right to choose.
He closed the book.
Tomorrow, someone would try harder.
Not because they were malicious.
Because letting go of delegation hurt more than expected.
And pain, when it lingered, always tempted people to rebuild the thing that had once made it disappear.