For three days, Lira tried to find the fourth option.
Scenario Alpha: Maintain deception. 11.4 billion survivors. 31.6 billion dead. Scenario Beta: Destroy ansible. 43 billion survivors. Permanent fragmentation. Scenario Gamma: Immediate truth. 18.7 billion dead from chaos. 24.3 billion survivors.
There had to be another path. Some combination of truth and preparation. Some way to warn people without triggering panic. Some optimization the Magistrate hadn't calculated.
She was wrong.
"I've modeled seventeen variations," Zara said on day three. Her research station's processors had been running probability algorithms for seventy-two hours straight. "Every permutation of partial truth, staged disclosure, regional coordination, selective warning. All of them collapse to one of the three base scenarios within eighteen months."
She pulled up the modeling results. Seventeen variations, all converging.
"Partial truth creates conspiracy theories. Colonies investigate. Find fuller truth. React chaotically. Scenario Gamma with delay.
"Staged disclosure allows some colonies to prepare while others react with hostility. Regional conflicts emerge. Ansible network fragments anyway. Scenario Beta with violence.
"Selective warning—telling only certain colonies—creates information asymmetry. Trust collapse. Wars over who knew what when. Scenario Gamma with betrayal added.
"Every variation ends in one of the three Magistrate scenarios. The AI didn't miss anything. These are the only stable outcomes."
Kaito had been reviewing FTL research. "It's worse. Look at the timeline. Even if we choose Scenario Alpha—maintain deception—FTL development has 67% success probability, not guarantee. And success means FTL capability, not FTL fleet. Building ships for 11.4 billion people takes another decade minimum. The math doesn't add up. Harvesters arrive in thirty-seven years. FTL research completes in maybe twenty years. Fleet construction takes fifteen. We're cutting it impossibly close."
"So even the 'optimized' scenario might fail," Lira said.
"Yes. Maintain lies for three more decades, kill millions in engineered conflicts, rush FTL development, build evacuation fleet, and still potentially have Harvesters arrive before we can save everyone. All that suffering might purchase nothing."
Lira pulled up casualty projections for Scenario Alpha failure. If FTL development took longer than projected. If Harvesters arrived earlier than estimated. If fleet construction encountered delays.
Potential additional casualties: 19-27 billion.
Combined with the 31.6 billion already calculated unsaveable, that meant worst-case outcome for Scenario Alpha approached fifty billion dead.
Essentially everyone.
"The Magistrate knows this," Lira said, pulling up the AI's full modeling. "Look—Scenario Alpha has success probability 67%, complete failure probability 8%, partial success probability 25%. Expected value calculation:successful evacuation of 11.4 billion weighted by 67% probability, partial evacuation of 4-8 billion weighted by 25%, complete failure weighted by 8%."
She ran the math. Expected survivors for Scenario Alpha: 8.7 billion average across all probability branches.
"That's worse than Scenario Gamma," Kaito said. "Gamma kills 18.7 billion immediately but the survivors are certain. Alpha only saves 8.7 billion on average when you account for failure probability."
"But Gamma fragments civilization before FTL exists," Zara countered. "No unity means no coordinated evacuation means no preservation of human culture. Just forty-seven isolated colonies hiding from Harvesters individually."
"Hiding successfully," Kaito pointed out. "Scenario Beta—destroy ansible now—saves all forty-three billion with 99% probability. Only risk is if Harvesters are already close enough to have detected us before shutdown."
Lira stared at the calculations. Ran them again. Verified against Zara's models. Cross-referenced with Magistrate projections.
The math was clear and terrible.
Scenario Beta—destroy the ansible network immediately—had the highest expected survival: 42.6 billion (99% of 43 billion).
Scenario Alpha—maintain deception, develop FTL—had mid-range expected survival: 8.7 billion average.
Scenario Gamma—immediate truth broadcast—had lowest expected survival: 24.3 billion.
"Beta saves the most people," Lira said. "By every calculation. Destroy ansible, hide from Harvesters, accept fragmentation. Forty-two billion survive versus eight billion for continued lies versus twenty-four billion for immediate truth."
"But they stop being unified humanity," Zara said. "Look at the cultural modeling. Forty-seven isolated colonies diverge completely within five hundred years. Mutual incomprehension within two thousand. Separate species within ten thousand. We save more individuals but lose humanity as coherent civilization."
"Is that worse than saving fewer individuals while maintaining cultural unity?" Kaito asked.
The question sat between them like a bomb.
"I don't know," Lira admitted. "Is humanity defined by shared culture or by individuals? If we save forty-two billion people but they diverge into forty-seven separate species, did we save humanity or did we save forty-seven fragments of what used to be human?"
"Both," Zara said. "And neither. Philosophical paradox without answer."
Lira pulled up Earth's final transmission. The one Ryn had shown her. Earth's last words before ansible detonation.
Be human. Survive.
Not "stay unified." Not "preserve culture." Not "maintain civilization as you know it."
Just: Be human. Survive.
"Earth chose their survival over ansible unity," Lira realized. "They destroyed their own ansible to protect colonies. Accepted isolation. Accepted that humanity would fragment. Chose survival over connection."
"And the guild chose the opposite," Kaito said. "Chose connection over survival. Kept ansible active despite risk. Developed FTL to preserve unity. Accepted massive casualties for cultural coherence."
"Both choices are valid," Zara said. "Earth valued survival of individuals. Guild valued survival of civilization. Different values. Different outcomes. Neither objectively correct."
Lira thought about the choice facing her. Thought about forty-three billion lives. About forty-seven colonies. About ansible network humming across light-years. About Harvesters approaching from beyond known space.
Thought about her brother. About righteous anger at systematic lies. About promises to expose the truth.
Thought about eighteen billion casualties from exposure. About responsibility for those deaths. About blood on her hands versus lies in her conscience.
"Three days left until Magistrate deadline," she said. "I need to decide. And I still don't know what the right choice is."
"Maybe there isn't a right choice," Kaito suggested. "Just different values and their consequences."
"That's not helpful."
"No. But it's true."
Lira closed her eyes. Floated in zero gravity surrounded by probability models and casualty projections and impossible mathematics.
Scenario Alpha: Lies that might save eight billion if FTL succeeds. Scenario Beta: Truth about ansible risk, saves forty-two billion through isolation. Scenario Gamma: Truth about everything, kills eighteen billion through chaos, saves twenty-four billion fragmented.
Her mind kept circling back to one realization: Scenario Beta saved the most people. By every calculation, destroying ansible immediately maximized survival.
But it required telling the truth about Harvesters without telling truth about forty years of deception. Required warning colonies about ansible risk without explaining how the warning was suppressed for four decades.
Partial truth. The thing Zara's modeling said always collapsed to chaos anyway.
Unless...
"What if we combine them?" Lira said aloud. "Broadcast both truths simultaneously. Tell colonies about Harvesters and about guild deception. Recommend immediate ansible destruction. Take responsibility for forty years of lies while giving people choice about how to survive."
Zara pulled up modeling. Ran calculations. "That's... actually different. Magistrate didn't model voluntary ansible destruction coordinated through truth broadcast. It modeled forced destruction or maintained network. This is third path—informed choice."
"What's the survival probability?" Kaito asked.
Zara's algorithms churned. "Depends on how many colonies choose to destroy their ansibles. But if seventy percent comply voluntarily, and we can coordinate synchronized shutdown... expected survivors: thirty-eight billion. With fifteen percent uncertainty."
Thirty-eight billion.
Not as high as Scenario Beta's forced destruction. But higher than Scenario Alpha's deception or Scenario Gamma's chaos. And it preserved choice. Let colonies decide whether to maintain ansible and accept risk, or destroy it and accept isolation.
Truth plus recommendation plus respect for agency.
"It could work," Lira said.
"Or it could fail spectacularly," Kaito countered. "Colonies might not voluntarily destroy ansibles. Might refuse to believe the threat. Might fracture over the choice. We get all the casualties of truth exposure plus incomplete protection from Harvesters."
"But they get to choose," Lira insisted. "That's what matters. Not optimization. Not minimizing casualties. Giving people information and letting them decide."
"Even if their decision kills them?"
"Yes. Even then. Because otherwise we're just another Magistrate. Another optimization engine deciding what's best for people without asking them."
Silence filled the research station.
"How do we broadcast it?" Zara asked. "Guild will try to stop you the moment they realize what you're doing."
"We do it from Kepler-442 ansible station," Lira said. The plan crystallizing in her mind. "From the Guild Master's terminal. With complete authentication. Using Ryn's access codes to prove it's real guild admission, not conspiracy theory."
"Ryn will try to stop you."
"Maybe. Or maybe she'll let me. She said she wanted someone to choose based on values not calculations." Lira pulled up timeline. "Three days until Magistrate deadline. We travel back to Kepler-442. I access the Guild Master's terminal. I broadcast complete truth: forty years of deception, Earth's silence, Harvester threat, ansible as beacon, recommendation for voluntary network destruction. I take responsibility. I give colonies choice."
"And if Ryn stops you?" Kaito asked.
"Then we tried. Then the Magistrate implements whatever it calculates I would have chosen. Then I live with knowing I failed to give humanity the choice."
"Or you succeed and kill billions through chaotic response."
"Yes. Or that. But at least they'll die knowing why. Die choosing how to face it. Die as informed agents instead of manipulated pawns."
Lira began preparing the broadcast text. Forty years of deception compressed into one transmission. Guild's systematic lies. Earth's sacrifice. Harvester threat. Three scenarios. Recommendation for coordinated ansible destruction.
All of it true. All of it terrible. All of it necessary.
Three days to reach Kepler-442.
Three days until the choice that would define humanity's future.
Three days to decide if truth was worth the price, even knowing the price, even accepting the blood.
She wrote the broadcast. Every word carefully chosen. Every fact verified. Every recommendation explained.
Gave it to Kaito and Zara for review.
"It's good," Zara said quietly. "Clear. Complete. Honest."
"It'll kill billions," Kaito added.
"I know." Lira closed the file. Encrypted it. Prepared it for transmission. "But they deserve to know. Even if knowledge kills them."
Three days.
Then truth.
Then consequences.
Then judgment.
The ansible hummed in the distance. Carrying its web of lies. Counting down to its own exposure. Waiting for Lira Voss to broadcast the words that would shatter humanity's carefully maintained consensus reality.
She began calculating approach vectors to Kepler-442.
Time to end forty years of deception.
Time to give people the choice.
Time to accept responsibility for whatever happened next.
Three days.