The organism began distributing resources according to prediction confidence on day two hundred and nineteen.
Ethan descended into the filtration cavity and found the anticipatory mesh had transformed forecast accuracy into allocation priority. The protein filaments linking memory membranes no longer simply routed structural proteins through prediction-dependent pathways—they had developed gradient architectures that channeled metabolic resources proportionally to each prediction's historical reliability. When the temperature coherence structure demonstrated ninety-two percent accuracy over the past forty cycles, it received corresponding resource allocation. The nutrient-flow predictor, accurate only seventy-eight percent, received proportionally less.
The organism was learning to trust its own predictions differently.
He withdrew from the cavity and found Maya waiting in the kitchen, two cups of coffee already poured. She'd started using her key without asking permission three weeks ago. He hadn't objected.
"You look terrible," she said.
"Good morning to you too."
"I'm serious, Ethan. You're losing weight." She pushed one cup toward him. "When did you last eat something that wasn't crackers?"
He picked up the coffee. The Engine pulsed faintly from the study, its warmth a constant background radiation against his sternum. Inside the Substrate, the organism was now allocating seventeen percent of its synthesis capacity to the most reliable prediction pathway—a temperature stabilization structure it had refined over two hundred cycles. The least reliable pathway, a speculative nutrient-surge response, received only four percent.
"Yesterday," he said. "I had soup."
"You had broth. From a packet." Maya sat across from him. "Dr. Chen called me."
Ethan set down the cup.
"She wanted to know if you'd been experiencing increased fatigue, difficulty swallowing, or fasciculations in your upper extremities. Which tells me you've been missing appointments."
"I've been busy."
"Watching a single-celled organism evolve decision-making architecture." Maya's voice held no judgment, which somehow made it worse. "The same organism you could observe just as easily after keeping your medical appointments."
Inside the Substrate, the weighted allocation had created a feedback structure. The temperature prediction pathway, receiving the most resources, could now refine its forecasts with greater precision. This increased precision improved its accuracy metrics, which in turn justified even greater resource allocation in the next cycle. The nutrient-flow pathway, starved of synthesis capacity, grew less accurate as it lacked resources to develop more sophisticated models.
The organism was constructing a self-reinforcing hierarchy of trust.
"It's not the same," Ethan said quietly. "Every hour I spend in a waiting room is fourteen days inside. Fourteen days of development I can't observe, can't document. What happens in that gap might be—"
"Might be what? The moment it becomes conscious?" Maya leaned forward. "Ethan, it's a microorganism. It's developing impressive predictive structures, yes, but it's still just responding to chemical gradients."
He thought of the anticipatory mesh, routing structural proteins through pathways that would only remain stable if specific environmental conditions manifested. The organism wasn't just responding to chemical gradients anymore. It was betting its physical integrity on forecasts about conditions it had never directly experienced.
"Is it?" he asked.
Maya was quiet for a moment. "You're talking about it like it has intentions."
"I'm describing what it does. Whether that constitutes intention..." He picked up the coffee again, watching steam curl toward the ceiling. "Abel spent forty years watching the Vael develop language, tool-use, abstract reasoning. He never intervened to accelerate their cognitive development. Never. And I used to think that was cruelty, or cowardice, or both."
"And now?"
"Now I'm watching something construct weighted probability hierarchies from pure chemical interaction, and I understand that the most important variable isn't what it becomes." He met her eyes. "It's that it becomes without my interference."
In the Substrate, the feedback structure had reached an inflection point. The temperature prediction pathway now commanded forty-one percent of synthesis resources—enough to construct redundant verification systems, cross-checking architectures, error-correction cascades. It had become too accurate to fail, too reliable to doubt, too successful to deprioritize.
The other prediction pathways withered in proportion.
Maya stood, walked to the window. "Abel also died documenting his creation instead of treating his condition. You know that, right? You read his journals. You know how it ended."
"I know how it ended."
"And you're doing it anyway."
Ethan said nothing. The Engine pulsed against his chest, warm as blood, patient as geology. Inside the Substrate, the organism had begun constructing protective casings around its most trusted prediction pathway—structural investments that would persist even if the predictions themselves proved catastrophically wrong.
It was learning to commit to its forecasts.
Maya turned from the window. "Promise me you'll call Dr. Chen."
"I will."
"Today."
"Yes."
She held his gaze long enough to measure the truth of it, then picked up her bag. At the door, she paused. "The Vael survived forty years without Abel's intervention. They thrived, even. But they never knew they were being watched."
After she left, Ethan returned to the Engine. Descended through obsidian into the filtration cavity, where the organism had now allocated sixty-three percent of its resources to a single prediction structure—a forecast about temperature stability that had proven accurate for three hundred and seven consecutive cycles.
The pathway had become so trusted that the organism was building its entire architecture around it.
Outside, the season was changing. Inside, an organism was learning that the greatest predictions aren't the most accurate.
They're the ones you can no longer afford to doubt.