DAY 730
The data filled the conference room's main display in six-point tables, smaller than Tanaka would have preferred but larger than anything she'd projected for this audience forty-eight hours ago. She'd added the color coding after Commissioner Blackwell's aide had requested slides rather than raw printouts—disease categories in blue, confidence intervals in grey, p-values small enough to require the table's LED lighting to read comfortably. She didn't need the slides to remember the numbers—one hundred and nine days since publication, and they had not changed.
Alzheimer's disease. Four thousand two hundred patients across thirty-one countries, thirty-seven independent research sites. Biomarker improvement—amyloid-beta clearance, tau phosphorylation reduction, cognitive assessment scores—in 98.7% of cases. The outlier 1.3% traced to administration protocol differences at two sites in South Korea and one in Brazil; when those sites adjusted their synthesis procedures, the response rate corrected to 99.1%. The cure worked. The data said so in six languages across the slide deck.
Rheumatoid arthritis—the joint inflammation that had consumed TNF-alpha inhibitors at sixty billion dollars annually. Twelve thousand patients. Ninety-seven point two percent efficacy on primary endpoints: reduction in joint inflammation markers, DAS28 scoring improvement, reduced radiographic joint damage at three-month follow-up. The speed of resolution had surprised even Tanaka, who had run the molecular simulations. The protein pathway correction was that clean.
The cancer cures were slower to compile because the disease categories were heterogeneous and the patient populations spread across more sites. Two thousand eight hundred patients total: eight hundred forty with pancreatic adenocarcinoma, six hundred with non-small cell lung cancer, eleven hundred sixty with triple-negative breast cancer. Response rates ranging from 94.3% for the pancreatic subgroup to 99.1% for triple-negative. These were not comparable populations; the cancers were mechanistically distinct and the treatments distinct. The efficacy rates clustering in the upper nineties was a function of the algorithm's approach—targeting conformational vulnerabilities in each cancer's specific protein folding cascade.
The panel had been quiet through the presentation. Fifteen officials around the table: FDA, EMA, WHO, the Australian TGA, Health Canada. Blackwell was on the end screen from Silver Spring, her yellow legal pad visible at the edge of the frame. No one had interrupted for thirty-five minutes.
Tanaka clicked to the final slide. "The gap between computationally predicted and clinically confirmed has closed," she said. "These are no longer projections."
Blackwell was the first to speak. She had been writing since the third slide, a habit Tanaka had noticed across their four calls over the past month—the pad as a tool for processing, not note-taking.
"I want to propose a framework," Blackwell said. She paused the way someone pauses when they've spent a night drafting something and are now committing to saying it. "Computationally Derived Therapeutic Validation. A provisional designation—not formal approval in the NDA sense—acknowledging that these compounds have been validated through computational proof and retrospective clinical evidence. A separate pathway. A new pathway."
The EMA representative, Dr. Adriaan van Hooijdonk, set down his pen. "The problem with a new pathway is that a new pathway implies the old pathway's requirements still apply to subsequent cases. It implies we're creating an exception, not establishing a new standard. The moment we call this 'computationally derived,' we're encoding that computational derivation is extraordinary rather than addressing what the actual validation threshold should be."
"The alternative," Blackwell said, "is to acknowledge that cures currently in active clinical use worldwide require nothing from us." The room held the implication. The Australian delegate, a compact woman named Prescott who had said nothing for forty minutes, spoke without looking up. "They don't, though. Require anything from us. That's the problem we're actually discussing." She was not being provocative. She was being accurate.
Tanaka watched the panel work through the logic. Blackwell's framework was rational—a category fitted to the evidence, a regulatory structure where none had existed, cover for institutions that needed to respond. Van Hooijdonk's objection was also rational: naming the exception entrenched the old rule. Both were correct, and neither changed anything happening in the hospitals where the cures were already being administered. While this panel constructed language for what had occurred, a clinic in Nairobi was synthesizing the Alzheimer's protocol. A hospital in Seoul had begun retrospective review of its rheumatoid arthritis patient cohort. A physician in São Paulo—the same lab that had first administered the Alzheimer's cure eighty days ago—was treating his fourth pancreatic patient this week. The panel was not behind the evidence; they were behind the world, and the morning session broke at 10:47 for a twenty-minute recess.
Tanaka walked out into the WHO corridor. The building ran institutional—linoleum, fluorescent overhead strips, the smell of coffee from the vending alcove near the elevator bank. Her phone buzzed twice as she cleared the conference room door. The message was from GRIND-7. He sent updates on a schedule corresponding to algorithmic milestones, not any clock, and she'd learned to check her phone when his timestamps clustered. She read it standing in the corridor.
Parkinson's disease: 100%. Synthesis protocol available at current repository. Estimated clinical availability: 2-4 weeks at existing manufacturing sites.
Current target queue: Type 1 diabetes: 82% Multiple sclerosis: 79% Systemic lupus erythematosus: 71% Amyotrophic lateral sclerosis: 58%
She scrolled the list twice, holding the Parkinson's figure: six major diseases now, ten million patients worldwide, and the queue percentages already moving since last week—Type 1 diabetes up from 76%, multiple sclerosis up from 71%. The algorithm was still accelerating. It had been running for 730 days on top of 621 days of prior computation, and whatever ceiling its protein folding accuracy was approaching, it had not reached it. The regulatory panel in the next room was discussing the framework for approving the first five cures. By her estimate—and she'd checked the progression rates three times this week—they would need frameworks for eleven within a year.
She put the phone in her pocket and got coffee from the alcove machine. It was bad coffee, the kind from a machine that offered seven varieties and delivered one flavor regardless of selection. She took it back toward the conference room with both hands around the cup, stopped at the corridor junction, and told the aide she'd need the small room off the main corridor for the next hour—twenty minutes to review her notes before the afternoon session. The room had a table, four chairs, and a window facing the building opposite. She sat with her notes in front of her and didn't look at them.
David had sent the email at 6:12 AM Geneva time. She'd seen it before the morning session and flagged it to open later, because she knew what it was—he'd told her he was sending a video—and she'd needed to present the data first, without being inside whatever the video would do to her processing. Now she opened it.
The video was forty-two seconds. Hospital corridor. Linoleum, the same blue-grey as every hospital corridor, lit by overhead fluorescents. In the shot, slightly out of focus at first, a child ran from one end to the other. Small feet in grey socks. Blue hospital gown flapping behind. The child was laughing—that particular seven-year-old laughter that has no target and no joke behind it, only an excess of physical energy and a corridor long enough to sprint. The child stopped at the far wall, turned around, and ran back.
She watched the video once. She watched it again. The third time, she watched the child's feet—the way a child's body runs before anything teaches it form, the heel strike, the arm pump, the pure mechanical fact of legs that work and a space long enough to use them in.
Fifty days ago, this child had a primary pancreatic adenocarcinoma and an expected survival timeline that David had declined to specify in numbers when they'd spoken on Day 680. Specificity would have required him to say something he was not willing to say. The child in the video was laughing in a hospital because the hospital was familiar and because the corridor was long and because someone had said it was okay to run.
She thought about the regulatory panel in the next room, constructing the language of Computationally Derived Therapeutic Validation. She thought about GRIND-7 in Iceland, whose update sat in her pocket—six cures complete, more behind them, the algorithm still running at full throughput. She thought about the parent who had not waited for regulatory approval, who had found the synthesis protocol and found a lab and found the compound and found a physician willing to administer it, because their child had not had time for the other answer. She ran her thumb across the scar on her left index finger—a kitchen accident when she was eight, nothing significant—and put the phone in her bag without watching the video a fourth time.
The afternoon session began at 14:00. The panel had spent the recess circulating van Hooijdonk's objections in a document that had somehow produced two alternative frameworks and a minority opinion by 13:55, and Tanaka was given the floor.
She stood at the head of the table and looked at the fifteen faces—some attentive, some tired, all careful in the way people are careful when they know what they say will be cited.
"The process was wrong," she said. "No controlled clinical trials. No randomized cohorts. No placebo comparison. No regulatory submission. No informed consent documentation in any standard format. Every procedural safeguard we have built around the introduction of new therapeutics was bypassed, in sequence, by a cryptocurrency mining algorithm in Iceland." She let that sit. "And the clinical data says the same thing the treating physicians have been saying for three months. Of the compounds with full clinical verification—five diseases. A sixth complete as of this morning—Parkinson's disease. The next disease in the queue is Type 1 diabetes, currently eighty-two percent of the way to a complete synthesis protocol." She touched the edge of the conference table with two fingers; the room was still.
"What I'm recommending to this panel—and I want to be clear this is a recommendation, not a clinical directive—is that we stop debating how to approve the first five cures, because the first five cures do not need our approval. What they need is distribution infrastructure, manufacturing standards, and post-market surveillance frameworks that function at the scale of global public health. What they need is for us to be ready when the sixth cure enters clinical use in four weeks. And the seventh. And the eighth." She looked at Blackwell's face on the end screen. "The computation is still running. More protocols are coming. The question isn't whether we have the right framework for what has already happened. The question is whether we have anything ready for what is still coming."
Van Hooijdonk's pen was still. Commissioner Blackwell wrote something on her yellow legal pad and did not look up. The session ended at 18:30, and by 19:15 Tanaka was in her hotel room, sitting on the edge of the bed with the lights off.
The Geneva hotel room had the same window-facing-building-opposite arrangement as every hotel room she'd occupied for the past six months, which was fine. She watched the lit windows across the gap and didn't think about anything particular, then watched David's video again—the corridor, the small feet, the laughter, forty-two seconds—and called Mika.
The call connected on the second ring. Mika's face on the screen: seven years old, David's dark eyes, Yuki's jaw, her own expression worn with no awareness that it was an expression. She'd lost another tooth since the last call. There was a gap visible when she opened her mouth immediately, something that couldn't wait.
"Mama, Nori caught a moth. We let it go but then it came back. Papa says moths do that sometimes."
"Moths do do that sometimes," Tanaka said.
They talked for twenty-two minutes. The cat—Nori, a grey tabby with a chronic interest in insects—had also knocked a glass off the counter, which Mika described in more detail than the damage warranted. There was a school project involving a papier-mâché volcano. Mika's friend Priya had a dog now. None of it touched the conference room, the clinical data, the child running in the hospital corridor, the Parkinson's cure sitting complete in the algorithmic queue. Mika talked about the moth and the volcano and the dog, and Tanaka listened and asked follow-up questions, and Mika's face was seven years old and lit by the screen and entirely indifferent to the words "amyloid" or "therapeutic validation" or "synthesis protocol."
After the call ended she sat in the dark. In Iceland, in the server racks, the algorithm was processing the protein folding data for Type 1 diabetes—82% complete, closing. Parkinson's protocols already in the repository. Multiple sclerosis at 79%. The diseases had names and patient counts, and the algorithm was diminishing them one computational solution at a time, and the world was not ready and would not be ready and was building readiness out of the materials it had, which were never enough and were always what there was.
She was grateful and she was afraid. Both, without one softening the other, the same way the clinical data was both confirmation and the beginning of a longer question about what came after confirmation.
She did not try to separate them.