DAY 680
The credential desk had misspelled her name three times. Three attempts, three versions: first TANAKA YIKU, then TANAKE YUKI, and finally, printed with the finality of a system that had moved on, YUKI TANAKE. She clipped it to her lanyard without raising the issue again. The WHO Geneva security office was processing forty-three visiting experts that morning and had reached the point where the spelling of an advisor's name was a problem the schedule could not support.
She walked the A-wing corridor with the badge slightly wrong, past officials in mid-stride and assistant directors carrying stacked folders and two EMA representatives having a rapid conversation she caught enough of—three words in Dutch, two in English—to understand they were debating jurisdictional scope. The corridor had the feel of a place that normally operated on process and was now operating on velocity. Every door open. Everyone moving.
She had been invited as a scientific advisor because she had been among the first to verify the data, which made her the closest thing to an authority on computationally derived therapeutic compounds—a category that had not existed until twenty-eight days ago. This was the logic of crisis: whoever arrives first gets the chair. She had not planned to be an advisor to anyone. She had planned to write the verification paper and return to her lab and let the regulatory bodies handle what regulatory bodies handled. Then Commissioner Blackwell's office had called, and then the WHO Secretariat, and then the EMA, and now she was here with a misspelled badge, and they were waiting for her to explain something she was still, herself, working out. The conference room was at the end of the corridor. Through the glass wall she could see the screens already running.
The São Paulo report had arrived at 06:18 Geneva time, routed through encrypted transfer from the University of São Paulo's Laboratory of Computational Therapeutics—a lab that had existed for three years as a bioinformatics consulting group and had reconstituted itself as a synthesis and clinical program in the twenty days since the data published. Tanaka had read it in her hotel room at 06:45, before breakfast. She read the methodology first, then the results, then the methodology again. Inside, she connected to the conference room's display system and loaded the first slide.
"Day one post-administration," she said. She kept her voice at conference register: factual, mid-speed. "The patient is seventy-one years old, diagnosed with moderate Alzheimer's disease, MMSE score of fourteen at screening, beta-amyloid PET scan positive with bilateral temporal-parietal uptake. Informed consent obtained. The São Paulo team synthesized the compound from the published algorithm output following the synthesis pathway included in the data. Administration was intravenous."
The first slide showed the biomarker charts. CSF beta-amyloid 1-42 levels, tau protein concentration, phospho-tau ratios. On day zero: the numbers that belonged to moderate Alzheimer's disease, the numbers she had seen in a thousand papers. On day three—seventy-two hours post-administration—the numbers had moved. She advanced to the second slide. "CSF measurements at seventy-two hours. Beta-amyloid 1-42 increased from 412 to 631 picograms per milliliter, moving toward the normal range. Tau decreased from 495 to 389 picograms per milliliter. Phospho-tau at 181: decreased from 68 to 47. Significant improvement across all three primary Alzheimer's biomarkers within seventy-two hours of a single administration." She advanced again.
The São Paulo team had included a photograph. Tanaka had not put it in the deck herself—the hospital had included it with the report and she had not removed it when she built the presentation. A woman in a white hospital gown, seated in what appeared to be a consultation room. The photograph was unremarkable: a patient, a chair, fluorescent light. The woman's expression gave away nothing about whether she felt different. Her hands were folded in her lap. A nurse stood slightly out of frame.
The biomarkers gave away everything.
Tanaka said nothing. She waited. The room had its own silence to arrive at, and she let it. Around the conference table: six WHO staff, three EMA liaisons who had come in from Amsterdam that morning, two epidemiologists whose disciplinary connection to this situation was still being worked out. On the main screen, Commissioner Blackwell's feed from Silver Spring showed her making a note by hand.
Forty-seven million people had Alzheimer's disease. The past tense of that sentence remained to be proven. Seventy-two hours was not a cure; seventy-two hours was a biomarker shift and a mechanism confirmed and a woman in São Paulo whose test results were different this morning than they'd been four days ago. But the mechanism was now operating in a human body. The computational prediction had crossed into biology. The silence went long enough that Tanaka counted her own heartbeats—seven, eight—before someone at the table said, quietly, "Questions?" She had said it herself. She hadn't realized. Blackwell was still on the second screen; it was 2:00 in Silver Spring, and she looked like someone who had stopped checking the hour.
"The Emergency Use Authorization framework requires an IND, an established safety profile, manufacturing quality assurance, and a defined patient population," Blackwell said. She was reading from notes. "The compound from São Paulo had none of these. The patient is the first IND. There is no safety profile. Manufacturing QA in this case means 'the synthesizing lab followed published instructions.' And the patient population is—" She paused. Looked directly at the camera. "The patient population is everyone who has Alzheimer's disease."
Dr. Hofmann, the EMA representative, said: "The European position is equivalent. Expanded access under Article 83 of Regulation 726/2004 requires a marketing authorization application to be pending or active. There is no application pending. There is no applicant."
"I understand the frameworks," Tanaka said. "I also understand that this compound has been downloaded from three preprint servers 110,000 times in twenty-eight days. Independent synthesis is confirmed in at least forty countries. The FDA can regulate what American laboratories manufacture. The knowledge is not American."
Blackwell did not disagree. "We can't unapprove information," she said. Flatly. A statement of physics. On the video feed, her Silver Spring office was visible behind her—beige walls, framed documents, the fluorescent lighting of American federal architecture. She looked at her notes and then at the camera, and Tanaka identified in the expression something she'd seen in her own mirror that morning: a professional whose expertise was intact and not sufficient.
She stepped into the A-wing corridor to take a call, twenty meters from the conference room, near a window that looked out onto the Palais des Nations garden. A row of flags she couldn't identify from this angle. The March light was thin and gray.
"I have a patient," David said. He used his careful voice—not cautious, but the register he used when he was carrying something and managing it precisely. She knew this voice from when Mika had fevers at three in the morning and he was walking her through why they should wait rather than medicate. "Seven years old. Pancreatic adenocarcinoma, unresectable, stage IV. Diagnosed six weeks ago. Gemcitabine plus nab-paclitaxel, marginal response." A pause. "Prognosis with continued conventional treatment is five to seven months."
She pressed the phone closer to her ear. Outside, two diplomats crossed the garden path below, coats against the cold. She watched them.
"The parent found the published data," David said. "She's been in contact with a synthesis lab in Cambridge. The lab is prepared to synthesize the compound and the family is willing to proceed. They're asking—" He stopped. "I'm asking."
The pancreatic adenocarcinoma data had published on Day 658. Twenty-two days ago. The synthesis pathway was complete. The molecular mechanism was predicted with 94% confidence by the algorithm. By the standards of computational biology, 94% was unusually high. By the standards of a seven-year-old on stage IV pancreatic cancer, it was a number that lived next to a smaller number. "What's the current clinical trial status for pediatric pancreatic protocols?" she asked. "Phase I. Brazilian cohort, adults only, first dosing scheduled for next month."
She touched the scar on her left index finger, the window glass cold through her sleeve where she'd leaned against the frame without realizing. "I can't tell you to proceed," she said. "The compound has one confirmed administration, in an adult Alzheimer's patient, via a different pathway. Pediatric oncology pharmacokinetics are uncharacterized. I don't have the data to tell you it's safe."
A beat of quiet from his end. She knew what he was doing in that quiet: the same calculation she was doing, from different inputs. "I can't tell you not to proceed, either," she said. "I don't have that data either."
He didn't say anything. She understood why. The hallway filled briefly with three rapid sets of footsteps—more officials—and then quieted again.
"I'll let you know what they decide," David said.
She returned to the conference room forty seconds after the call ended. Blackwell was still on screen. Someone had set water on the table; she took a glass without looking at it and pulled up the slide she had built at 4:30 that morning when the sleep she had needed hadn't come.
"I want to separate what we know from what we're assuming," she said. She moved through the slides at a deliberate pace, because this was the assessment that mattered and speed was not the asset here. "Known: the algorithm's computational predictions have been validated against crystallography data for all five protein targets. The predicted structures match experimentally solved structures to within 2.1 angstroms on average—within the margin of X-ray crystallographic error. This is not an approximation. This is precise structural biology, arrived at through an unconventional method. Known: the São Paulo clinical administration produced significant biomarker improvement in seventy-two hours, consistent with the predicted mechanism of action. One patient is not a trial. One patient is confirmation that the synthesis pathway is accurate and the mechanism operates in vivo."
She advanced to the final section. "Assumed but not verified: that the remaining four compounds will show equivalent clinical efficacy. That long-term effects are manageable. That the synthesis pathways as published are safe across the range of manufacturing environments now attempting to follow them. That pediatric pharmacokinetics match adult predictions. None of these assumptions are unreasonable. None of them have been tested."
She put down the clicker. "My recommendation: parallel tracks. An emergency clinical verification program, coordinated internationally, that sequences through patient populations and documents safety and efficacy across all five compounds. Simultaneously—not sequentially, simultaneously—continued open-source distribution. The knowledge is already distributed. We build the evidence base while the compounds are being used, not as a precondition."
Commissioner Blackwell said: "That inverts the normal process." "Yes," Tanaka said.
A pause. Blackwell looked at something off-camera, then back. "I'll take the parallel track proposal back to the Center for Drug Evaluation and Research. I can't commit to a timeline." Another pause. "That's the most honest thing I can tell you." "I know," Tanaka said.
By evening she was back in the hotel. The room was on the eighth floor, overlooking the Quai du Rhône; she had pulled the curtains when she arrived and not opened them since. Beige walls, a desk she hadn't used, a lamp she'd left on by accident.
She texted David at 21:17: I can't tell you what to recommend. I don't have the data to cover them if it goes wrong. I'm sorry. He replied in twelve minutes: I know. Thank you for being honest.
She checked the verification spreadsheet—a University of Melbourne postdoc had started it and approximately six hundred researchers were now updating it in real time. Three more independent confirmations of the pancreatic adenocarcinoma compound in the past eighteen hours, all adults, all early-stage administrations, all showing the predicted tumor marker response. The arthritis compound was being administered in clinic settings in India and Brazil without regulatory supervision. The São Paulo Alzheimer's trial was enrolling its third patient. The cure data was in use. The regulatory framework was adapting to the use. The use was not waiting. She opened GRIND-7's shared dashboard—he had given her access the week before, a link and a password, no context—and scrolled to the target queue.
Alpha-synuclein pathway, Parkinson's disease: 78.3% computation complete.
Myelin sheath repair pathway, multiple sclerosis: 65.1% computation complete.
78.3%. 65.1%.
She read the numbers twice. Parkinson's disease: ten million patients worldwide. Multiple sclerosis: three million, skewed heavily toward adults under forty. The algorithm hadn't stopped when the cures published. It was still running in Iceland, still solving, still generating conformational data that would produce the next set of synthesis pathways when the percentages reached their thresholds. She had known this—the mining network was still operating, the race was technically still on—but seeing the specific numbers made it concrete in a way that knowing it abstractly hadn't.
She set the phone face-up on the nightstand. The room was quiet. Through the curtains she could see the ambient glow of Geneva's waterfront lights, orange-white through the fabric. She had presented the parallel track recommendation to the panel this afternoon: verify while distributing, build the evidence while the cures moved through the world. She had believed it when she said it. She still believed it. But the panel's verification program would take twelve months minimum to generate meaningful clinical data across all five compounds, and GRIND-7's dashboard was showing percentages that would cross their thresholds in weeks. The verification program would still be in its first phase when the Parkinson's data published. It would still be in its first phase when the MS data followed.
The cascade was not slowing. It was running at the rate of computational throughput in an Icelandic facility that had no reason to slow down, moving toward protein targets she had not yet been asked to advise on, producing data that would propagate at network speed the moment it crossed whatever threshold the algorithm required.
She did not reach for the curtains. She left the lamp on.