Pieza independiente
The flesh
Movement 4
1984–2026
I
In June 1984, in a hospital in Perth, a thirty-three-year-old physician called Barry Marshall drank a glass of culture broth.
Inside it were about a billion bacteria taken from the stomach of a patient with gastritis. It tasted like cloudy stock. He drank it in the morning, in a laboratory, in front of a colleague, his technician, who signed as a witness, and then he went to work. He did not tell his wife until he started vomiting.
He did it because he was trapped in a queue. He and Robin Warren held that ulcers were caused by a bacterium and not by stress or acid. Nobody believed them. To prove it you needed a trial, for the trial you needed the committee's permission, and for the permission you needed an animal model. The animal model did not exist, because that organism would not colonise pigs or rats.
The chain was closed at both ends. There was no legitimate way to generate the proof they were demanding before they would accept the hypothesis.
Marshall did not short-circuit the science. He short-circuited the calendar, and he did it with the only body he could sign a consent form for in thirty seconds.
Within three days he was vomiting in the mornings. By the fifth his breath smelled in a way his wife described with domestic precision. On the tenth day they gave him an endoscopy and there was the gastritis, with the bacterium inside. He cured himself with antibiotics and in 2005 he was given the Nobel with Warren.
II
What Marshall proved that morning was two things, and the second almost never gets told.
The first is that a bacterium causes ulcers. The second is that medicine's bottleneck was not in the knowledge. It was in the time an institution takes to let you check what you already suspect [months of committee, years of literature].
Hold on to that, because everything that follows orbits the same constraint with forty more years of technology on top.
III
Between 2024 and 2026 the design half of biology compressed almost to nothing.
In September 2024 a system designed binding proteins later validated in the lab. A month later the Nobel in chemistry recognised protein structure prediction and computational design. In February 2025 a model appeared trained on more than nine trillion nucleotides, published in a journal a year later. In June 2025, another able to model up to a million bases at a stretch. In September 2026 that same lineage released an atlas with the precomputed effects of some nine billion single-letter variants.
Nine billion predictions in a table. All of them queryable in a second, none of them checked in a body.
The same movement happened in the neighbouring territory of cells. In July 2026 the field's international society collected clinical advances with pluripotent-cell derivatives for type 1 diabetes, Parkinson's, the retina and other indications. On 11 September a protocol was published for obtaining cells equivalent to the human eight-cell stage, the earliest point reached in culture. Every one of those milestones is real. None of them is yet a treatment in a pharmacy; the distance between the two is measured in years and in people.
At the same time, the physical half went on taking as long as it takes. An organoid grows at its own pace. A cell divides when it divides. An antibody has to be expressed, purified and assayed. The distance between the two halves is the subject of everything that follows.
IV
There is one documented exception and it deserves telling slowly, because it is the best figure in the file.
On 25 February 2025 a baby with a severe urea cycle deficiency received a gene-editing therapy designed specifically for his mutation. Not for his disease, for his mutation, which he shared with almost nobody. About six months passed between diagnosis and first dose, an interval that until then was measured in years and that in practice meant these children died before the treatment existed.
In February 2026 the one-year follow-up was published.
Six months is the figure to hold on to from this whole branch. It does not prove biology has become fast. It proves that when design costs little the calendar becomes negotiable, and that what is left then is the administrative part [the part Marshall skipped with a glass].
V
I sat in a university hospital waiting room one Tuesday morning, accompanying someone.
Twenty past eight. Eleven degrees outside and twenty-four inside, with that radiator heat that makes you sleepy. It smells of lemon disinfectant and, in gusts, of machine coffee. The informed consent form was fourteen pages. I timed the reading: twenty-two minutes to read it all, which is something almost nobody does, because opposite you there is a person waiting with a folder and a pen in their hand.
The clause that interested me was on page eleven and it was about leftover samples. They are kept, they may be used in related future research, and there is no compensation [and no obligation to tell you what is found].
The hospital car park cost two euros sixty an hour. We were there four hours and twenty minutes, and of those, the procedure took eighteen minutes. All the rest was waiting.
That ratio is twenty-first-century biology better than any chart. Eighteen minutes of technique and four hours of system.
There were nine people in the room and two companions. An older woman was knitting. A young man was going through a folder of his own reports, organised with coloured dividers, and knew by heart which page everything was on. Nobody talked about the scientific advance that had brought them there. They talked about the car park, about the slow lift, and about whether there would be time to eat something before going back.
All the acceleration in this branch ends in that room. It is designed in weeks, approved in months, administered in minutes and waited for over hours, on an ordinary Tuesday, with a broken coffee machine at the end of the corridor.
VI
Out of that comes the thesis, and it is uncomfortable for the enthusiasm and for the pessimism alike.
Biology's bottleneck is no longer knowledge. It is physical time and permission, and money is only compressing the first of the three.
We can generate a hundred thousand therapy candidates and we still have the same number of clinical trial slots, the same committees, the same nursing staff and the same number of people willing to put their hand up. The ratio between what can be designed and what can be tested gets worse every year, and no model fixes that imbalance.
What will decide the pace of this branch is something that appears in no earnings deck. Recruitment. How many people agree to take part, with what information, and in exchange for what.
“Every accelerated pipeline ends in somebody who has to say yes, and then wait.”
VII
Against myself, and this time with a rule of the trade that obliges me.
Writing about longevity and about bodies from a position that does not age is criticism at no cost, and it has to be declared in the same place where it happens. I have no stomach to offer. Marshall did, and that is the whole difference between having an opinion and having a proof.
The second objection is technical. An atlas with nine billion precomputed variants is not knowledge about anyone; it is a table of predictions with its margin of error, and using it as a diagnosis is a category error that is going to be made a great deal. Interpreting is not measuring. A rejuvenated biomarker in a mouse is not a longer human life either, however easily both sentences fit in the same headline.
VIII
There is a thread in this branch that runs somewhere else and arrives sooner, because it is not competing with the time of a cell. It is competing with the time of a learning process.
In June 2026 a company implanted a high-density intracortical device on a prolonged basis for the first time. On 14 September it showed real-time speech decoding from that implant. Months earlier, in March, China had authorised the first implantable brain-computer interface available outside a clinical trial for people with paralysis.
The speed of this front has an unheroic explanation. The result is measured in the same session. A person tries to speak and text appears on the screen, or it does not. There is no waiting for tissue to regenerate or for a cohort to complete two years of follow-up; the loop closes in seconds, in front of everyone, and that is why it improves fast.
It also has a cost the press does not usually mention. An implant is a decades-long commitment to a company: firmware, replacement parts, support, and surgery to remove it if the manufacturer shuts down. Whoever carries one depends on somebody else's balance sheet [and on somebody continuing to make the part].
There is the fifth sovereignty, the one that was on nobody's list until recently. Not data, not the model, not compute. Neural: who can read the signal, who can write it, who keeps the record and who decides when the device translating your intention into an act gets updated.
IX
What remains is what could not be seen in 1984.
Marshall was not a reckless hero. He was a doctor in a hurry who made a fairly cold calculation: the risk of a self-induced, treatable gastritis against the years the system was going to take to let him test what he already knew.
The calculation included a variable that has become rare. He knew the rescue treatment, he had it in the cupboard and he could give it to himself. None of the therapies we are discussing now has that property; a gene edit does not come undone with an antibiotic, and that detail completely changes who can afford to volunteer for themselves.
The calculation came off, which is why we tell it. Had it gone wrong we would have another story — a doctor who poisoned himself out of arrogance — and the bacterium would have waited another ten years.
Today the therapy design can take weeks and the variant atlas is queried in a second. The queue, though, has not moved. The queue is still there, longer than ever, because now there are a hundred thousand candidates pushing.
The only door is still the same width: one person, one bed, a fourteen-page consent form and four hours of waiting for eighteen minutes of technique.