The world that is protected

JuneLast time the whole hour came down to one line. Nature was never the ceiling. The ceiling is what moved.

TheoAnd once you have said that out loud, there is only one honest question left. If the worst case is not capped by whatever nature happens to produce, what does a world that can take it look like?

JuneThat is today. This is the last episode of Unit One, and the one that builds the shape of everything after it. By the end you will have four layers, in order, and know which unit of this course is which layer.

TheoWhich is unusually practical for us.

JuneIt is. And it matters on Monday because the day's own exercise asks you to imagine the year 2050, with the world much more pandemic-proof, and to say what technologies, policies and organisations exist, and how we got there.

TheoSo the piece we are mostly on is Richard Williamson's essay from 2022, Pandemic Prevention as Fire-Fighting. And he does something generous at the start. He opens with the success story everybody reaches for, and then takes it away from himself.

JuneWhich success story?

TheoSmallpox. November 1977, the Somalian port city of Merca. An eradication team arrives and vaccinates 54,777 people in two weeks. House to house, and checkpoints at the city edge.

JuneFifty-four thousand, seven hundred and seventy-seven. In a fortnight.

TheoIn a fortnight. And they were there because of one man. Ali Maow Maalin, twenty-three, a hospital cook. He had taken a car ride of no more than fifteen minutes with two children who were infected, after they asked him for directions.

JuneFifteen minutes.

TheoFifteen minutes, and the contact tracers missed it. Ten days later he was sick. In May 1980 the thirty-third World Health Assembly declared smallpox eradicated. And Williamson notes that the last push cost only about one billion dollars in today's currency.

JuneAnd then he deflates his own opening. Smallpox was unusually eradicable. Four things were true of it that are not true of most of what we fear. A human disease with no animal reservoir. Distinctive symptoms. No asymptomatic transmission. And a DNA genome, which mutates far more slowly than many viruses.

TheoHold on. Animal reservoir. What is that in plain words?

JuneA non-human population where the pathogen lives on even after you clear it out of people. Birds, pigs, bats. Smallpox had none, which is the biggest reason it could be finished. Most diseases with pandemic potential have one, so there is always somewhere to come back from.

TheoAnd no asymptomatic transmission means nobody passes it on before they look ill. So you can watch it move.

JuneWhich is why contact tracing worked at all. And Williamson's conclusion is blunt. Most natural diseases of pandemic potential differ from smallpox in one or more of those ways, and that is before you consider engineered ones. So he goes looking for a different model.

TheoAnd lands where?

JuneFire.

TheoFire.

JuneFire. And he earns it with a list. Boston 1760. New York 1776, and again in 1835. Chicago 1871. Boston again in 1872. Baltimore 1904. A dozen more. Every one of them is called the Great Fire of somewhere.

TheoSo this is a hazard that used to shape whole civilisations and now barely registers.

JuneThat is the thesis, and it is the one worth defending all day. Fire stopped killing people in the developed world. And it did not stop because firefighters got braver or faster. It stopped because we rebuilt the buildings. Materials, codes, alarms, insurance, inspection.

TheoAnd you could say the pandemic version of that is mostly unbuilt. That is the sentence this whole episode is really about.

JuneMostly unbuilt. So let us go and find out what it costs to build.

JuneOkay. Let us walk through what fire costs, slowly, because this is the number to carry into the discussion.

TheoStart at the top.

JuneHis estimate is that the total cost of fire today in the United States is in the order of two percent of gross domestic product. In the order of. He is not claiming a decimal point.

TheoAnd two percent of everything the country makes sounds like a lot of fire.

JuneIt is almost no fire. That is the point. Of that total, actual losses from fire are only seventeen percent. Seventeen. And most of that seventeen is not burnt buildings. It is the statistical cost of death and injury, translated into dollars.

TheoTranslated how?

JuneUsing a value of a statistical life, and in this particular estimate that value is 9.6 million dollars. It is the figure an analyst uses to put deaths and injuries on the same page as money. He tells you which figure he used, which is the respectful thing to do.

TheoSo if losses are seventeen percent, the other eighty-three is what?

JuneProtection and prevention. And in 2014 that came to 273 billion dollars.

TheoWait. Two hundred and seventy-three billion dollars a year, to stop fire, in a country where fire has almost stopped?

JuneYes. And here is the line inside the line. Of that 273 billion, only 40 billion is the direct cost of fire departments.

TheoForty, out of two hundred and seventy-three.

JuneFar more of it goes on making buildings and products firesafe. Indirect protection. The things nobody thanks anybody for.

TheoAnd how firm are these?

JuneHe calls them crude high-level estimates, and he shows you one of his own assumptions. The cost of making nonresidential buildings firesafe is assumed to be construction expenditure times twelve percent. That is an assumption, and he puts it on the table rather than burying it.

TheoAnd for scale he reaches for Apollo. The entire lunar programme is estimated at about 280 billion dollars in today's dollars. So the United States spends roughly an Apollo programme on fire protection and prevention, every single year.

JuneEvery year. Not once, for the moon. Every year, for buildings not burning.

TheoAnd his verdict on that is one short sentence. It is worth it, because the cities used to burn.

JuneSo now set that against the thing we are actually here about.

TheoThe deadliest fire in United States history was triggered by the 1906 San Francisco earthquake, and it killed around three thousand people. All civilian fire deaths in the United States between 1977 and 2020 come to approximately two hundred thousand.

JuneAnd COVID-19?

TheoOn deaths inside the United States alone, Williamson's comparison is that COVID-19 is around three hundred and fifty times worse than the worst ever fire, and five times worse than all fires in the last forty-five years combined.

JuneAnd the cost?

TheoEstimated in the order of sixteen trillion dollars in the United States alone. Which he points out is about fifty years' worth of the estimated annual cost of fire.

JuneSo the country pays an Apollo programme a year to hold fire down to almost nothing. And one pandemic cost it fifty of those years, in one go.

TheoAnd this is not a once-ever event. Since 1900 there have been no fewer than five pandemics estimated to have killed a hundred thousand people or more in the United States alone. Three influenza pandemics, COVID-19, and HIV and AIDS.

JuneThen the punchline is not that we underspend. It is the shape of the spending.

TheoThe shape. Of what the United States spends protecting itself from fire, only about 40 billion of the 273 billion goes to the people who turn up while it is burning. The rest is the building.

JuneAnd you could say the honest translation is this. If pandemic defence ever takes the shape fire defence took, then most of the money is not medical.

TheoMost of the money is not medical. Hold on to that one. It will be useful in the room.

JuneRight. So Williamson takes fire and splits the strategies four ways, and this is the bit that organises your whole week.

TheoFour what?

JunePrevention. Detection. Passive protection. Active protection. And his line is that managing fire risk is not all, or even primarily, about more firefighters and more engines, so managing pandemics needs a suite of strategies beyond new pharmaceuticals.

TheoAnd those four are your course. BlueDot calls them Prevent, Detect, Withstand, Combat. Unit Two is prevention. Unit Three is detection. Unit Four is passive protection. Unit Five is active protection.

JuneAnd the course gives it a picture borrowed from medicine. The Swiss cheese model. Every layer has holes, and a threat that gets through a hole in one layer should, in expectation, be stopped by the next.

TheoWhy cheese and not a wall?

JuneBecause a wall implies one perfect barrier, and the course's premise is that any single layer may fail against a determined adversary. So you stack layers that fail in different places. Its own four are delay or entirely prevent the creation of engineered pandemics, detect them early to buy time, withstand the initial shock, and combat and ultimately eradicate.

TheoGot it. Four slices. Now Williamson's version of each.

JuneHe starts with active protection, which is the medicine. Vaccines, antivirals, antibodies. And his way in is extinguishers. Four broad types of fire need different technologies. Ordinary combustibles. Flammable liquids and gases. Metal. Electrical. Chemical foam and carbon dioxide extinguishers arrived in the early twentieth century, and for the first time a flammable-liquid fire could be put out.

TheoAnd biology has classes too.

JuneBiology is less kind to us than chemistry. There are approximately twenty-six families of viruses known to infect humans, and the differences between families matter the way the classes of fire matter. Within a family they share structural features, so knowing how to fight one can help you with the others.

TheoAnd how many of those families do we have an answer for?

JuneFor around half of the virus families that infect humans, we do not have a licensed vaccine for any virus within the family. Not a good vaccine. Any vaccine, for any member of the family.

TheoHalf the extinguisher classes are empty.

JuneWhich is why he says that has to be the starting point, and why he wants a prototype pathogen approach.

TheoPrototype pathogen. Meaning?

JuneYou pick one representative virus from each family and do the groundwork now. Antivirals, monoclonal antibodies, a vaccine. So when the family eventually produces something, you are not starting from nothing. It is stocking the right extinguisher before the fire.

TheoAnd there is a real case for why prior work matters, which is the COVID-19 vaccines. One crucial piece was that our immune system had to meet the coronavirus spike protein in the shape it has before it unlocks the door into our cells.

JuneThe prefusion shape.

TheoThe prefusion spike protein. And it was worked out years earlier, by people designing a vaccine for a different coronavirus, MERS, Middle East respiratory syndrome. Williamson's point is direct. Had that research not been done before the pandemic started, we would have waited longer for effective vaccines.

JuneOkay. Detection. And he does something clever here with alarms.

TheoGo on.

JuneIn the United States, fire alarms cut the time it takes for a fire to be reported by an average of around twenty minutes. Set that against an emergency response measured in single-figure minutes, which is where all the effort goes.

TheoSo everybody is fighting over seconds at the response end, and the alarm is worth twenty minutes at the start end.

JuneAnd the alarm has outcomes to show for it. A 2016 meta-analysis found that smoke alarms halve the risk of fire death. A Canadian study found a seventy-one percent reduction in the risk of a fire spreading beyond the room where it started.

TheoAnd our pandemic alarm?

JuneThe first complete genomic sequence of SARS-CoV-2 was posted on the tenth of January 2020. Wuhan reported the first death the next day. And the virus is thought to have been circulating two to three months earlier.

TheoSo the alarm went off about ten weeks into the fire.

JuneAnd in this kind of fire, ten weeks is not ten weeks. A virus doubling weekly infects a thousand times more people per day after ten weeks. Doubling every two and a half days, as estimated in Hubei in January and February 2020, turns one new infection a day into millions in under two months.

TheoSo why could we not see it?

JuneBecause of what our test is. PCR, polymerase chain reaction, amplifies a sequence, but it needs a known target to copy. That target is called a template. No template, no test.

TheoOh. So targeted testing can never find something new, however cheap it gets. It can only confirm what somebody already thought of.

JuneWhich is why he says it has to be complemented by target-agnostic metagenomic sequencing. Reading every piece of genetic material in a sample, rather than asking the sample one question.

TheoAnd here he is careful with himself. He says such a system is possible now in theory, then says the sequencing would be prohibitively expensive, and that nobody yet knows how to analyse the data it would produce. The Nucleic Acid Observatory at the MIT Media Lab is working on that second problem.

JuneAnd his nearer complaint is that we fail to use what we have. Poliovirus is found in wastewater worldwide using PCR. In New York, they only started looking after the first case of paralysis in an unvaccinated patient.

TheoThe alarm existed. Nobody had switched it on.

JuneNow the layer he is most optimistic about, and it is the one nobody writes headlines about. Passive protection.

TheoPassive meaning what? Protection that needs nobody to do anything?

JuneProtection that is already in place before anything happens. With fire, it is the whole built environment. Materials, codes, enforcement. In the thirteenth century the city of Lübeck restricted the use of wood in construction, after its second devastating fire in a hundred years.

TheoAnd he does not pretend it was easy. After the Great Fire of Chicago in 1871, proposed restrictions on wood in construction were materially watered down, because working-class constituencies did not want the cost of a new home to rise. A real objection from real people.

JuneAnd the other direction has its own cost. He names Grenfell Tower, and the seventy-two lives lost, as where failure to supervise and enforce can end.

TheoSo what is the biological version.

JuneNear-term, two things we already have. HEPA filters, high-efficiency particulate air filters, shown to remove SARS-CoV-2 from indoor air on a hospital ward. And ventilation, which he calls extremely effective against airborne pathogens. His view is that funding trials in high-risk buildings is a no-brainer.

TheoAnd the reason it might get built is that it pays in ordinary years. Hospital-acquired infections are a significant burden with no pandemic anywhere, and the airborne share is estimated in the region of ten percent. Reducing that alone, he says, could be worth billions.

JuneHospital-acquired. There is a term for that.

TheoNosocomial. An infection you catch in a healthcare setting, from being there rather than from what you came in with. If you have ever worked a ward, you know exactly which infections he means.

JuneAnd then the emerging one, far-UVC. Ultraviolet C light is a potent antimicrobial, and the whole question is wavelength. Above two hundred and thirty-five nanometres it penetrates skin far enough to reach the DNA in our cells, and can cause cancer. Below two hundred and thirty-five nanometres it is absorbed by even a small layer of protein.

TheoSo below that line the photons are stopped by the outer layer of dead skin cells, and never reach anything living.

JuneWhile staying effective against microbes in the air. That is why it is interesting, and he calls it the most promising new candidate for a pandemic firewall.

TheoAnd then comes the most responsible paragraph in the essay, a list of what is not known. Effects on materials, cosmetics and surfaces. How it interacts with existing ventilation. Whether it works on larger respiratory droplets. Whether real sources emit above that line as a by-product. Trade-offs between wavelengths below it, since most studies used two hundred and twenty-two nanometres. And edge cases, like an open wound.

JuneSo: most promising candidate, and a list of open questions as long as the pitch. Both, together. That is how to say it on Thursday.

TheoAnd the last piece of passive protection is the one you wear. His verdict on what we have is unsparing. Current protective equipment is uncomfortable, unattractive, impractical, insufficiently reusable and imperfectly protective.

JuneWhich is the moment for the other reading, because it tells you what all of this has to survive.

TheoThe Geneva Paper.

JuneThree pages of it. Securing Civilisation Against Catastrophic Pandemics, by Anjali Gopal, William Bradshaw, Vaishnav Sunil and Kevin Esvelt, from 2023. The course assigns pages ten to twelve, a section called Two scenarios for catastrophic pandemics. And both scenarios end in the same place.

TheoWhich is?

JuneEssential workers. Not hospitals. The people who move food, water, power and law enforcement.

TheoThe first scenario they call Wildfire. One or more sufficiently debilitating and transmissible pandemic agents trigger civilisational collapse by interrupting the distribution of those four things. Food, water, power, law enforcement.

JuneAnd the mechanism starts with what COVID-19 showed, which is that some high- and middle-income countries can contain an agent temporarily by locking down. But a lockdown holds only if essential goods and services keep moving. And the people who move them were infected at particularly high rates during COVID-19.

TheoAnd the authors give numbers instead of an adjective. The vast majority of essential workers have nothing beyond N95 masks. An N95 theoretically blocks ninety-five percent of particles down to 0.3 micrometres. In practice they have been found to block only fifty-seven to eighty-six percent of coronavirus particles worn loosely, and seventy-nine to ninety percent worn correctly.

JuneFifty-seven, if it sits loose.

TheoFifty-seven to eighty-six worn loosely. And then the cascade, stepwise. Lose too many essential workers, to death, to debilitation, or to refusal to work through fear, and some households fail to get the food, water, heating and protection they need to survive. People leaving home in search of supplies spread it further and disrupt services further. As law and order break down, societies collapse.

JuneAnd then there is one passage in those three pages that changes what kind of problem this is.

TheoGo on.

JuneThey say a visibly severe Wildfire scenario, together with the spread of misinformation, could trigger societal collapse even if the actual risk of infection is low.

TheoEven if the risk is low.

JuneEven if case numbers are falling across the population. Because if essential workers judge their own peril to be unacceptably high, for any reason, they will decline to risk carrying an exceptionally deadly virus home to their families. And their conclusion is that perception, at least among essential workers, is just as important as epidemiological reality.

TheoAnd the word they use for that refusal is reasonably. They are not describing panic, and they are not describing irrationality. They are describing a correct decision by an individual that adds up to a collapse.

JuneAnd you have names for this already. You write the Moral Health essays, and two of your own frameworks are strategically overburdened and systems fatigued. Compassion is renewable. Tolerance for a broken system is not.

TheoWhich is exactly the distinction the authors need. The person who does not walk in has not run out of compassion. They have run out of tolerance for a system that sends them in with something that blocks fifty-seven percent when it sits loose. And it turns the usual public health move inside out. If perception is as load-bearing as the epidemiology, reassuring people is the wrong intervention. The right one is protection good enough to deserve the reassurance.

JuneOkay. Scenario two, and this is the one the rest of the course keeps coming back to.

TheoStealth. An initially mild or asymptomatic agent, with a long incubation period, infects most of humankind before its harmful effects become apparent.

JuneIncubation period being the gap between catching it and feeling it.

TheoThe time between infection and the appearance of symptoms. And a long one is exactly what makes this scenario undetectable while it is still containable. They also use latency, which is a pathogen persisting in the body without causing illness, sometimes for decades.

JuneAnd they build this out of real pathogens rather than inventing one. SARS-CoV-2, tuberculosis and HIV killed more people in 2021 than any other pathogens. All three show some minimally symptomatic transmission or latency. But only SARS-CoV-2 spreads rapidly. And only HIV natively has both a long incubation period and a very high infection fatality rate.

TheoInfection fatality rate, as opposed to case fatality ratio.

JuneInfection fatality rate is the share of everybody infected who dies, including those never diagnosed. The honest denominator, and the harder one to know.

TheoSo a pathogen that put those traits together, which they call colloquially a fast-spreading HIV-equivalent, could infect much of humankind before anyone noticed its existence, and then debilitate or kill most essential workers.

JuneAnd here is the specific reason it would almost certainly go undetected today, and it is not a technology gap. Patients with common colds are rarely sequenced. So nobody would have cause to look for adversarial engineering in the genome at all.

TheoWait. So the gap is not that we cannot see it. It is that nobody is looking in the samples where it would show.

JuneThat is the gap. And it is why Unit Three is about reading everything in a sample rather than asking it one question.

TheoAnd their figure for this is a timeline with three marks on it. Four months. Three years. Six years. The first stretch is labelled mild or no symptoms. The later stretch is labelled delayed-onset death or disability.

JuneAnd the thing to notice is not any one number. It is that the whole window where acting is cheap sits inside the stretch where there is no visible evidence of harm.

TheoWhich is why their hardest question is political, and they leave it as a question. If no one is in hospital, no one has died, and only the scientific community is alarmed, will governments take politically costly actions against an agent that might or might not be harmful?

JuneMight or might not. They are not assuming the warning is certain. They are asking a politician to impose visible costs now, on the strength of a prediction that only gets confirmed years later.

TheoAnd a second obstacle holds even if detection worked. SARS-CoV-2 showed how exceptionally difficult it is to contain a pathogen when a substantial fraction of the population does not believe it poses a threat. So people who are persuaded may still be unable to protect themselves from people who are not.

JuneAnd the last part of their section is about despair, which is unusual for a security paper. As more victims began to suffer, people would understand they were infected, and so was almost everyone around them. How they would react, the authors write, is unclear. Perhaps they would come together. Perhaps they would withdraw.

TheoAnd their one intervention there is not technological. They say preserving hope, in the form of newly devised treatments, could be critical to maintaining order.

JuneHope as infrastructure. A strange sentence, and I think they mean it exactly.

JuneOne layer left, and it is Williamson's fourth, and it is the one Unit Two is built on. Prevention.

TheoWhich with fire means what?

JuneHuman behaviour. The number one cause of fire is people. When it is deliberate we call it arson, but mostly it is the innocent, occasionally negligent pursuit of work and play. The leading cause of home fires in the United States is cooking. The main factor is unattended equipment. And more than a quarter of those who died in cooking fires were asleep.

TheoSaid plainly. Those were ordinary evenings.

JuneAnd then comes the argument I think is the sharpest thing in the essay. Because protection and detection are now comprehensive, unsafe behaviour with fire does not imperil a whole city. So we need not restrict deep fat fryers. If a city-wide conflagration were still realistic, our risk tolerance would be far lower.

TheoSo he is saying tolerance for risk should move with how good your defences are.

JuneTolerance for risky research should depend on how good our defences are. And that is the tension to bring to the discussion, because it cuts in both directions. It justifies restricting dangerous research now, while the defences are mostly unbuilt. And it also implies the restrictions should loosen once passive protection is in place.

TheoWhich is a much less comfortable claim than it first sounds. In either direction.

JuneHe is also careful with the words, which is worth copying. He says gain of function is imprecise, because a new function can be innocuous, like making something glow. What matters is dangerous functions, which the National Institutes of Health calls enhanced potential pandemic pathogens, inside a bigger umbrella called dual-use research of concern.

TheoDual-use meaning research that could do harm as well as good.

JuneResearch whose potential for harm comes out of the same work as the benefit. We stay at who decides, and on what standard. Unit Two does the governance properly.

TheoOkay. Last thing, and it is the shape of the field, because the course gives you two long overviews for exactly this.

JuneThe first is a Hoover Institution report from October 2025, Biosecurity Really: A Strategy for Victory, by Drew Endy and colleagues. Its distinctive move is that it refuses to treat biosecurity as damage limitation. It opens with two versions of the same Sunday family dinner.

TheoTwo versions?

JuneIn the first, a teenager has a runny nose and nobody knows what it is. In the second, the son knows which mild rhinovirus strain he has, his mother took a vaccine earlier in the year covering several common cold strains at once, his father's school is warned in advance when viruses are circulating, and modern air treatment stops them spreading in classrooms.

TheoSo their protected world is not heroic at all. A known diagnosis, treated air, advance notice. That is the fire story again. Nothing in that dinner is a firefighter.

JuneAnd their costing is no more than 40 billion dollars annually to the United States federal government, generating up to one trillion dollars annually in direct and indirect public and economic benefits. Both are given as expectations, in their own words, not as calculations you can check from the text.

TheoAnd the comparison they chose is nuclear. The United States is expected to spend ninety-five billion dollars a year to operate, maintain and upgrade its nuclear arsenal over the next ten years.

JuneAnd their sharpest claim is about order rather than content. Nine recommendations in three stages: the next thousand days, then through 2030, then capacities organised for by or before 2040. And the stabilising ones come first, starting with unified cross-agency biosurveillance, because everything else depends on knowing what is happening.

TheoAnd they flag their own failure mode by name. One of their three challenges is understanding changing risks without becoming consumed by unfounded hypotheticals, which they call too many mights and coulds.

JuneThe second is Christian Ruhl's guide for philanthropists, written for Founders Pledge in October 2023. His most usable finding is a measurement rather than an opinion. Between 2001 and 2014, only about twenty-one percent of what counted as United States biodefense funding was strictly biodefense, rather than multi-use.

TheoTwenty-one percent. So the headline numbers in this field routinely overstate what is actually being spent on the problem.

JuneAnd excluding COVID response, the 2023 President's budget requested about eleven billion dollars in discretionary biodefense funding, of which he argues only a small slice is plausibly relevant to worst cases. And excluding direct response and HIV-related funding, total philanthropic funding remains in the tens of millions.

TheoAnd he is honest about his own evidence. Budget line items, he says, do not always mean what they appear to, and his own funding spreadsheet is incomplete and, coming from manual searches, likely biased.

JuneAnd one figure turns up in both readings. Williamson says the American Pandemic Preparedness plan would have authorised 88.2 billion dollars over five years, less than one percent of what COVID-19 cost, and he says that is not nearly enough but would have been a start. Then one sentence, standing alone. Congress did not take up the call.

TheoAnd Ruhl's version of that same figure is that the administration touted a historic 88.2 billion dollar request in the 2023 budget, and he notes that it was requested rather than appropriated.

JuneAnd Williamson ends where he began. James Braidwood wrote the first firefighting textbook in English, and died in the line of duty at a warehouse fire in London in 1861, killed by a collapsing wall. A failure far less likely under modern building regulations.

TheoAnd his conclusion is that we did not solve the Great Fire through more and braver firefighters. The change in the hundred and sixty years since came from innovators, engineers, campaigners, regulators and legislators. And from the will to act.

JuneSo here is the line to walk in with. Of the 273 billion dollars the United States spends protecting itself from fire, only about 40 billion goes to fire departments. If pandemic defence ever takes that shape, most of the money is not medical, and almost none of it is heroic.

TheoAnd that is the end of Unit One. You have the four layers, in order, and the course calls them Prevent, Detect, Withstand, Combat.

JuneNext time you start at the first slice. Prevention. Stopping a pandemic being made at all. Deterring the state that would try, and denying the knowledge and the materials to anybody else.

TheoAnd Williamson leaves you a question, which is a fair one to carry in. How many more biological Great Fires do we need in order to act?

JuneWe will see you in Unit Two.