The detail doing the most work is in your footnote: as far as you could find, exactly one factory did this. That makes Braskem less a model than a measure of the gap. What protected the N95 supply was one creative idea, eighty willing people, and an employer who happened to pay them like the work mattered. None of it was designed in, and almost none of it was reproduced.
I'd resist reading this as "the market and a few heroes had it covered," but also the opposite reflex that what we lacked was a better central plan. The US did have pandemic plans. The thing that stayed with me from The Premonition is how little the formal apparatus mattered when it counted, and how much came from people at the margins of the system who saw further than the institutions did. Your "authorities and amateurs" point is that same tension, and Braskem is a data point for it, not against it.
So the question I'd sit with isn't "wasn't this heroic" or "why didn't we plan for it," but what institutional arrangement makes on-site continuity and supply resilience a default rather than a fluke, without designing out the thing that actually worked here: competent people improvising, trusted and paid to. The ALLFED project @jamesmulhall🔹 mentions seems exactly the right shape, turning a lucky one-off into something repeatable, compensation and trust included. COVID was a fairly forgiving test of all this; worth building the system while that's still true.
Thanks for this. I want to defend the principle in your section 2 and push two threads further: a near-term funding case that has shifted in your favour since 29 May, and the longer-term response-capacity case in your sections 5-6, which I think is where the real opportunity sits. From an impartial view it doesn't change things that this is unlikely to go pandemic, or that the suffering is in central Africa rather than closer to home; it counts the same.
1. The near-term case is stronger than when you wrote it
The "it's already funded" objection is being falsified in real time. On 5 June, WHO and Africa CDC launched a consolidated six-month plan asking for $518m, not ~$350m, and pledges are soft and falling: Africa CDC put them at ~$290m in late May, down from ~$500m a week earlier, with the headline figure still bundling in-kind and repurposed budgets. The structural reason matters for an EA audience: the US withdrawal hit this exact response (health-worker layoffs, supply shortages) and no other donor has filled the gap it left. The "vivid disaster, then money floods in, so low room for more funding" dynamic that made GiveWell cautious in 2014 is not operating this cycle.
That is the sharp point. GiveWell judged even the multi-billion-dollar 2014 response plausibly as cost-effective as its top charities, yet still declined to recommend marginal donations, because governments were flooding in. That single reason is exactly what is failing now. Honest limits: this is no longer an "early" intervention in your section-2 sense, so the case is stronger on room-for-more-funding, not on timing; and whether it matches 2014's cost-effectiveness depends on how large the averted caseload turns out to be, which we cannot yet know. With R still above 1, every week of delay keeps multiplying that caseload. The toll is also worse than a DALY count for the outbreak alone suggests: Ebola disproportionately kills health workers, and each loss thins an already-thin system for years, well beyond Ebola.
Where a marginal donation can and cannot go: no individual can move a shortfall in the hundreds of millions sitting at the sovereign and multilateral level (WHO, Africa CDC, national plans). MSF, the most visible operator, is ~98% privately funded and largely independent of that shortfall, so I doubt a marginal MSF gift is the binding input either, though I cannot show its outbreak-specific position. But the burial-and-community-engagement layer, the most agile and scalable bit and the slice the US defunded, is retail-reachable: IFRC's CHF 29m regional appeal funds exactly this, with a dedicated donation page and ~200 DRC Red Cross volunteers already working in Bunia and Rwampara. The honest gap is not the route; it is that no one has evaluated it GiveWell-style. Which is the bridge to the bigger point.
2. The longer-term case: the response-capacity gap EA keeps re-noticing
Your sections 5-6 are the more important half, and I would push them past "should we build something?", because most of it already exists. That is the more useful answer to your "what's already there so we don't reinvent it" question. The counter-cyclical fund meant to disburse before the appeals are written is the WHO Contingency Fund for Emergencies, and the funding collapse this cycle is partly the story of that fund being starved: it has released only ~$3.9m into this outbreak so far. The layer that places skilled responders into a live outbreak is GOARN, which already deploys epidemiologists, surveillance, data and risk-communication experts within 48 hours. So the gap is not building those two; it is that they are chronically underfunded and disconnected from EA money and people.
What genuinely doesn't exist is the third piece: independent, EA-style evaluation of whether a marginal dollar does most good through the CFE or through an operational appeal like IFRC's. That sits squarely in this movement's comparative advantage. The work is about money, not bodies, and it starts with that evaluation: do it in advance rather than mid-crisis, and money can flow to whatever clears the bar. The failure mode that should worry us is the reverse: channels that would clear the bar never get funded, because no one has done the work to show that they do. That, not new infrastructure, is what "a vetted route from day one" actually takes.
On the x-risk link you raise: I would not claim this outbreak is existentially relevant in scale. Its value is as a drill. Engineered-pandemic preparedness suffers from tiny probabilities and absent feedback, and the most recent real stress test was sobering: COVID was neither the Big One nor engineered, yet wealthy governments still failed on testing capacity, surge coordination and public trust. A live, p=1 outbreak is the cheapest available place to build and test exactly those muscles, plus surge logistics and diagnostics for a strain the standard field tests miss. The transmission dynamics differ, so the clinical specifics don't all transfer, but the operational and trust layer that broke in COVID is the layer a real outbreak exercises. Building response capability this way is plausibly the most tractable thing we can do for the engineered-biorisk class outside life-science R&D, and it pays off even if the tail risk never arrives.
Net upshot: if you're giving now, the IFRC appeal looks like the cleanest retail route; the larger and more durable prize is standing evaluation of these channels, so the next outbreak starts with a vetted route instead of this scramble. Is anyone already doing that, formally or informally? If not, it seems worth a few people scoping, which is the question your section 6 is really asking.
(Disclosure, in the same spirit as the OP: I used Claude for research synthesis and a first draft; the framing and judgements are my own.)
The detail doing the most work is in your footnote: as far as you could find, exactly one factory did this. That makes Braskem less a model than a measure of the gap. What protected the N95 supply was one creative idea, eighty willing people, and an employer who happened to pay them like the work mattered. None of it was designed in, and almost none of it was reproduced.
I'd resist reading this as "the market and a few heroes had it covered," but also the opposite reflex that what we lacked was a better central plan. The US did have pandemic plans. The thing that stayed with me from The Premonition is how little the formal apparatus mattered when it counted, and how much came from people at the margins of the system who saw further than the institutions did. Your "authorities and amateurs" point is that same tension, and Braskem is a data point for it, not against it.
So the question I'd sit with isn't "wasn't this heroic" or "why didn't we plan for it," but what institutional arrangement makes on-site continuity and supply resilience a default rather than a fluke, without designing out the thing that actually worked here: competent people improvising, trusted and paid to. The ALLFED project @jamesmulhall🔹 mentions seems exactly the right shape, turning a lucky one-off into something repeatable, compensation and trust included. COVID was a fairly forgiving test of all this; worth building the system while that's still true.
Thanks for this. I want to defend the principle in your section 2 and push two threads further: a near-term funding case that has shifted in your favour since 29 May, and the longer-term response-capacity case in your sections 5-6, which I think is where the real opportunity sits. From an impartial view it doesn't change things that this is unlikely to go pandemic, or that the suffering is in central Africa rather than closer to home; it counts the same.
1. The near-term case is stronger than when you wrote it
The "it's already funded" objection is being falsified in real time. On 5 June, WHO and Africa CDC launched a consolidated six-month plan asking for $518m, not ~$350m, and pledges are soft and falling: Africa CDC put them at ~$290m in late May, down from ~$500m a week earlier, with the headline figure still bundling in-kind and repurposed budgets. The structural reason matters for an EA audience: the US withdrawal hit this exact response (health-worker layoffs, supply shortages) and no other donor has filled the gap it left. The "vivid disaster, then money floods in, so low room for more funding" dynamic that made GiveWell cautious in 2014 is not operating this cycle.
That is the sharp point. GiveWell judged even the multi-billion-dollar 2014 response plausibly as cost-effective as its top charities, yet still declined to recommend marginal donations, because governments were flooding in. That single reason is exactly what is failing now. Honest limits: this is no longer an "early" intervention in your section-2 sense, so the case is stronger on room-for-more-funding, not on timing; and whether it matches 2014's cost-effectiveness depends on how large the averted caseload turns out to be, which we cannot yet know. With R still above 1, every week of delay keeps multiplying that caseload. The toll is also worse than a DALY count for the outbreak alone suggests: Ebola disproportionately kills health workers, and each loss thins an already-thin system for years, well beyond Ebola.
Where a marginal donation can and cannot go: no individual can move a shortfall in the hundreds of millions sitting at the sovereign and multilateral level (WHO, Africa CDC, national plans). MSF, the most visible operator, is ~98% privately funded and largely independent of that shortfall, so I doubt a marginal MSF gift is the binding input either, though I cannot show its outbreak-specific position. But the burial-and-community-engagement layer, the most agile and scalable bit and the slice the US defunded, is retail-reachable: IFRC's CHF 29m regional appeal funds exactly this, with a dedicated donation page and ~200 DRC Red Cross volunteers already working in Bunia and Rwampara. The honest gap is not the route; it is that no one has evaluated it GiveWell-style. Which is the bridge to the bigger point.
2. The longer-term case: the response-capacity gap EA keeps re-noticing
Your sections 5-6 are the more important half, and I would push them past "should we build something?", because most of it already exists. That is the more useful answer to your "what's already there so we don't reinvent it" question. The counter-cyclical fund meant to disburse before the appeals are written is the WHO Contingency Fund for Emergencies, and the funding collapse this cycle is partly the story of that fund being starved: it has released only ~$3.9m into this outbreak so far. The layer that places skilled responders into a live outbreak is GOARN, which already deploys epidemiologists, surveillance, data and risk-communication experts within 48 hours. So the gap is not building those two; it is that they are chronically underfunded and disconnected from EA money and people.
What genuinely doesn't exist is the third piece: independent, EA-style evaluation of whether a marginal dollar does most good through the CFE or through an operational appeal like IFRC's. That sits squarely in this movement's comparative advantage. The work is about money, not bodies, and it starts with that evaluation: do it in advance rather than mid-crisis, and money can flow to whatever clears the bar. The failure mode that should worry us is the reverse: channels that would clear the bar never get funded, because no one has done the work to show that they do. That, not new infrastructure, is what "a vetted route from day one" actually takes.
On the x-risk link you raise: I would not claim this outbreak is existentially relevant in scale. Its value is as a drill. Engineered-pandemic preparedness suffers from tiny probabilities and absent feedback, and the most recent real stress test was sobering: COVID was neither the Big One nor engineered, yet wealthy governments still failed on testing capacity, surge coordination and public trust. A live, p=1 outbreak is the cheapest available place to build and test exactly those muscles, plus surge logistics and diagnostics for a strain the standard field tests miss. The transmission dynamics differ, so the clinical specifics don't all transfer, but the operational and trust layer that broke in COVID is the layer a real outbreak exercises. Building response capability this way is plausibly the most tractable thing we can do for the engineered-biorisk class outside life-science R&D, and it pays off even if the tail risk never arrives.
Net upshot: if you're giving now, the IFRC appeal looks like the cleanest retail route; the larger and more durable prize is standing evaluation of these channels, so the next outbreak starts with a vetted route instead of this scramble. Is anyone already doing that, formally or informally? If not, it seems worth a few people scoping, which is the question your section 6 is really asking.
(Disclosure, in the same spirit as the OP: I used Claude for research synthesis and a first draft; the framing and judgements are my own.)