One downside is that you may end up with a suboptimal filter. The one you linked, for instance, is P2 (~N95 in US) rather than P3 (~N100 or P100) -- so less protective against particulates. It does add some protection against organic vapors ("A1") -- but I don't think that would add anything in a pandemic scenario and likely reduces breathability. 3M does make some P3-rated models in that series, but I bet the cost is higher and breathability worse (because they also incorporate ratings against other gases and/or a higher rating against organic vapors).

The costs of masking in 1:1 -- in terms of lost social benefits -- strike me as much higher than the costs of masking in a 1:many presentation (if you're not the presenter).

I'm not current on what an infective dose is in 2025, but it's not obvious to me that the risk reduction from distance/non-direction in the quiet-room scenario is greater than the risk reduction from (e.g.) 98-99% fewer people who might be infectious in the 1:1 scenario.^[1] This study from 2023 suggests that the dose from a person who walks into a room will be high enough within minutes to infect someone else in the room. So I'd be hesitant to assign a massive risk reduction without more evidence. 

1. ^{^}

   Of course, there may be other people around to account for in the 1:1 scenario.

Without testing, I would assume the imitation filters have similar performance characteristics to the "fraudulent" ^[1] ones NIOSH recently tested. They did better than I would have expected, but the differences were still significant. Maybe the imitations have a use case for particularly cost-conscious users, or for those who want a stockpile for friends and neighbors.

1. ^{^}

   The fraudulent models generally included NIOSH markings (e.g., P100) that were not authorized. As far as I know, they did not include fake 3M markings.

What's the current thinking about risk of counterfeit items? The filter link currently points to a small, random third-party seller on Amazon. I recall seeing reports of some fake disposable N95s in years past, although I never could get a good sense of the prevalence.

Thanks -- this is helpful in understanding different assumptions at play here.

At the outset, I'm inclined to defer to AIM -- not because I am inclined to defer to EA orgs in general, but because experience suggests that nonprofits (and other types of entities witout market discipline as well) are much more likely to err by expanding to fill available budget than by constricting their activities. So while it's certainly possible that AIM has misjudged the tradeoffs, I start at a place of some deference.

All you have to do is spread it to all the members of AIM, rather than pay for a new department, ie. each member of AIM takes on mentoring a few people.

If the AIM staff have available bandwidth, I'm guessing there are a number of different projects they could take on. We don't know what the counterfactual use of staff time would be.

As for the quota, isn't that self set, so the budget is there to choose 25 people. If instead you choose only 20 because your criteria rejected the rest, now you have a budget excess. 

An organization does not have to spend its budget; it can spend less and then has to fundraise less for next year. If I recall correctly, AIM is very conscious of what the likely counterfactual use of funds donated to it would be.

I think this is a stronger point insofar as the relevant costs are fixed / sunk at the point of selection. I know some are, and some aren't, but don't know the relative proportions. (Note that I would include the broader ecosystem's costs, such as seed funding from non-AIM sources.)

Everyone knows there's no magic formula for picking a hit record, you just have to sign a bunch of bands and let them go crazy and see what happens.

This metaphor doesn't work for me very well. In GHD/AW work, we have the ability to get a great return off of the existing catalog of "artists" (non-profits). In contrast, assume it is hard to invest money at good returns in any musical artist who has already proven themselves. Also, the hypothetical music investor should be willing to make investments as long as the expected value of the investment is positive; there is no pre-determined hard cap on funding available (and the investor should be able to get as much funding as they can find good investment targets). In contrast, the charitable "investor" obtains impact (which isn't convertible into money) and so is limited by the size of their bankroll.

There's a limited amount of seed and mid-stage funding available, so I would have concerns about exceeding the ecosystem's carrying capacity. The practical effect of significantly increased cohort sizes may be moving more of the culling decisions from AIM to the early funders. That strikes me as having some upsides and downsides.

• To the extent that one thinks there should be more seed/mid-stage funding (and is willing to accept the counterfactual reduction in funding for established charities), that isn't really in AIM's power to control.

In the end, my napkin model (low confidence) goes something like this:

• Investing in AIM and its early-stage incubated charities, under current operating conditions, is slightly more cost-effective than the GiveWell or ACE alternatives.
• AIM has moderate ability to identify founders that are more likely to be successful.
• AIM has moderate ability to identify projects that are more likely to be successful.

That model doesn't rely on a belief that AIM is great at identifying good founders or projects. But it does suggest that being less selective could easily flip the decision to donate within the AIM ecosystem vs. the GiveWell or ACE ones.

And there may be a tradeoff between attracting more experienced candidates and moving to a model where more charities are founded but a greater percentage of them fail. For example, some career fields are less likely to look kindly on someone doing something in left field and then trying to return if the new career was unsuccessful. Some of that is unavoidable, but I expect that being selective with founders and not having more incubated charities than could plausibly get midrange funding would reduce the risks to mid-career folk.

In reading the discussion of elastomerics and triethylene glycol, it seems to me that the emergency logistics would be extremely challenging to pull off on a short time frame. I know particularly little about the latter, so I'll comment on the former instead.

When compared to a mundane FEMA (Federal Emergency Management Agency in the US) action, I think a biodefense nightmare scenario would be a significantly more difficult scenario in which to execute quickly (e.g., within 24 hours as stated in the victory condition for elastomerics). One major difference would be the global (as opposed to local/regional) nature of the extreme emergency. 

I think you'd have have the elastomerics stored in the right locations, know exactly where they need to go, have enough willing hands and other resources to move them from the stockpile to end users quickly, protect them against diversion, and so on. Not sure who would be charged with this, but I wouldn't want to put the country (or world's) survival on (e.g.) FEMA's shoulders in light of the logistics complexity and FEMA's mixed past performance.

Does the estimate of less than $500M for achieving a elastomeric victory condition include everything necessary for a solution that could reliably pull off a quick distribution? 

I think people should give back/away FTX money they hadn’t already spent or earned. But I take a liberal definition of spending and earning.

This sounds reasonable to me, with the caveat that I think the definition should be significantly more liberal for natural persons than for non-profits. By performing work, an individual like Elizabeth incurred a detriment to herself (she could have sold her labor to someone else). Although I don't know the specifics of the person who incurred "the cost of moving back home and reestablishing her life" when FTX exploded, that sounds like a detriment too.

But if a fraudster gives a cancer-research charity $500K, and it spends $500K on cancer research, I don't see any detriment to the charity in the same way. Given that the sole purpose of the charity is to fund cancer research, it's hard to call the funding outlay a detriment. Maybe the charity dropped its funding bar slightly and so repaying the whole $500K would put it in a slightly worse place than if it had ever accepted the fraudster's money in the first place. But disgorging the dirty money doesn't leave the nonprofit holding the bag in nearly the same way as it would leave an individual worker or even a for-profit who exchanged reasonably equivalent value in goods or services.

when we haven't even effectively stopped people from buying puppies from puppy mills

I'm curious about effectively stopped here (cf. also unresolved). It's often the case that the addressing the first (say) 50% of a problem is significantly easier than the last (say) 20%. And so it doesn't strike me as surprising that it would be more effective to move on to mitigate new issues rather than devote a large amount of resources in an attempt to fully resolve previously worked-on issues.