On Tuesday morning as I was grabbing croissants, the man in front of me turned around and told me to back away. ‘Because of the pandemic.’ He was the only one in the bakery wearing a mask, yet in every way he gave off the clear vibe that I, not he, was the one being unreasonable. There was a time he would have been right. There was even a time I was the person being crowded and upset about it.

Now the pandemic, for this kind of purpose, is over. Now I had to tell myself to let the Wookie win. There was space behind me, in no way was it Worth It to do anything but put my hands up and walk away.

I remember it because it is in such sharp contrast to the everyone and everything else. Covid over. People sick sometimes, sure, and yes I missed a 1-on-1 meeting because of someone’s Covid exposure. Eric Topol continues to warn us how bad things are and not to get numb, despite it being way too late for that. Still.

I was going to say that this week marks the end of the Covid posts being majority Covid content. Then it turned out that this week’s is still largely Covid content, as it snuck up on us with a lot of content this week (none of which is alarming). So the point of transition isn’t quite there, but we will get there soon. I predict slash hope there will only be a handful more of these posts with this high a ratio of Covid to non-Covid news.

The biggest development this week in things-relevant-to-my-interests was that PredictIt’s no action letter was revoked, effectively forcing it to shut down its markets in February 2023 unless something changes by then. The most credible story is that this is new prediction market website Kalshi performing a hit job to take out the competition, but I have not had the time to look into it yet. Whether or not that was the reason, the shut-down is quite bad, speaks quite badly of everyone who let this happen, and is something worth fighting against if there is a practical way to do that. PredictIt was far from perfect, but it was a dim light in a dark world.

Executive Summary

1. Reports from people on their quest to get Paxlovid.
2. PredictIt was shut down, will cover that at some point.
3. Monkeypox growth slowing worldwide but not in America.

Let’s run the numbers.

The Numbers

Predictions

Prediction from last week: 750k cases (+3%) and 3,000 deaths (-2%)

Results: 690k cases (-6%) and 3,208 deaths (+5%).

Prediction for next week: 650k cases (-5%) and 3,200 deaths (+0%).

Decline in cases continues in the West with the rest mostly looking like a random walk. I am going to predict a continued small decline, whereas for deaths I will predict a continued small increase, but mostly this is all random walk at this point.

Deaths

This number came in disappointingly high. I could go state by state and nitpick but nothing is that blatant and we get random bad reporting in both directions each week.

Cases

Biden is feeling much better now. As usual, the rebound was harmless.

The Quest for Paxlovid

Several people wrote about their efforts to acquire Paxlovid.

From Adamsb6:

From Tom VanAntwerp, looks like his path may no longer be available, alas:

From my email:

Plushcare is a telemedicine service, got me Paxlovid for around $120 with a 15m same-day appointment, didn’t even have to lie (though I think it helps). Have heard people say something like 80% success rate for getting same-day pharmacy prescriptions out of it. Getting some now, ahead of time, is probably wise if one can stomach the deception, so you don’t have to deal with the uncertainty when sick and tired and stressed and probably handling 3 other logistics issues. 

My strong anecdotal impression is that the Paxlovid rebound thing is pretty real (my social circle tests quite regularly and around half of Paxlovid cases I know had a rebound while I can’t think of anyone whom I know with a rebound that didn’t take it). 

To the extent I have a coherent model around this after talking to bio folks, it’s something like “Your immune system figuring out how to defeat COVID is bottlenecked in part by the number of viral particles around to try things on; if you cut back on viral particles artificially via Paxlovid, you will test negative after a bit, but when the few remaining ones come back after you’ve stopped the drug, your immune system still won’t be set up to eradicate them and the infection recurs.”

Physical World Modeling

New Long Covid study claiming it can be identified from immunological data with high accuracy. I’m out of time this week so I haven’t looked in detail yet.

Another pointer to Long Covid failing to create unexplained workforce absences or other signs of poor population health, which continue to track acute Covid cases. His analysis that finds very few people reporting being involuntarily out of the labor force due to illness. If people can’t work due to Long Covid they are doing quite a poor job of reporting it in official paperwork.

BA.5 still slowly gaining dominance, up to 87%.

A deeply cool graphic, as the years and antibody levels go by.

Theo Sanderson @theosandersonAntibody waves in England, Mar 2021 – Jul 2022

August 5th 2022299 Retweets953 Likes

Study from Qatar says three doses provide strong protection against infection (paper), as did prior infection and hybrid immunity.

Agreed, not a surprise, but the waters keep getting muddied. And another new source:

There were too few unvaccinated adults in the country to study, which is a different kind of good news.

Meanwhile, the updated boosters would be better, but hey, it’s fine, no worries?

This in ‘I don’t know who needs to hear this but some people somehow do’:

Paper on Covid viral shedding, of course already somewhat obsolete since it doesn’t use BA.5. BA.1 and BA.2 interestingly had same shedding levels. Also points out that different people shed different orders of magnitude of virus, which emphasizes the superspreader principle once more. And also:

Study showing very little in-class transmission at Boston University (paper), where there was a mask mandate and a vaccine mandate. 140k in-class events and only 9 identified infections. Of course, what does ‘identified’ mean in context?

Results  More than 600 000 polymerase chain reaction tests were conducted throughout the semester, with 896 tests (0.1%) showing detectable SARS-CoV-2; there were over 850 cases of SARS-CoV-2 infection identified through weekly surveillance testing of all students and faculty on campus during the fall 2021 semester. The rolling mean average of positive tests ranged between 4 and 27 daily cases. Of more than 140 000 in-person class events and a total student population of 33 000 between graduate and undergraduate students, only 9 instances of potential in-class transmission were identified, accounting for 0.0045% of all classroom meetings.

Conclusions and Relevance  In this cohort study, the data suggested that under robust transmission abatement strategies, in-class instruction was not an appreciable source of disease transmission.

That is indeed plausibly sufficient testing to conclude they weren’t missing anything.

In Other Covid News

Child vaccination progress as good news.

Young child vaccination progress as chart.

This is very much not a lot of vaccinations. Even 17% is not that many.

White House pitch to get kids vaccinated. About what you’d expect.

Many kids (28%) don’t get vaccinated because parents don’t want to take time off work and/or (38% of) parents don’t know where to get the vaccine. Is vaccinating your child worth missing work? It is not obvious.

80,000 tourists trapped in Sanya City (population ~1mm) when it suddenly went into Covid lockdown. China continues to hold onto its zero Covid policy and pay the price.

Maps of who is considered ‘up to date’ on their booster shots.

Number is expected to increase slowly as more shots occur, then decline sharply at one time as requirement for being up to date requires additional shots.

The problem, clearly, is this recommending of additional shots. Without that we’d have so many more people up to date.

Always look on the bright side of life.

Explanation of what’s wrong with the study is here, but I bet you can figure this one out on your own.

Summary of state of regulation around these parts.

It’s not as clear a proof as one might think. I can imagine worlds in which we need to and can hope to control spread, so the travel/isolation hassles dominate but they are correct. As you know, we very much do not live in that world.

Some schools continue to have this in the worst way.

If I am required to test my children, I will administer a test and report the results. If I am not required to do so, and they are not sick enough to require medical attention, no way am I giving them a test.

Meanwhile, the CDC continues to encourage more masking, including as a way to address kids’ anxiety and other mental health issues about Covid. Have these people met children?

Monkeypox

Global spread is slowing a bit, while the infected population remains 97%+ MSM, but American counts continue to rise.

That’s a 10x over 21 days, then a 10x over 27 days, then a 5x over 16 days, then a 2x over 12 days (so a 10x over 28 days combined). Exponential growth is not yet slowing much here in America. If we didn’t have the example of other places, or this was spreading into other communities more, I would be a lot more worried.

In terms of predicting future cases, Ross Rheingans-Yoo points out (full post) that Metaculus lists the median case count as its community prediction, whereas one likely is more interested in the mean expectation. The mean is a lot higher, as the distribution is skewed. The problem is that one does not want some large outlier prediction to greatly warp the result, and even Ross is mostly cutting the predictions off at the 95th percentile because of this. So it is tricky. My guess is that both the mean (probably with extreme outliers removed) and median should be reported in cases where they differ substantially, as both numbers are important to know.

New paper is out (direct link).

Having failed to secure its vaccine supply, but with the FDA ‘fully committed’, America is going to reduce dosage and changing method of administration to make up for it.

Instead of giving a full dose subcutaneously (into the fat layer under the skin), a lower dose of the vaccine could be given intradermally (injecting the vaccine within the skin). The FDA may soon issue an emergency authorization for Jynneos to be administered in this way.

However, if that did not work, it would be bad.

But this strategy could also backfire. If there is a risk to effectiveness, it may be better to focus the available full doses of vaccine on individuals at highest risk — men who have sex with multiple male partners — to provide the best chance of getting the outbreak under control.

Uncertainties in the effectiveness of Jynneos will be magnified by switching to a new mode of administering a lower dose outside of a clinical trial. While the data support Jynneos as effective in reducing the severity of monkeypox, the data do not show that it prevents mild disease or transmission of the monkeypox virus.

I do not know if the math supports more lower doses but it seems like our system is likely to be overly conservative about making this kind of switch under uncertainty. Given they are actually doing it, I am confident it was the right decision based on what they knew at the time.

Other errors aside, this is a very positive development, and brings hope for other good decisions in the future.

Note that the virus not stopping transmission would not in any way be an argument against lower doses, since transmission would not care about how many people are infected. What matters then would be which method in practice prevents more severe disease and death, which given our current state of knowledge I presume favors smaller doses.

Another concern is that we do not know how effective the vaccine is. The one concrete number in the article is 86%, but it is based on limited data. That should still mostly be good enough for practical purposes, as decisions in the 70%-90% range in practice probably don’t differ that much.

There is potentially another vaccine, ACAM2000. This article argues against approving it until we have more data, and seems like it makes an unusually good case for this as taking ACAM2000 is very much not a free action.

Speculation on what other groups might get Monkeypox if exposed to it. Short video with the basic Monkeypox info if you need that for some reason.

How to avoid Monkeypox? NYC department of health continues to say the obvious.

More importantly, this is saying: In other contexts, don’t worry about it. Draw a distinction between what is possible and what is actually dominating in practice.

Washington Post covers the continuing debate about whether or not to give people these obvious physical facts, with the other side of the debate saying that warnings and admonitions ‘don’t work’ and that ‘it is futile to tell people to have less sex.’ Which is partly true, in the sense that many people will ignore them. And mostly Obvious Nonsense, in the sense that many other people will not ignore them and adjust their behaviors, at least on the margins, in ways that make everyone better off.

Several said they planned to avoid casual sex at after-parties. Attendance was down by thousands compared with previous years, and participants remained spaced apart as they browsed booths hawking leather harnesses and gawked at men dressed as dogs.

…

For the time being, he said, “no dark room sex parties, no orgies.”

…

But at the street festival itself, warnings about monkeypox were hard to find. Only one of the attendees interviewed said he received an informational pamphlet about the virus, even as organizers checked for proof of coronavirus vaccination.

There’s the contrast. Even at such a gathering – outdoors at a street festival – they check for proof of Covid vaccination but don’t warn about Monkeypox.

Another perspective is that a focus on ‘anyone can catch this at any time from any contact with skin or clothing’ is going to perhaps cause more fear and more stigma than the alternative? While being massively misleading in terms of real risk.

When people do get Monkeypox, the medical system mostly has no idea what to do when it isn’t too busy worrying about potential stigma, including often misdiagnosing patients causing delays in treatment that are often extremely painful, forcing the impacted communities to figure it out for themselves. They have some practice.

Trevor Bedford talk on Monkeypox. Seems almost entirely reiteration, but he’s top notch, so good source if one is desired. And yes, if Monkeypox was super infectious we would know.

What does it mean that monkeypox is now an official public health emergency and what are we up to? Mina is a big fan of the switch to lower vaccine doses, which should not have required an emergency. Then of course he moves on to at-home or rapid testing, where he confirms:

1. Rapid tests for monkeypod exist.
2. Those tests are illegal in America because FDA, probably for a year or more.

Holding steady at 8k tests/week while case counts increase.

Nothing ever changes. We’re also going to do the blood supply fears thing again because ‘in theory’ pre-symptomatic Monkeypox could be transmitted through a blood transfusion. I’d suggest testing, but how would we do a thing like that?

Vaccination and its Discontents

From Bloomberg, work continues on a universal Coronavirus vaccine, but the FDA is back to its usual glacial pace and is being profoundly uncooperative, slowing progress greatly. Also from everything I’ve heard, the Biden administration is not there to logistically help such efforts, whereas Operation Warp Speed did what it had to do. The correct approach would have been to do this in parallel with other vaccine efforts the whole time or at least after work on the original shot. But it sounds like only after the Omicron variant boosters failed did the push begin in earnest. We are very bad at doing things in parallel to ensure at least one of them gets the job done.

New York has a polio outbreak. Score one for wastewater monitoring.

Vaccines, man.

Most school-aged children have received the polio vaccine, which is a four-dose course, started between 6 weeks and 2 months of age and followed by one shot at 4 months, one at 6 to 12 months, and one between the ages of 4 and 6. According to the health department, about 60% of children in Rockland County have received three polio shots before their second birthday, as have about 59% in Orange County — both below the 79% statewide figure. 

According to the CDC’s most recent childhood vaccination data, about 93% of 2-year-olds in the U.S. had received at least three doses of polio vaccine.

…

Meanwhile, adults who are not vaccinated would receive a three-dose immunization, and those who are vaccinated but at high risk can receive a lifetime booster shot, according to the health department.

The vaccine is 99% effective in children who receive the full four-dose regime, health officials said.

Orange and Rockland counties might not be the most enlightened places in America, and you can say what you want about New York State in general, but things being so much worse than the overall 93% rate seems quite odd and also alarming.

It is very important that we not allow Covid vaccine hesitancy and politicization to carry over into other far more crucial vaccinations. So when, for example, Gavin Newsom asks whether it is ‘hypocrisy or fraud?’ to not mandate Covid vaccines for kids while mandating seven other vaccines, he is making it more likely we are going to see more general resistance to vaccinations and face measles and polio outbreaks.

The policies in question, of course, are neither hypocrisy nor fraud. It is common sense that many other vaccines have much stronger cases to make than the Covid vaccine, and one can reasonably think it is right to mandate those other vaccines but that the Covid vaccine does not rise to that level, or even that the Covid vaccine isn’t worthwhile in children. You might disagree with that assessment, but there’s nothing unreasonable about it (or about disagreeing with it), and the last thing you want to do is point out that this requires ‘unfortunate exceptions’ that need to be corrected.

Unless the point is to sabotage vaccinations to win elections. In which case, you might be the baddies.

You know who is even more anti-vax on this one? Denmark. They not only don’t mandate it, they are banning Covid vaccinations under 18 outright.

I am not a fan of mandating Covid vaccinations for children, and wouldn’t be even if it didn’t risk increased anti-vax sentiment elsewhere. I am even less of a fan of mandating not vaccinating children, of making this action illegal – even if the costs do outweigh the benefits, this is a clearly reasonable decision. Instead of what seems like the obviously correct option – letting parents decide – we have a world in which the CDC strongly urges one thing, Denmark outright bans not only it but versions with much better cost-benefit profiles, and we have governors dunking on states for not making the thing mandatory in ways that endanger our general vaccination program. Great.

Mason offers anecdotal evidence from her baby’s birth month forum that this is a real thing, note the 122 comments.

No idea how common this is.

Not Covid

Venkatesh Rao gives an Official Hot Take: That there is a conspiracy afoot to get people to quit working from home and get back to the office. I doubt there is an explicit conspiracy. I do think the Implicit Coalition is backing this play, which largely amounts to the same thing, and that we should ignore them. Working from home is where it is at and if anything we should be taxing commutes.

UK subsidizing indoor dining with your friends during Covid as a metaphor for likely future AI policy interventions. Consider that at least some policy decisions will be approximately this stupid.

Standard ‘checkmate, libertarians’ challenge as applied to AI. No, it’s not weird to have a general philosophy of it generally being good to allow things but be against one particular thing because you think that particular thing would be extremely bad. For some reason Scott Alexander feels compelled to respond in detail as if there is something to wrestle with here. Whereas the only actual challenge is ‘if we think AI is a dire enough threat perhaps we should completely reverse polarity and support bad things over good things’ which I have very much considered and rejected.

Good news, we solved AI alignment, and the secret was corrigibility.

Dan Luu keeps running into bad/fake products on Amazon, finding it impossible to be confident he’s getting the real thing. I notice I am confused because my family orders from Amazon constantly and never runs into this issue. The example Dan links to about a stolen camera seems frustrating but the mechanism there is ‘someone stole the $7k item during shipping and Amazon tried to avoid refunding it’ which is bad but a different mechanism. Maybe this is a bigger problem in Canada?

Democrats attempted but failed to cap out-of-pocket payments by the insured for insulin to $35/month. Link presents an utterly unsurprising-on-any-level theory of why insulin costs so much. Spoiler alert, it’s the United States Government again. It is still not obvious that such a cap would make things worse. Demand is fully inelastic so presumably the cap would have purely distributional effects?

Right now 14% of our 7 million insulin users (so ~1 million citizens) are spending more than 40 percent of their non-food non-housing (post tax) income on the medicine. Both things can be true at once, that the government is mostly responsible for this mess through allowing regulatory capture, and also that the current situation could be improved by further messing with things even if it could be much more improved by no longer messing with things and no longer causing the mess in the first place.

Robin Hanson asks for examples of moral progress caused by new moral arguments, gets Tyler’s response of ‘don’t beat you kids,’ which as Robin points out seems like fact-based progress – we found out it isn’t necessary and it backfires.

Food expiration dates don’t have much science behind them, ‘and are responsible for about 20% [of the 31% of all food] wasted at home.’ The awful truth is that they are numbers someone made up and could be based on anything at all.

For example, a food producer may survey consumers in a focus group to pick a “use by” date that is six months after the product was produced because 60% of the focus group no longer liked the taste. Smaller manufacturers of a similar food might play copycat and put the same date on their product.

The claims seem to be…

1. The food expiration dates might be too soon.
2. The food expiration dates might be too late.
3. The food expiration dates are not be the right way to think about it.

The central complaint is that people are throwing out ‘perfectly good’ food because it has expired, thinking it is unsafe when it is not. And then at some points people don’t throw out other food that is safe, because food in some conditions (e.g. wrong temperature) spoils faster.

The dates are highly useful, and in my mind one of the most successful mandates, exactly because they do not have to mean anything.

The important thing is that they do have relative meaning. I can notice that this gallon of milk is at day +7 and this gallon is at day +3, and now I have enough information to know the distribution of possible quality states of the milk. Or I can look in the store, and know what I should buy. I can compare the whole milk to the skim milk and notice how much faster the skim milk spoils.

Could they have put different labels on and I’d never know? Sure, but the important thing is they have consistent labels for a given brand that have relative meaning.

Is it a problem that people throw out ‘perfectly good’ food before expiration? In some situations, mistakes are being made by people who are following rules without knowing why, but mostly I am going to go with ‘no.’ Food that might have gone bad and probably won’t taste right generally is not worth keeping around.

I also call BS (based on a hunch rather than digging into the evidence, to be clear) on the ‘20% of food’ and ‘$161 billion dollars per year’ numbers, which I am guessing are based on ‘what if everyone magically knew exactly which food was and wasn’t unsafe and we treated all safe food as good as new.’

This news source ‘advices consumers to use their own judgment’ on expiration dates rather than treating them as sacred pronouncements. Yes, I would do so as well, in both directions.

Then there’s the whole Montana milk situation…

Like most public health advice, an expiration label is a compromise between giving people accurate and complete information versus giving them rules and explanations simple enough they can remember and follow them, and concerns about what people will believe given what they are told. That compromise is often not made well, but here I think it is reasonable under normal circumstances. What would be a superior equilibrium?

IRS is adding a bunch of funding and new employees, including agents. This is badly needed. Paper returns are endlessly backed up. Computer systems are ancient and can’t handle things like ‘I have five dependents’ without a manual fix. Is some of this to go after tax cheats and do audits? Yes, and that makes sense too given that for a while there has been almost zero doing of that particular thing. When you do zero of it a lot of people start doing a lot of cheating.

My personal experience is that I am damn well trying to pay the taxes I owe and the IRS has managed to make a supremely confusing series of mistakes. It took me and my accountant quite a while to figure out what they had done, because they do not explain what is going on. Once we figured it out, we tried to get it sorted out, but it was impossible to get anyone on the phone without the line cutting out within a minute or two (well before we could explain the situation), and it has been quite a long time without getting a response from our written inquiries as well. No one can get an agent on the phone.

Everyone thinks they hate the taxman, but what they actually hate are the taxes. Both the money owed, and the giant mess that is dealing with them. The taxes suck. The paperwork sucks. The distortions suck. The taxman is mostly trying to sort out the giant mess that is the taxes and it is good that someone might actually do that.

This previous lack of resources is not an accident. There was a deliberate choice to starve the IRS of funds. Because of reasons, you see.

Dustin Moskovitz asks and answers: How many EAs does it take to change a lightbulb? I love how you’re thinking about theories of lightbulb change, here’s a $50k FTX regrant to move to Berkeley and think about Lightbulb Safety. Any one EA could change the lightbulb on their own so we should worry about the unilateralist curse.

Good News, Everyone

Colorado could be next state to allow non-lawyers to provide some legal services. Lawyers have ‘very serious concerns’ about the proposal.

Treasury thinks use of Strategic Oil Reserve cut between 17 and 42 cents from price of gas. Quite the choice of exact bounds there. Also lots of room for improvement in execution.

Bruce Springsteen correctly charging market prices for concert tickets. Others, naturally, do not see it that way, but I’m glad the boss has embraced efficiency.

Some of us still remember our shopping skills.

Time article about Effective Altruism, which seems to have the EA community seal of approval. Vox has a profile. There is also Tyler Cowen’s interview with Will MacAskill. And, by Kelsey Piper’s report, a lot of babies, which is excellent news, and she is here to share her takes. A noble attempt at 227 takes that I haven’t yet fully processed. The rate of ‘take bankruptcy’ whenever someone says ‘1 like = 1 take’ remains very high. Zohar Atkins has a thread with very good thoughts. I spent much of the last week at Future Forum, largely among many EAs, and things seem to be trending in good directions there. I have big disagreements and I often criticize, but that is because EA is worthy of criticism.

Some talks from Hereticon posted online, haven’t yet had time to watch any of them. Still not sure how to interpret having not been invited.

The federal bureaucracy can save Congress from itself when it does things like pass subsidies no one can technically qualify for. Or even better, it could choose not to do that on the basis of electric vehicles having waiting lists everywhere so a subsidy on purchasing them seems highly unnecessary. Spend the money on lithium mining.

Politics is divisive, but doesn’t have to be. We can all agree on something.