Monthly Roundup #8: July 2023

As always, plenty of non-AI things are happening, most of which do not justify their own posts. This is where all of those go. Skimming for things relevant to your interests is encouraged.

Before I begin, a Point of Order: I plan to have a post later regarding the rate limitation and disruption on Twitter, which disrupts many features of Tweetdeck and means that if you aren’t logged in Twitter links no longer work. Adjustments will need to be made if the situation is not fixed very soon, despite my having no fear of ever hitting the 10k rate limit itself. It is not practical to go back and fix links or other decisions on already written material but I am adjusting the process going forward.

Bad News

RIP 538. Nate Silver leaving was always going to be a hard blow to recover from, as the models are the heart and soul of the entire model. It also depends on a deep commitment, visible to all, to fair play and avoiding putting a finger on the scale or using the scale as leverage. Alas:

Nate Silver: Very bad to do a Spanish Inquisition with pollsters based on their political orientation. I love my ex-colleagues (this is coming from a new guy they hired) but if this is their practice, hope ABC will stop use of 538 brand so it isn't associated with me.

Washington Examiner: "Rasmussen must explain the nature of its relationship with several right- leaning blogs and online media outlets, which have given us reason to doubt the ethical operation of the polling firm," demanded Elliott Morris, the editorial director of data analytics at liberal ABC News and FiveThirtyEight.

No. I have listened to many a Model Talk, with in-depth discussion of how to (among many other considerations) construct a polling aggregation systems that best resists partisan manipulation, despite the need to ‘play defense’ where the pollsters respond to your policies. Using logic like this is incompatible with such an effort.

Cocaine Bear, meet Government Warrantless Survey of Private Property Bear.

It is an important fact about banks that both they will pitch beyond atrocious products at you constantly, and also that no overloaded number of indicators, however explicit, that you know better, will get them to stop. Unlike McKenzie, I do begrudge the bank and the salesperson here, actively pitching a 2% CD right now is stealing. Worse, it’s wasting your time. The amount of idiocy I had to push through to get a Chase Bank representative to tell me their real interest rate offer was… not small.

Claim that a third of software job applicants, and over 5% of actual job holders, can’t write fizzbuzz, which implies they can’t do anything else either, and then the claim that many people literally never produce any productive output yet stay employed indefinitely. Which is good news if you are competing against such businesses. It is ambiguous news if you are trying to get hired. It means you have a big edge on those people in this area, also means that area is less important than you thought.

We have a list of jobs for which workers are in short supply, called Schedule A, to streamline immigration. It hasn’t been updated since 1991. Might want to do that.

Good News, Everyone

Worth repeating: Across the Spiderverse is the best new movie I’ve seen in years. If you have not yet done so, see it now, ask no questions. Unless you never saw Into the Spiderverse, in which case see that first and same advice applies.

Excellent new long Paul Graham essay on How to Do Great Work. In so many ways no one else could have written it this way. I do not agree with everything and do not forget to consider reversing all advice you hear, and yes this is too long, yet this is an excellent essay and you should read it. I hope he invests the time to make it shorter. Plausibly deserves a full response post at some point.

Chinese friend of Paul Graham estimates 90% of highly educated 25-year-old Chinese citizens would come to USA if handed a green card. If we made this our policy, we would cripple China and supercharge America. If you think we have to ‘beat China’ in any sense, or are worried about Chinese AI, or arguing that we can’t slow down on AI because we must beat China, and are not screaming for this immigration policy from the rooftops, your position is not coherent.

Excellent writing advice from Larry McEnerney (2 min + 9 min). Write to understand. Your writing must be valuable to the reader. Recommended for a watch.

Joe Weisenthal says personal trainers are observing gym use patterns shifting (locally in NYC) to reflect a return to the office.

The statistics reflect this. New York City office occupancy breaks 50% of pre-pandemic levels for first time.

There is a clear pattern here. If it holds we should get to at least 60% later this year, and be mostly back to old levels before the decade is out.

Claim that of the four self-driving car accidents cited by San Francisco officials, one wasn’t an accident and the other three involved the self-driving car being rear ended, which by default means it wasn’t at fault, and no determinations of fault were made. None had major injuries. I’d assume it would be easy to find out who was responsible? Don’t these cars have cameras actual everywhere?

Michael Ashcroft makes the case for holding down a job. About half of it seems like ‘if you get an unusually good job’ style arguments. The other half is that not having to think and only do a narrow thing you are told within fixed bounds can be freeing and provide a narrative where you are fitting in and responsible and making progress and not blameworthy and such. Which is in fact a big deal. Eyes open, though.

Lionel Messi signs with Inter Miami over Riyadh, passing up a billion dollar payday, presumably because he has plenty of cash and gets to live in Miami and grow his brand in America. Also he’s getting true skin the game upside potential for the whole of MLS, this could be the heir to the Michael Jordan shoe contract (if you haven’t seen the movie Air about that, it is excellent).

While I Cannot Condone This

It really is strange and wonderful that we can’t trust the airline to get the bag (or, lately, you) to your destination, but we completely trust the other passengers not to steal our bag, such that when our bag isn’t there we don’t even consider the possibility someone stole it despite zero precautions being taken.

My name is Inigo Montoya. I am representing my own interests. Prepare to vote.

Claim that The Simpsons is good again. I am skeptical that one can go home again.

New York City bans discrimination based on height and weight. Knicks are on it.

Paul Graham declares peak woke.

Claim by Brian Manookian that if you challenge the chain of title validity of your student loans using the proper legal magic invocations, there is a very good chance that your loan gets wiped out. Huge if true. The causal theory offered is: Almost no one ever does this, so the companies involved stopped bothering to properly handle the paperwork. No one involved is a lawyer, none of this is legal advice and so on. Note that this type of affordance gets very interesting as AIs lower the cost of both noticing and exploiting the opportunity.

Poland simplifies visa procedures for citizens of many countries, including Ukraine, Turkey, India, Pakistan, Vietnam and Nigeria. The right wing ‘law and justice’ party intends to accept over 400,000 such visas per year, including simplifying the ‘benefit’ procedure to allow them to work. This would be more immigrants than any other country even in absolute terms, who would then get to move throughout the EU.

There is particular talk that demand is high in typically ‘male’ industries like construction, welding and transport. This indicates that Poland is not using this opportunity to choose those with high human capital, instead looking for cheap labor, and trying to use guest workers, likely without intending on offering a path to citizenship. That tends to change over time.

This is remarkably close to open borders for Poland, which then has open borders with most of Europe. Poland gets the short term economic benefits of new workers, while offloading many of the longer term concerns from taking on so many immigrants so quickly as they diffuse throughout the continent. This has always seemed like a deep flaw in the EU system for those who want to limit immigration. If you think the public favors too little immigration and too much democracy, it is a feature.

A theory of the popularity of anime.

Devon: I recently heard the screenwriting trope "Villains Act, Heroes React" for the first time and it destroyed me While there are counterexamples, it does seem like the Good Guys are more likely to protect the status quo rather than try to change the world

Patrick McKenzie: One of the reasons I think anime is popular is sometimes avoids Cambell’s hero’s journey in favor of “I’m going to achieve mastery for it’s own sake.” If you say that out loud in English fiction, *you are certainly the bad guy.*

Much like there have been ten thousand reskins of Harry Potter I’ve been waiting for more central examples of English-language cultural products to take that story archetype and just run with it. There is clearly a demand.

It wouldn’t be a minor tweak to have Peter Parker grow up hearing “You have a great responsibility to earn great power.” but one can imagine utterly sympathetic characters with basically that motivation.

Yes, I am absolutely saying that quite a lot of objections to things are exactly this. Usually they make it slightly less obvious.

Daniel Eth: Wtf is this crap? “as more governments seek to reduce their plastic waste, many fail to account for the workers who not only depend on that waste, but are also mitigating much of its environmental harm.”

~”Before we go ahead and stop this harm, we must first consider that some people make a living mitigating that harm” I am *begging* self-described progressives to stop with this nonsense.

AgiDoomerAnon: that's a parody article right?

Daniel Eth: No.

Trump arrested for retaining classified documents under the espionage act. Nixon reads the indictment, detail after detail is essentially ‘It doesn’t look good.’ Former Trump-appointed Attorney General Bill Barr says ‘if even half of it is true he’s toast, it’s a detailed indictment and it’s very, very damning.’ Jane Rosenzweig explains that the indictment is highly readable exactly because of good writing decisions, with the explicit goal of readability and the telling of a page turning story. All coverage I have read confirms that, as a matter of law, they have Trump dead to rights here.

Like everyone else I still expect Trump to somehow not be held to account, but from what I can tell a normal person would be super duper toast here.

I share Patrick McKenzie’s ambivalence here. Is it good news that the internet can often run laps around official investigators? Or is it bad news that the internet can often run laps around official investigators? Highly related to whether this is a ‘me question’ or a ‘you question.’

Patrick McKenzie: This feels like one of those rare Twitter threads that will end up in the footnotes of a criminal indictment, or at least the investigatory logs.

As something of a proponent of the Internet I’m torn between satisfaction at seeing the crowd ridiculously outperform on institutional functions like e.g. investigation of complex international money laundering and pique that the institutions keep being years late to the party.

Like you would think that the obvious suspects would have family trees and org charts of these people/firms already on a corkboard and frequently they do not.

Source: conversations which surprised the obvious suspects with corkboard-worthy facts.

Apologies for having to be oblique here on the specifics of the conversation but once had one with:

Me: That rhymes with *shocking assertion of malfeasance.*

Institution: Wait what.

Me: You know how $X did $Y.

Institution: That’s some theory! Goodness.

Me: Oh no I have documents.

Institution: You have what. But you’re not an insider there?

Me: They’re all public records. Want a copy?

Institution: Wait what. What’s your angle?

Me: I have weird hobbies.

(This was downstream of a tip-off from crypto skeptic Twitter and a bit of digging from your’s truly to turn up the smoking guns.)

Institution: How did no one important find this already?

Me, to myself: Feels like that’s a you question not a me question, dawg.

The other thread is quite the rundown. I definitely haven’t checked the claims, but here’s the first third or so minus its images:

Adam Cochran (linked thread from the first Tweet above): Back in May, TUSD threatened to sue me for even remotely suggesting they'd be related to Justin Sun [screenshot shows threat]. So naturally, I dug in to prove the ownership web of Justin for both TUSD, Huobi and his shell companies around the world.

This is just a summary. Personal/Sensitive info has been removed where reasonable. Full docs, additional pics and info have been provided to journalists to dig deeper. Thanks to the dozens of people who shared their info as well.

The first red flag I came across was on their own reports statements from their auditor. It noted: 1) "An agent" of Techteryx opened an account. 2) It was a regular corp account. Not an FBO account, not bankruptcy remote.

These have been repeatedly signed by two supposed executives of TrueUSD: -Jennifer "Yiying"/"Yiyang" Jiang -Steve Liu Both supposedly of Techteryx a Singapore company. Which of course Singapore says doesn't exist.

In fact, Steve Liu, Jennifer Jiang, Yiying Jiang and Yiyang Jiang do not show up as the officers of any Singapore company. But, a Yiying Jiang does show up on another company. DTV Limited - the UK holding company of DLive by Bittorrent. Owned by Yuchen (Justin Sun).

Now BitTorrent is owned by a holding company called Rainberry - which lists both Justin Sun, and Steve Liu, our other supposed executive of Techteryx.

Now maybe these are just former BitTorrent execs that made their own fund, right? No. If we go back to Poloniex which was initially held under Augustech LLC, we'll see Jennifer Yiying Jiang again, holding it on behalf of Sun before he admitted ownership.

[thread continues, also it contains images of various documents]

What would happen if you fed all the public information on such matters into a computer system and set it up to do this type of analysis, ideally using refined LLM tools, in the style of Person of Interest? You would learn quite a lot of ‘Hayekian details’ that way, and uncover quite a lot, no need to add in universal illegal wiretaps. Our secrets largely survive via security through obscurity, which is why someone who is willing to spend the time can figure out things like this. What happens when they don’t need to spend the time? What happens if you combine this with governmental information assets and powers?

Useful Vibes Only

Consumption should not be conspicuous. The same applies to its absence.

Daniel Eth: Note that while extreme lavishness can give vibes of grift, extreme asceticism can give vibes of cultishness.

I don't think people should up consumption against their personal desires for this reason, but I *do* think this is a reason to avoid creating norms of asceticism or assigning status for it

Oliver Habryka: Either of them creates vibes of "primarily optimizing for how you are perceived instead of its outcomes", which I think is the underlying issue. The whole "I should decide how much to spend based on signaling" thing feels really confused to me.

Daniel Eth: Hmm, I don’t think that’s true? Like, if an org is spending a lot on company dinners so employees can go to expensive restaurants all the time, I don’t think it’s a signaling game, but instead that employees like the food and figured out a way to get the org to agree to the arrangement (which feels ~grifty if it’s a nonprofit). Similarly, if someone drives themselves into/near poverty for a cause, I think the more common reaction is “they’re likely neglecting their basic needs for this cause - were they brainwashed or something?”

Cate Hall (other thread): It is totally bizarre to me that this post & nearly all comments are about the marginal effect of compensation on employee productivity and not its effect on organizational productivity via recruitment and retention.

The idea that your people will be equally good, just more pure of heart, if you pay them less is one that can only flourish in a community of people low on life experience.

[links to this Lincoln Quark post called ‘Why Altruists Can’t Have Nice Things’ where Lincoln essentially endorses providing clearly worthwhile stuff but not luxurious perks or things that otherwise send the wrong type of message, and to consider charging for things including food at events.]

The distinct tax advantages of providing employees with experiential goods should not be dismissed, especially in high-tax areas like New York and San Francisco. Nor should one dismiss the value of making someone’s life nicer or easier. People really do appreciate it, and the time saved is more valuable on the margin than it appears. Free is powerful. Offering people free food in particular has very strong returns in positive vibes and ability to not distract. I am all for using markets and prices to allocate scarce resources, sometimes it’s even worth a major tax hit to use them, but sometimes the optimal price most definitely is $0.

Can you go too far with upgrades and massages and retreats? I mean, yes, of course.

I still strongly endorse that if you want someone to work hard at an important task, you want to pay them an amount such that they’re not constantly worried about money (not profligate, but they can act in reasonable ways they are used to without worrying about it). And you definitely don’t want to be underpaying people, or not providing useful services, to ensure people are ‘on mission’ or to send a signal. Whenever you do things for signaling reasons, be suspicious.

People at nonprofits are, as far as I can tell and in my experience, constantly put under a kind of siege where they are expected to get none of the surplus they generate - they should always get the minimum they need to deliver the goods, nothing more, or else it is, as Daniel puts it, ‘a grift.’

In my experience, paying less (in total compensation, including perks) than someone’s happy price and dramatically less than their opportunity cost, asking people to make the maximum amount of sacrifices they possibly can, is a fools errand. Absolutely act responsibly. Don’t throw money around, don’t do crazy perks, especially don’t do perks that cost more than the person values them plus what they add to productivity. Also don’t guilt trip the exact people out to do the most good, or most others either.

Outback Steakhouse dominating Brazil with over 150 spots all packed. Article makes clear that the chain has much better both absolute and relative style in Brazil, and that the contrast in food styles is doing a lot of work. It had better, because the food quality very much does not justify the price level. Whereas Fogo de Chao is spreading here in America, and it’s awesome.

Government Not Working

FTC situation is not great, complete with a lot of threatening emails from certain commissioners if other commissioners don’t vote the right way. They are also going after Amazon for enrolling people in prime and not making it easy enough for them to quit, claiming millions have been unwittingly enrolled. With complaints like this:

At another point, the FTC notes that one mobile solicitation to sign up for a 30-day free trial of Prime clearly states, in prominent text, "After your FREE trial, Prime is just $14.99/month." But it does not explicitly state in the same section that the membership will auto-renew after the free trial—for that, customers must scroll slightly further down the page to a section laying out the terms.

..

The FTC has also deemed the Prime cancellation process too complicated to be legal. To cancel Prime, customers can contact customer service directly or go through an online process (which was changed in April 2023 to be even more simplified). The FTC's big complaint is that the online process required at least six clicks, over the course of four pages. On these pages, customers were presented with options like switching from a monthly plan to an annual plan (or vice versa), ending their membership on a specific future date, pausing their membership, keeping their membership, or ending it immediately.

To make matters even less bad, Prime members could get a refund of one automatic monthly charge. Someone who forgot to cancel after 30 days or didn't realize they would be auto-charged could get that $14.99 back. Still, the FTC is unsatisfied because someone "who discovered Prime charges after a few months could not obtain a full refund online."

Stratechery devotes a whole post to pointing out how good a deal prime is and how it is a customer friendly business model that benefits from scale, and then walking through the details of these so-called ‘dark patterns’ and showing definitively that Amazon’s actions here are all perfectly reasonable. The deck isn’t completely unstacked, but it never is, I really don’t know what anyone was expecting.

I actually do think it would be good to take action to deal with abuse of free trials and automatic renewals across the board. In particular, automatic renewals for extended periods, in the absence of active use of the subscription, should require periodic explicit consent. I’d also be down for universal enforcement of ‘two clicks to cancel.’ Until then, this is the worst kind of selective enforcement, for one of the best deals in the world. There are much, much worse cases out there getting away with it, as Ben points out. New York Times, anyone? Or your local gym?

Call-to-cancel is the toxic thing that must be wiped from the planet. If your website is a little obnoxious about its six clicks? Seriously, stop complaining now.

(The post then goes on to discuss a new Texas anti-porn law that applies if your platform is ‘more than one third’ porn, and asks how a social network could know if this was true, complaining it is not well defined. And I guess so, but wouldn’t you know? I feel like you would know.)

Niskanen Center publishes one of the most depressing documents I’ve seen in a while, about how Culture Eats Policy when it comes to government hiring and tech procedures. Trying to hire someone is a nightmare, you submit a request and then HR judges who can cut and paste and give themselves master rankings in everything and maybe eventually you’ll get someone through the ‘cert’ process after multiple six month cycles that can even in theory do the job at all - such as a job for web designer where zero applicants are web designers. Senseless requirements in contracts that no one ever intended, such as routing messages through an ESB loop that when it works properly does nothing, as every stage of authority solidifies requirements more and more and no one has any way to point out they make actual sense, with costs sometimes running into the billions for no reason at all.

Congress yells about such things, then everyone goes back to being terrified of justifying every step of the process in pure process terms, again regardless of whether anything involved made any sense. And so on.

Ben Hoffman suggests creating common knowledge that such systems treat costs (the requirements) as benefits, rather than the benefit being that you build the thing and it does what it was meant to do. Then perhaps we can offer solutions. That seems better than the Niskanen Center’s exhortation of ‘have less process requirements.’ Yes, have less process requirements, but how? A process requirement to minimize process requirements?

What won’t work is tinkering along the edges of the problem. The core structure is fatally broken.

What we need is a plan that works in practice, not one that works in theory, that gets everyone involved away from all of these checklists and required practices, while not opening the door to corruption, corner cutting and the whole host of principal-agent problems. The government system is designed around only paying for, requiring getting what you can technically measure. Can we redesign the system to do the thing the rest of the world does, and pay for and reward practical results? Keep in mind incentives will need to be fixed up and down the line.

Biden Administration provides aid to ports, on the explicit condition that the funds not be used for automation. Yeah, AI might run into some problems.

Periodic reminder that our immigration system is completely bonkers, including letting a single officer arbitrarily deny entry to a fully funded PhD headed to Johns Hopkins because they didn’t like their stupid face.

Reasons not to take advice from professors: World is changing rapidly so their experiences don’t apply, survivorship bias, conflict of interest.

Someone is the change they want to see in the world.

Merle Hazard: And they said there are no more heroes. He is right, to be clear.

backstory: Wikipedian Giraffedata has removed "comprised of" as many as ~90k times AND wrote a thoroughly-researched 4300 word essay about why the phrase is bad

Thread from Sandeep Parikh about his experience working with Robin Williams, great stuff. The power of engaging for real with those around you no matter what, and also that you know more than you realize you know.

Caviar really is the best delivery service and it isn’t close.

Eliezer Yudkowsky: ...okay, as much as there is no obvious reason why Caviar "by Doordash" - which supposedly doesn't even let Doordash drivers know whether or not their delivery is for Caviar - would be consistently fast and on time, Caviar is consistently fast and on time. ¯\_(ツ)_/¯

It’s true. Doordash itself is consistently a disaster. Caviar is highly reliable. Their time estimates are bizarrely terrible, but if you ignore them and use common sense, you’re golden.

Agency can be taught to humans.

Emmett Shear: The basic formula was a little like coaching. Both people I was very close to. Every time they said something in victim mindset I’d challenge, and ask more or less the same series of questions. “What’s the stupidest easiest one thing you could do to make even a little progress?”

“What if it was possible? What might be a good first step?” “It sounds like you’re sure you won’t succeed, what’s going on with that?”

Oh, and then occasionally they’d claim something couldn’t be done, we’d argue, they’d refuse to try. So I’d go make it happen, and every step of the way tell them what I was doing. After 2-3 times each they both noticed when I said things were possible, I was right.

Visikan Veerasamy: this is almost exactly what my ex-boss did with/for me, with great results

[Quoting old thread]: in the 5+ years I worked for my ex-boss Dinesh, he constantly, casually-yet-intensely would ask me about my goals & desired outcomes "what do you want to get out of this?" "how will you measure your progress?" "what's the next step?" "what's the limiting factor?"

Emmett Shear: I’ve multiple times had the thought that we need “agency therapy”…basically this process, but purchasable like therapy or coaching. I bet some coaches already offer it but it’s very distinct from other types of problems coaches focus on. Needs its own brand.

Game of Thrones sound suddenly went from world-class to awful, Lauren McKenzie claims, because AT&T forced HBO shows to use AT&T’s sound tech. I file this under good news because that means there’s a very easy way to fix it. Could also be filed under the efficient market hypothesis is false.

My general rule at this point is that I will use subtitles on all television shows except comedies, since jokes don’t land without their details and timing. It’s too important not to be confused. For movies sound quality and details can be good enough that, given I’m giving my full attention, I’ll go without.

Government Working

UK government plans ‘controlled spontaneity’ in response to terrorist attacks, to focus public reaction on sympathy for victims rather than outrage at terrorists, and to prevent giving the terrorists exactly what they want. Clearly framed as terrible in the post. This seems very good?

Senator Sinema finally gets one right, holds up major bill unless it reduces the absurd 1500 hour training requirement for pilots. The response talk of ‘blood on your hands’ is completely unhinged. If anything, this should be reversed. If you make it harder to fly and force people to drive instead, you are the one with blood on your hands. Also the rhetoric comes from Senator Duckworth, whose army pilot training was 180 hours.

I loved this comment on r/flying on the topic. Refreshing levels of quiet part out loud.

The doctor says this is good and his career field is very similar. He is a senior resident at our nation's top neurosurgery institution and apparently they have a VERY TIGHT control on how many neurosurgeons are released into the market - they do this by severely restricting incoming students as well has having crazy high entrance barrier. Overall his entire field only produces a handful of neurosurgeons each year. The result? Seven figure incomes out of residency for those surgeons.

Nothing groundbreakingly new, but I think they all agree that limiting supply is key to maintaining job security and high pay. Look what happened to lawyers and dentists - every law school and dental school started pumping them out and now their pay isn't what it used be.

So whatever bottleneck is creating this shortage for the pilots (the 1500hr rule & high cost of entrance...etc) we need to KEEP IT. Actually, go further do our best to make the entrance even HIGHER is what they recommend. You do not want fresh new labor flooding the market easily by lowering standards.

I understand that unions are well aware of this and they are against lowering the barrier.

Deeply Suspicious

If you think that’s suspicious, you can add that to a very long list of officially suspicious things.

Trace Mitchell: In IJ’s unconstitutional traffic stop case in Texas, the officer said he wanted to search our client’s truck as he was acting suspicious. But our client was calm and forthright throughout the entire interaction! So, what will an officer label “suspicious”? A suspicious amount:

• It’s suspicious if you act too nervous. 612 F.3d 341.

• It’s suspicious if you act too calm. 551 F.2d 991.

• It’s suspicious if you make too much eye contact. 622 F.2d 1218.

• It’s suspicious if you make too little eye contact. 799 F.2d 704.

• It’s suspicious if you walk with too much purpose. 912 F.2d 1014.

• It’s suspicious if you walk with too little purpose. 908 F.2d 497.

• It’s suspicious if you travel with luggage. 625 F.2d 9.

• It’s suspicious if you travel without luggage. 555 F.2d 594.

• It’s suspicious if you take a one-way trip. 878 F.2d 469.

• It’s suspicious if you take a round-trip. 555 F.2d 594.

• It’s suspicious if you deplane too early. 912 F.2d 1014.

• It’s suspicious if you deplane too late. 446 U.S. 544.

• It’s suspicious if you travel with others. 799 F.2d 704.

• It’s suspicious if you travel alone. 890 F.2d 1413.

I guess my question is, in the eyes of law enforcement officers, is there any behavior that isn’t considered suspicious?

I would have so much more respect for this strategy if they put the contradictory conditions together rather than scattering them. What’s suspicious is anything a cop didn’t expect or that looked weird, or the cop is racist, that’s it, full stop. I mean, isn’t that the job? To be suspicious?

Why was more than $200 billion in Covid economic aid funds stolen, as reported by a watchdog? Justin Amash points out this was completely predictable. Patrick McKenzie explains that in general we have the capacity to either get money out the door fast or not get a bunch of it stolen, and in this case we made a choice to get it out the door.

In Medical and Health News

Air quality is a really big deal.

That point was driven home rather forcefully this past month, as New York City’s sky briefly turned orange and spent several days being not blue, linked thread has both good details and good advice from Zeynep Tufekci, including instructions on making your own air filter in a pinch. My oldest child reporting being more dizzy than he’d ever felt, upon waking up. My youngest was coughing.

I rushed to finish the Thursday morning AI update, because I was having trouble thinking straight while working on my desktop. When I then moved next to the air filter and we turned on the air conditioning for help, my brain returned. The next day, on Friday, things were mostly back to normal, but my head was still noticeably not right.

The new plan is to permanently run three air filters, one in each room. I expect a hard-to-notice but quite real cognitive improvement, on top of health benefits. That came up on 6/30, as the AQI broke 150 again, and the filters are working as designed.

All of which is to say, yes, air quality really is a really big deal, as per Alex Tabarrok’s redux and this Matt Yglesias write-up.

This is highly underserved cause area around the world. Huge gains are available.

Lilly’s new drug Retatrutide also a really big deal assuming anything like its current results are sustained (direct link requires login).

Drew Armstrong: These are astonishing results in weight loss. 24.% weight loss at 48 weeks. Average loss of 58 pounds. Blows out of the water everything else we've seen so far. This is a hell of a slide:

NJEM: Late breaking at #ADA23: In this trial involving participants with obesity, 48 weeks of treatment with retatrutide, an agonist of the GIP, GLP-1, and GCG receptors, resulted in substantial reductions in body weight.

For context: Every other weight loss drug we've seen is around 15%. Which was already awesome. This is 10pp higher than that.

The debate has been around "will the new pill versions wipe out the injections" and this just reset the conversation.

This drug is still in mid-stage testing. A lot could happen between now and then. And the market for these products will likely fragment based on how they're administered, how strong they are and the side effect profiles. But also...wow.

... and here is the side-effect profile, which is more or less in-line w/ other competing drugs.

Unsurprisingly, there's a relationship between dose and side effect. Higher dose = bigger weight loss = more nausea, diarrhea, vomiting, constipation etc (all common on these drugs)

Lilly has over the last little bit surpassed J&J in market cap. It's now the biggest drugmaker in the world, on that measure, I believe. Which... was not the case a few years back.

If I had to guess: Given the many options out there, what patient need looks like, side effects -- the "winner" here won't be "who is strongest" but will be a lot more fragmented than that.

Perhaps results will fragment, if different drugs solve different issues, especially with different side effect profiles. It is also entirely plausible that varying the dose is mostly the best strategy. Too early to tell.

These drugs are, as I understand it, not especially fun. They are still a lot less not fun than trying and failing over and over to lose weight via diet and exercise, often less not fun than succeeding ‘the hard way,’ and a lot less fun for many than giving up entirely.

Also, being fully serious, how much to buy out the patents here and offer this free to whoever wants it? Can we for once not have a massive deadweight loss triangle, and instead have massive weight loss? I bet the government turns a profit after health care spending cost reductions and productivity gains.

NHS situation in the UK continues to deteriorate.

Prompt engineering works on humans. Our doctors are willing to largely spill the beans on the right decisions if given the right prompts, because no one is much paying them not to do that. Like with LMMs, we are only teaching doctors to hide this info in a shallow way.

There is high willingness to pay $100+/month for effective weight loss drugs. If anything the mystery is why willingness is still modest, not why willingness is high. I personally do not want drugs lie Wegovy, because they do not help with my particular issues. If they did, I’d be all over it. At $100/month you could plausibly come out ahead from reduced food costs alone.

UK did get three very important things right during Covid: Approving the vaccine fast, first doses first and the Recovery trial. All three saved many lives. Similarly, the USA did one thing very right, Operation Warp Speed. What does that say about the post title, state capacity? That in theory there is some. In most practice, not much. What other high impact options could we have used state capacity for that we didn’t?

Talk back and forth about Lab Leak probabilities. In addition to analyzing the evidence value of the new info, Tyler Cowen claims that the lab leak world is a worse world. I see why one would think so, but is this true? If Covid-19 came from the wild, then that suggests we may have little hope of preventing another such virus from developing, although we can and very much should develop vaccines and other responses in advance.

If Covid-19 escaped from a lab, then that gives us another affordance. We can avoid doing gain of function research, tighten up lab security, and reduce the chance of another pandemic quite a bit.

Or, alternatively, if we take the risks as givens, we can use the cause to implement better safety in labs, stifle gain of function research, and also rally anti-pandemic measures as it is now a man-made threat from a geopolitical rival rather than a natural cause, justifying very high ROI investments in vaccines and other responses. And the CCP losing face is certainly not obviously bad.

I do see downsides as well, especially involving political tensions rising. The sign here is not obvious.

What percentage of colleagues you would trust depends on both the distribution and also the alternative. It seems psychiatric standards of care are not so good.

David Mordecai: Perhaps the problem some people have with psychiatry as a whole is that they have never had an encounter as a patient with a really good psychiatrist? I've dealt with at least ten as a patient, dozens more in other roles. Out of that I can think of maybe three.

Jonathan Shedler: Nearly all my patients have had "psychotherapy" before they came to me. Most of what I hear about their previous therapies is dismaying.

Coop: As a former patient who regular speaks with other former patients about their experiences, the difference in outcomes with a skilled, careful, thoughtful practitioner vs someone just trying to get through the day are VAST, and the ratio is like 1/30; FAR WORSE in public sector. Professionals often have limited opportunity to observe their colleagues in practice; if they did, I suspect there would overall be greater credence granted to patient claims of widespread Iatrogenesis.

Jonathan Shelder: A colleague & I did a survey of psychologists & psychiatrists & asked what % of professional colleagues they would trust to treat a family member/loved one. Answers converged around 5% (20:1). But that's among people w top credentials. God knows % is for "therapists" generally.

A plausible model of therapy is that the good version of it is great, while the median version of it is terrible, and we have no good tools as prospective patients for identifying the good version.

When it’s not therapy, the bar can be a lot lower. If you need an SSRI and a benzo, then all you need is someone not too crazy to run the sanity checks. That still rules out maybe half the profession, but the other half is fine. Whereas if you want something more complicated, you’ll have to search, presumably with a highly itchy trigger finger on switching providers.

For Science!

Alex Tabarrok correctly points out that it is insanity to ask you to do proper formatting for a given journal prior to submission, rather than after acceptance. It’s a huge waste of time. The only explanations are an ego trip, or more sensibly as a costly signal to filter applications. The cost seems too high for the costly signal plan to make sense, you could instead charge a modest fee.

There is not enough funding of long term science, because it does not generate short term wins for funders. This is an echo of scientists not doing long term research work, especially if it might not pan out, because they have too-short feedback cycles to attend to, despite such long term risky projects being key to science progress. This still seems like massive underinvestment even with the AI situation. I don’t know what to do about this at a structural level.

Whereby a rather large problem with our system of diffusing science knowledge is highlighted by Justice Jackson’s recent SCOTUS opinion, thankfully not in a load-bearing way.

Ted Frank on October 30, 2022: Anyway, anyone want to place a bet whether the game of telephone works and takes a bad legal writeup of a bad study and the entirely fictional (but striking!) claim in the brief ends up in a SCOTUS opinion?

Andrew Rettek: So this happened and will hopefully be a "you can't trust anyone totally" moment for some people.

The exact quote from the brief is: “For high-risk Black newborns, having a Black physician more than doubles the likelihood that the baby will live, and not die.”

A fun game, by the way, is to ask Bing to fact check this claim, and have it say it’s true, while also explaining in detail why the claim is false.

Me: Please fact check the following claim: For high-risk Black newborns, having a Black physician more than doubles the likelihood that the baby will live, and not die.

Bing: According to a study published in the Proceedings of the National Academy of Sciences in 2020, the claim is true. The researchers examined data on hospital births in Florida from 1992 to 2015, and found that when attended by white physicians, Black newborns experienced 430 more deaths per 100,000 births than white newborns. However, when Black newborns were cared for by Black physicians, the mortality penalty they suffered, as compared with white infants, was halved1. The study also found that these effects were more pronounced in more complicated cases and in hospitals that delivered more Black babies1. The study did not find any significant improvement in maternal mortality when birthing mothers shared race with their physicians.

The authors of the study suggested that racial concordance between the physician and newborn patient could ameliorate outgroup biases, boost communication, and increase trust. They also called for more research to understand the mechanisms behind this phenomenon and to identify potential interventions to reduce racial disparities in newborn mortality.

Me: I notice I am confused. The claim is that the chances are more than doubled that the baby will live. Isn't that very different from cutting the mortality penalty in half?

Bing: I understand your confusion. The claim and the study use different ways of expressing the same effect. Let me try to explain with an example.

Suppose there are 100,000 white newborns and 100,000 Black newborns in a hospital. If all of them are attended by white physicians, the study found that there would be 287 deaths among white newborns and 717 deaths among Black newborns. This means that the mortality rate for white newborns is 0.287%, and the mortality rate for Black newborns is 0.717%. The difference between these two rates is 0.43%, which is the mortality penalty for Black newborns.

Now, suppose that all the Black newborns are attended by Black physicians instead. The study found that the mortality rate for Black newborns would drop to 0.497%2. This means that the mortality penalty for Black newborns would be reduced by half, from 0.43% to 0.21%.

Another way to look at this is to compare the odds of survival for Black newborns under different scenarios. The odds of survival are calculated by dividing the number of survivors by the number of deaths. If all Black newborns are attended by white physicians, the odds of survival are 99,283/717 = 138.5. If all Black newborns are attended by Black physicians, the odds of survival are 99,503/497 = 200.2. The ratio of these two odds is 200.2/138.5 = 1.45, which means that the odds of survival for Black newborns are increased by 45% when they are cared for by Black physicians.

To say that the chances are more than doubled means that the odds are more than doubled. In other words, if the odds of survival for Black newborns were 276 or higher when they were cared for by Black physicians, then we could say that their chances were more than doubled. However, this is not what the study found. The study found that the odds were increased by 45%, not 100% or more.

Therefore, the claim is true in the sense that it captures the direction and magnitude of the effect, but it is not precise in the way it expresses it. A more accurate way to phrase the claim would be: For high-risk Black newborns, having a Black physician increases their odds of survival by 45%, and reduces their mortality penalty by half, compared with having a white physician.

I hope this clarifies your confusion. Please let me know if you have any other questions.

A modest proposal by Stuart Buck: You can put your name on lots of papers without doing the work if you want to do that, but if there is fraud in those papers, you are held responsible, no saying ‘I didn’t do it.’ Real skin in the game. I approve.

The NIH to not allow AI to be used for peer review, because it is a ‘breach of confidentiality.’

“Reviewers are trusted and required to maintain confidentiality throughout the application review process. Thus, using AI to assist in peer review would involve a breach of confidentiality. In a recently released guide notice, we explain that NIH scientific peer reviewers are prohibited from using natural language processors, large language models, or other generative AI technologies for analyzing and formulating peer review critiques for grant applications and R&D contract proposals. Reviewers have long been required to certify and sign an agreement that says they will not share applications, proposals, or meeting materials with anyone who has not been officially designated to participate in the peer review process. “

Tyler Cowen was alerted to this by Raghuveer Parthasarathy, who points out that the policy makes no sense as stated whatsoever, and certainly can’t reflect any actual confidentiality concerns as written.

 My own blog comment: “As others have pointed out, this policy rationale makes little sense, regardless of one’s thoughts on whether AI should or shouldn’t be used in peer review… Presumably the authors mean non-local AI services, but if so, that should be stated. Or do the authors mean ideas that come from unsourced others? (If that’s the case, why is confidentiality the issue?)”

Tyler Cowen confirms this is Obvious Nonsense and asks: dare we expect just a we bit better from our scientific authorities?

I mean, no? Why should we? Remember Covid, also everything else? These are our scientific authorities. They are who we thought they were. Do not let them off the hook. Don’t pretend they are better than this. They are not.

That does not mean we have a better alternative. It does not mean it would not be strategically unwise to participate. The system still, in some ways and some places, tends towards truth over a sufficiently long time horizon. Eyes open, though.

Gamers Gonna Game Game Game Game Game

The most epic deck photo of all time competition. Current entry:

Last year’s entry from Ryan Grodzinski:

Whereas this was merely a normal one from championship quarterfinalist fpawluszmtg, still warms my heart.

Sam Black wins a major premodern tournament, his second back-to-back, this time with a deck built from theory, no prior games played. Man remains a legend. So many details here are so Sam Black. Why not play all the best cards?

I have never played a game of Premodern, but it is impossible not to appreciate what is happening here. It is also amazing to see the declaration that no, creatures are nothing to worry about, I will begrudgingly board in two cards I guess but don’t waste precious space.

I do think this should be sufficient to show that the engine of Enlightened Tutor, Mox Diamond, Land Tax and Scroll Rack cannot survive. It was fun for a while, it is going to rapidly get less fun over time. I’m sorry to have missed it, but ensuring I miss this is 100% the necessary move.

Get in gamer, we’re going to MagicCon Las Vegas for the 2,000 person, $100,000 prize pool limited open. I mean, no, I’m probably not because I would have to learn the set, but this is still awesome and I’m actually tempted.

Quite the Twitch. They came out with these guidelines, before quickly backtracking the face of a widespread creator revolt.

Zach Bussey (quoting since-walked-backk guidelines): Twitch has new Branded Content Guidelines.

- On-stream logos are limited to 3% of screen size.

- Burned-in video Ads are NOT allowed.

- Burned-in Display Ads are NOT allowed.

- Burned-in Audio Ads are NOT allowed.

What is allowed is: - Channel Page Panel ads. - Products in the background of your stream. - Links in chat. - Discussing/unboxing products - Playing sponsored games.

(Also a list of what types of branded content are banned, the usual stuff.)

Brian Kibler (streamer, used as example): Uhh what? Between this and the sub split reduction it might be time to consider YouTube streaming. And this is just on an individual stream level. Can a tournament or league really not run embedded ads or sponsor graphics anymore? @Twitch you can't be serious.

The more I think about this the more insane it is. Literally no event can run on Twitch under these guidelines. Tournament, league, charity, etc. All of them violate these in so many ways. It's just an incredible failure of understanding how people use the platform.

Zack Bussey (later): Twitch responded to their policy change about 5 hours later after witnessing the community reaction. They say they missed the mark, and apologize for the confusion and frustration. They are going to rewrite the policy.

Twitch: Today’s branded content policy was overly broad. This created confusion and frustration, and we apologize for that. We did not intend to limit streamers’ ability to enter into direct relationships with sponsors, and we understand that this is an important part of how streamers earn revenue. We wanted to clarify our existing ads policy that was intended to prohibit third party ad networks from selling burned in video and display ads on Twitch, which is consistent with other services. We missed the mark with the policy language and will rewrite the guidelines to be clearer. [Apology and promise to update].

It is very difficult to get things right on the first try. It’s not this hard. If you want to make it clear you don’t allow third party ad networks, you can say up front ‘this is a clarification of our existing policy that we don’t allow third party ad networks, if this is functionally doing other things that are a problem please let us know and we will fix it.’ If you want to actually put on much harsher restrictions and then you quickly realize you’d face a revolt, so you pretend that all you were doing was clarifying your existing position… that’s also a thing one can do.

Yes, Twitch will probably mostly be fine, but they sustained serious damage from failing to get this right on the first try. There are constantly examples of people who very much have the clear ability and the strong incentive to get something right on the first try, yet they fail. Then in hindsight the right way seems obvious.

Also, this introduced to me the Frog of Shame, whose conditions were very much met, and a concept I definitely endorse.

Twitch also has an alignment problem regarding its metrics. Streamers do not benefit from having lots of performance metrics forcibly shoved into their face.

Jorbs: something that gets lost a bit in interface design is that telling people more information can be bad. I turn off almost every interface element on twitch that I can turn off, and would turn off more if I could. I don't need numbers shoved in my face while I'm trying to work.

like imagine if you were coding on a project, and part of your screen was dedicated to a number which showed how many people had bought the project that day, and you couldn't turn it off, and also there were 20 other numbers on your screen like that. why?

also sometimes there's stuff that actively feels bad being told to you. a big red down arrow on a number because it's worse than it was yesterday. a number which communicates that your ex is streaming the same game as you. a number which says you're behind on rent for the month.

so many of the numbers are like "compare yourself to other people". what if I don't want to spend energy doing that? what if i find that stressful? or "this will help you grow your channel". what if i don't want to do? why is your generic UI trying to be my producer and manager?

When I do stream, I go further and don’t look at the interface at all. That is not why jorbs is a successful streamer and I am not, it does however make my experience much better. I do want to know how many viewers I have so I know what drives people away, and when one of them subscribes or pays, because I want to thank them - sub hype and all that - but otherwise leave me alone.

Chance Kornuth speaks truth in call for ban on poker player Martin Kabrhel, former Magic professional Tom Martell helps explain. Was Martin Kabrhel cheating at the WSOP? We don’t know. I was convinced by a Doug Polk video that it is unlikely he cheated. Except via what we do know, which is that he was intentionally making everyone else think he was cheating.

Which is cheating. This creates a distraction that generates an unfair advantage, also makes the experience of playing miserable for everyone involved, as did his other antics.

As a former professional Magic player from back in the days cheating was a big deal, I can strongly attest that it is a huge advantage if your opponent is spending a substantial portion of their effort ensuring you don’t cheat. If you intentionally abuse that, claiming or implying you are cheating, that itself is cheating. It deserves a ban. I wish we had that rule back in the day.

College football FBS teams often use larger personnel on offense in short yardage situations. The few teams that sometimes do this outside of short yardage situations don’t benefit from doing it in short yardage situations, but they also aren’t harmed by it. Other teams do actively worse and would be better off running their normal offense every time. Or at least, that’s what the basic stats say. I would want to check to ensure the samples represent similar situations, since more difficult plays and especially goal-line stands potentially involve more big packages.

I am not typically one for cage matches. I will always be one for dumb money.

Peter Wildeford: Pollsters ask the important questions of our time.

This seems like an excellent data point for exactly how much of a hell of a drug is partisanship. Also helps explain why the odds are as close as they are. If the match does happen, Zuckerberg isn’t 100% to win, but it could not be more obvious he is the right side. I shudder to think what Kelly would tell me to bet if I end up still laying less than 3:1.

As expected, I spent my month of gaming on Final Fantasy 5, where I’m near the end of world two, and primarily on Diablo 4. I do regret moving in on Diablo 4 so quickly, as the first quality of life improvements are a massive upgrade, something like 80% more experience in the core endgame loop while reducing required travel time. That’s hardcore punishment for not waiting.

Mostly my summary of the game is, this is More Diablo. If you want More Diablo, this game is for you. It offers much better character customization and interesting decision making than Diablo 3, and it is extremely well polished to keep you playing and doing things that feel satisfying if you find Diablo games satisfying. The plot is fine, about what you would expect at this point. The side quest writing is solid.

The big issue is that Diablo 4 does not much innovate, or offer that much variety of things to do. You also get remarkably few upgrades to gear, because a few key features are more important than everything else, so you often get stuck with equipment for a very, very long time. This can cause the auto-leveling to be quite punishing if you are unusually weak for your level, although this issue goes away once you approach the endgame.

The endgame content loops are for now pretty basic. The nightmare dungeons are the World Atlas and maps from Path of Exile, except without anything to motivate or tie them together. The Helltide and World Tree game loops are undifferentiated pointers to more of the usual stuff. I mean, it’s fine, you kill things, you get loot. In general the game needs to do a better job of tying its good rewards to its more fun gameplay loops, and finding more good gameplay loops.

The level of bugs and glitches is fine for normal mode game play. If you go hardcore, expect them to kill your character at some point. There are at least two times that I was killed in ways that were completely unavoidable bugs, which I verified online.

Definitely go straight to World 2 when you start. I recommend playing the game in ‘no spoilers’ mode, in terms of figuring out character abilities and other mechanics for yourself, the game is highly forgiving on letting you undo mistakes and run experiments. Once you complete the main quest and enter the endgame, if you decide to continue, you’ll want to look up where to find the Alters of Lilith, but the game is more fun if you avoid doing that until the end.

As with all such games, one can think of it as two distinct experiences. Leveling up and going through the plot is cool. Playing the endgame, which will constitute most hours put into such games in the end, is less obviously cool or interesting. Know when to quit, also I will soon need to take my own advice.

No Content For You

A lot of what you see these days is not for you.

Often that is because the business models is focused on a few heavy spenders or otherwise important customers called whales. Most of the value and profits come from whales. Everyone else, even if technically profitable and where most of the consumer surplus is generated, is an annoyance. If you are never going to be a whale, you mostly don’t matter.

We are now much better at these types of focused optimizations, and much worse at resisting the temptation to do them.

Emmett Shear: Remember: the hours-and-views-weighted-average-person is very far from the average-person. Casinos are designed for habitual gamblers, not you. Mobile games are designed for whales, not you. (Netflix is designed for the average person. Per-customer vs per-engagement revenue!)

Patrick McKenize: Two words which were seared into my brain by an otherwise mostly forgotten look into the economics of the Japanese cultural industries: “hyperconsuming outliers.” The existence of them is something of a Rosetta Stone for things which are otherwise counterintuitive.

There’s a phrase in media criticism “If you see it, it’s for you” and it needs to be asterisked with “Actually, you could conceivably be just the collateral exposure from an attempt to reach Taro, who will spend 4X your total yearly entertainment budget on this this month.”

(Logic applies also to metered ad models where some people consume 5 minutes of video and one ad a month and some consume 200 hours of video and many, many ads.)

Emmett Shear: “If you see it and it is supported by an economic model where you produce within one standard deviation of the median profit, it’s for you.”

Andrew Rettek: This is a very good reason to prefer consuming things with a flat fee as opposed to per use.

By default, the revenue model determines who things are for, with the caveat that those who don’t raise revenue can also serve other purposes such as marketing and community. The experience you get is deeply intertwined with the ultimate revenue model. Beware when that revenue model is predatory, even if the target of this predation is clearly not you, because they will punish your experience as a result.

In a similar vein, oh no.

John Carmack: AAA game dev often does a lot of analytics around where they lose players, and how “stuck” they get in various places. Amazon could provide similar info for kindle books, and authors could learn a lot from it — where are readers bored, and where are they binging?

The obvious benefit is for future writing, but it is interesting to consider “patching” books. Revising textbooks is common, but tuning fiction is probably also valuable. It feels “wrong” to me, but hard to articulate why.

Nish: Do you think there's an analogy to the kind of metrics that YouTube creators get? Makes me think that this could lead to low-quality, metrics-gaming content, but still helpful to have

John Carmack: YouTube is an even better example than game dev.

Let me try to articulate. Standard Beware Goodhart’s Law situation only more so.

A little of this in a contained way is good. Too much of it, or using it on an ongoing basis, results in generic over-optimized dreck.

For a fiction book to be good let alone great, it has to be unique, it has to follow its vision, and it has to challenge you and be willing to lose you, in order to bring its vision to life. It must also remain unpredictable and surprise you, you need the feeling anything can happen if it could happen in the fictional scenario. And often there is a subtle magic to the details of a book and its world, or even a hidden meaning, that risks being lost. Finally, there is a kind of improvisational wonder, or historical artifact, or knowing of the mind of another, aspect to the whole process. Oh, and there is the pressure to censor, or simply to tinker.

As a result, we have a very strong sacredness taboo on books. Once it is published, it does not get revised. We can then bond over this sacred text, compare our notes, cherish it forever. And we have seen examples of people who modify books of others, including posthumously, in ways that would have the author rolling in their grave and which destroy much of what makes the book special.

At one point, in one of his books, Nabokov says ‘at this point, dear reader, you are ready to throw this book out the window.’ Which my father reports was exactly what he was thinking at the time. That is an amazing moment.

All of that will get ground into dust if there is pressure at the micro level to optimize for how many people keep reading. It would be the nightmare version of Han Shot First.

It also confuses costs and benefits. Reading a book is a cost - a good reader knows to stop reading.

You still need an editor (cough George RR Martin cough) who will fight you on everything from individual words to general bloat, and make you better. And during the writing process, some amount of this type of feedback is super valuable, especially on a macro scale, allowing you to spot stupid mistakes.

This applies less to a non-fiction book or textbook, which already are often revised, and whose components more stand on their own. You still risk trying to solve the wrong problem using the wrong targets and the wrong model of what matters, your mistakes will not cancel out.

You can of course say, oh, you are free to ignore all that if you like. Maybe this is true, especially if there are no feedback mechanisms in the metric - if no one is going to feed pages read into an algorithm for book recommendations, if the revenue model is a flat fee, there will be far less pressure.

Yet despite this, we see more and more games that focus on keeping players playing, the same way Netflix wants to compete with sleep and win. We have run this experiment, and that is the result.

Another way to look at this is to ask, once again, who and what is the book for?

If you are modifying your work based on when people stop reading, then your customer is the marginal person who might stop reading, and your book’s purpose is to be optimized to stop that from happening. That is not going to result in great literature.

Productive Productivity Discussion

Tyler Cowen notes productivity has been down for five consecutive quarters, which explains how wages can be down and yet the labor market robust. You need more workers to do the same work, so you need more workers, but you can’t afford to pay them more when they are producing less.

He considers work-from-home an insufficient explanation for the changes, but the changes don’t seem so big that the explanation obviously fails us. This is especially true if your model of work-from-home is that it is a short term dramatic increase in wages (commuting and going into an office really suck) and also neutral-to-good for existing team short term productivity while everyone is on the same cultural page, everyone knows what needs to be done and expectations are that everyone actually works.

Whereas over time, as many have theorized and I strongly believe is true, you’re spending down your cultural capital, new employees aren’t integrated properly, team cohesion declines, and people slowly slack off more. That all fits the data, and the trend to increasingly try to force workers back in the office.

Tyler Cowen asks why a deal isn’t struck for higher pay and higher productivity. My answer is that this isn’t an efficient deal. The lower productivity level is mostly efficient, people are taking large experiential wins and effective pay increases, they don’t want to give that up for a few more dollars. Where is the trade to be had?

Arnold Kling correctly sees work from home as a massive human welfare improvement, irrespective of the measurements from the GDP factory. He suggests that the blip in productivity could be merely a measurement error, or due to workforce compositional effects. I’d also check if commute times are being handled properly, presume they aren’t, and see how that impacts the calculations.

Or:

Paul Graham: I've talked to multiple founders recently who have changed their minds about remote work and are trying to get people back to the office. I doubt things will go all the way back to the way they were before Covid, but it looks like they will go most of the way back.

Cate Hall: “I’ve talked to multiple founders recently who have changed their minds about having raised employees’ salaries and are trying to lower them again. I doubt salaries will go all the way back down to where they were, but it looks like they will go most of the way back.”

Paul Graham [continuing thread]: Why were all these smart people fooled? Partly I think because remote work does work initially, if you start with a system already healthy from in-person work. (It's like communism in that respect.) And partly because it seemed to solve recruiting, which is always a bottleneck.

There will definitely continue to be remote-first companies. There were before Covid. It works for some businesses. And there will be types of jobs, like customer service, that will commonly be done remotely. But remote-first won't be the default.

People will vote with their feet, or their refusal to use them. We will see what happens. I think we are doing fine in this particular trade-off, in a way that isn’t showing up in the productivity statistics.

Selection effects may throw a wrench in this. If you are a true superstar and want to perform like one, or otherwise care deeply about the company’s or product’s success, you are more likely to actively embrace a return to office. If you need to be hiring superstars, many of the superstars will be willing to do whatever it takes. They really do want to do more work, and take their pay in money and in equity and career building.

Paul Rosania: You will get a lot of pushback on this with claims about labor dynamics and worker preferences, but crucially, many of the highest performers *want* to spend time in close proximity to each other, so returning to an office might be a net win on talent quality.

Paul Graham: Interesting point.

Paul Rosania: Optimizing for labor market size is for suckers. I don’t want to work with a big team, I want to work with a talented (and dedicated) one.

Flo Crivello: Have definitely noticed this the moment we started hiring only in SF. 1 or 2 standard deviations shift in the drive level of our candidates. Remote candidate: “I’m really looking for a remote job because I wanna maximize time with my dog” (true story)

SF candidate: “I don’t care about the poop, I don’t care about the needles — something crazy’s going on over there and I’ll do anything to be a part of it.” (Also true story)

The ideal would be to not impose too much deadweight loss in order to solve this sorting problem or competition. The needles might filter out the unmotivated, it would still be better to not have our best performers paying extra to step around needles. It is still better to pay a high price and assemble the highest performing possible teams, rather than not doing that.

We really do pay a much higher price than we realize to align incentives, solve principle-agent problems, sort for the best people and keep everyone on the same cultural page. Then we pay additional huge alignment taxes to enforce norms and relative social status, to ensure everyone is conforming and thinking the right ways. There is, as Adam Smith put it, ‘a lot of ruin in a nation.’ There is an even bigger amount of ruin, surplus to burn off in various ways, in almost any productive activity.

That is why I am increasingly optimistic about our ability to pay a high ‘alignment tax’ on our AIs. We pay a huge alignment tax on humans for highly imperfect results, because the alternative is too dangerous even locally to do otherwise. Have you seen what teams can do when they can truly set such worries aside? The sky’s the limit.

Thus, there’s every economic justification for paying extremely high alignment taxes on AIs, purely to control local self-interested dangers. That still isn’t as much as we should be willing to pay, the externalities involved and lack of understanding of the dangers is still quite likely to get us killed, but if you say ‘this alignment solution will add an order of magnitude to costs’ I do not treat that as a dealbreaker anymore.

Think Like an Economist

Insurance for home and auto is becoming difficult to find in California, which is unsurprising given rate increases have been rejected for three years. It turns out that they only sell you insurance when they think it makes them money.

California electricity prices to vary based on your income. The electricity company gets to know your income. This is going to be so enraging. Isn’t this what the taxes are for? Why not simply charge people progressive tax rates?

Somehow James Medlock is continuing to get the best bets on the planet, where you can hedge out at a single digit percent of your risk and the rest is pure profit. Turns out if you dare the hyper-inflationist advocates to bet on their beliefs they don’t understand or don’t care that they can trade in the market. What a gig.

Kenya signs trade pact with EU, other East African Community members invited to join. Most promising EU move I’ve seen in a while. How much the EU then works with its regulatory arms to cripple the resulting activity is an open question.

A simple model of taking down Binance: When the ~100th biggest financial institution is a brazen criminal enterprise, either you crack down on it, or your regulatory regime fails. The choice was clear.

More historical grounding to Mint the Coin.

Presidential Trivia: Woodrow Wilson’s portrait is on the $100,000 bill, the highest denomination ever issued by the U.S. government. Printed in 1934, they were used for transfers between branches of the Federal Reserve. It’s illegal for individuals to own them.

Joe Weisenthal: You're telling me large denomination money has been used in this country, strictly for the purpose of moving one part of the government to another? Just a little something to remember in 2025.

Is college actually getting more expensive? This is less clear cut than you might think. Private college tuition is up, but private average tuition paid on net might even be slightly down. Price discrimination, featuring 100% marginal tax rates on your entire family and its assets, is the order of the day, unless you want to downgrade to a place that needs you more than you need them.

What’s the move here, though, if this is true? What’s the point of ‘saving for college’ if you’re about to face a 100% asset tax? So you can either go big or stay home.

Two places in Switzerland raise minimum wage to about $25/hour, the city of Zurich and the canton of Winterthur. A quick Bing Chat inquiry said this will only bind for about 2% of Zurich jobs, so this might not be as crazy high as it sounds.

Is a 4% inflation target now proven to be politically unsustainable? We know that large amounts of sustained or unexpected high inflation, noticeable large price hikes, are deeply unpopular. A sustained 4% inflation rate could potentially be accepted, but any misses on the high side would be quite unpopular. If people are willing to tolerate up to about 4%, then you can’t target 4%. I essentially agree that to the extent we have a target 2% is about right, except I would advocate for switching to targeting the level of NGDP.

Tyler Cowen asks: Should Nairobi burn elephant tusks it confiscates to indicate they are sacred, or should they sell in the open market to lower the price and thus the incentive to poach? My instinct says at a minimum hang onto the confiscated tusks, perhaps displaying them in some sort of museum as a tourist attraction, so you have the ability to lower or crash the market at any time, thus lowering prices now. If I had to choose, I’d sell on the open market. Yes, I see the long run argument for trying to create sacredness, but even if it works people collect sacred things and there will always be people willing to break the principles involved. I don’t see it.

Crypto bot borrows $200 million in a flash loan to secure $3.24 in profits.

Joe Wiesenthal: If you're not willing to borrow $200 million in order to secure a $3 profit, well I got news for you, there's an AI bot ready to take your job.

One must understand, no. This is dumb. Something has gone horribly wrong. There’s picking up pennies in front of a steamroller but this is ridiculous. If you use $200 million to secure a $3 profit over a series of transactions, you are going to end up broke. At some point the result will not be what you expected.

That does not mean the bot has gone horribly wrong. The $200 million does not belong to the bot. It is borrowed. Who in this chain of agents is taking on the inevitable calamity risk is not clear. It still has to be someone.

Will stores be able to sustain 30-day no-questions-asked return policies? Patrick McKenzie points out that this is an excellent policy when applied to almost all people. Mostly the cost of returns is well worth it for the goodwill, even efficient. Except there is a fat tail of anti-social behavior where people will treat such policies as an invitation to maximize - buying a TV to watch the Super Bowl and returning it, or returning 99% of the clothes they order.

What is needed is some form of ‘show that you are one of the people who generally acts reasonably,’ and that you have what this EconTalk episode discusses, a concept introduced by Lord Moulton: Obedience to the Unenforceable. Over time, people are getting more efficient at exploiting such weaknesses, and are less willing to obey the unenforceable, both in terms of techniques used and willingness to use them. AI is the extreme version of this. AI will obey rules to the letter, but the common sense middle will fall away.

Snowing Pine offers (via ACX) an overview of trucking from a while back. The first thing that struck me is that we massively subsidize roads, damage to which is almost entirely due to trucks. We don’t similarly subsidize railroads, leading to widespread abandonment of them. If we were serious about our green goals, we would fix this.

From a while back: Why do so many prices end in 99 cents?

The Lighter Side

Not the answer I was expecting.

The weightlifting forum thread about how many days there are in a week. So much better than you expect.

Twitter community notes continues to overperform.

(Text of above): Reuters: Exclusive: Aspartame, one of the world's most common artificial sweeteners, is set to be declared a possible carcinogen next month by a leading global health body, pitting it against the food industry and regulators.

Readers added context they thought people might want to know: Group 2B is the IARC's lowest-risk classification of carcinogens. This group also includes aloe vera & pickled vegetables; higher-risk groups include red meat, hot drinks, & being a hairdresser. To face increased risk, an adult would have to drink 12 to 36 cans of soda daily.

Usually with emergent order I can at least guess what the original planner was thinking. Not always.

Last month’s diagram, expanded.

An angle I had not considered.

Nate Silver: Sort of ironic that Vegas is going to win this series 5-2, 7-2, 3-2, 8-3, basically all the worst possible poker hands.

An angle someone else had not considered.

It is still previously there (original was posted Nov 27, 2022).