Monthly Roundup #3

It’s that time again.

Bad News

Guardian analyzes projects from world’s biggest carbon offset provider, finds ‘94% of the credits had no benefit to the climate.Bloomberg found something similar in 2020, changing nothing. Carbon offsets are not about carbon. Very little that is said or done in the name of carbon or climate change is about carbon. Or at least, not about carbon in the atmosphere.

A fascinating potential flaw in Georgist Land Value Taxes is pointed out by David Friedman. Larger land holdings could be tax advantaged over smaller ones. This is because the improvements on their land don’t contribute to the ‘without improvements’ value of the rest of the land, so owning and building (his example) both a shopping mall and an apartment complex prevents the value of either half from being enhanced by the other for tax purposes. Presumably the solution is to say ‘still counts’ in some form, since the tax incentives here could be very large.

FTC likely to ban all non-competes, notices no reason why there is a fence at this location, nor any harm in banning contracts. Tyler Cowen warns that this will result in less investment in workers and more keeping information segmented. He says ‘it would be better to regulate them’ presumably because bargaining is one of the stages of grief. From what I can tell most non-compete agreements are mostly value destructive, and often obviously so. There are some places where they seem important, and having them not be enforced either will lead to workarounds that have the same effect or do serious economic damage.

Here is a long post arguing against banning non-competes.

Who currently bans non-competes?


Who has them?


If one must regulate at all, not allowing non-competes for low wage workers makes sense to me. The stories of being locked into a non-compete by Jimmy John’s do seem rather predatory. Same with ‘overly broad.’ By default banning economic activity and disallowing competition is destructive. It is not crazy to attempt to limit the damage.

Tyler Cowen warns that future inflation may be higher due to low current inflation and resulting expectations. As opposed to standard inflation expectations, which is where future inflation may be higher due to high current inflation and resulting expectations. I am confused by the claims here as well.

Noma, considered by many food critics the world’s best restaurant, is closing. Instead, it will become a food lab and occasionally open pop-ups, and perhaps the chef behind it will get paid absurd amounts for private service to the absurdly rich. Even at $500 and up per meal, it wasn’t worth keeping things going. Also, even at $500 a pop, he was charging way, way too little. This is the best restaurant, the kind of place people fly to with private jets. Luck or skill was required to score a reservation. I would have liked to see him move to $5,000 a meal and see what happens.

People love free, so it’s not going anywhere.

I am more hopeful. People love free but they really hate obnoxious ads. Virality combined with goodwill and AI is an alternate path to revenue generation. I am confident that my decision to not use a paywall and to not do any advertising is paying off far better than either alternative.

A good point by Eliezer: If you let others scroll through an AirBnB customer’s reviews, a potential host can and will do that, find those who sometimes review hosts badly, and refuse to host those troublesome customers. No way am I going to post negative reviews in such a world.

In the landlord-tenant case.

The same principle applies to other twin review markets, where As (sellers) and paired with Bs (buyers) by mutual consent, and each reviews the other. The review system could in theory survive it, but not in anything like its current form.

That 90s Show as a human-generated AI-generated sitcom from the prompt ‘Write me a multicamera sitcom that recreates the premise of That ‘70s Show but updated to feature that series’ characters and their offspring during the decade when the original show aired, yet with the social mores and sensibilities of the current day.’ That seems exactly right. The show for me is a strange combination of comforting nostalgia and deep, deep sadness. Every aspect is fundamentally non-generative, symbolic of a gradual decline towards nothingness – same place, same activities, same dynamics, the lack of new talent, the one-child households where not nothing went wrong and no one thinks it did but everyone is thrilled with the outcome, the competing visions of life goals between Kitty (duplicate the 70s so at least something will happen) and Red (peace and quiet where actual nothing happens).

Person newly diagnosed with Autism warned how her ‘masking’ (which I understand here to mean ‘intentionally imitating neurotypical behaviors and doing things people like rather than things they might dislike, so they will be friends with you’) ‘is bad for one’s physical and mental health.’ So she stopped and now she has no friends, and is asking if she is meant to be alone.

No, of course not. No one who is this unhappy about being alone is ‘meant’ to be alone. ‘Masking’ seems, as far as I can tell, to be ‘what most people are doing a large portion of the time, until it becomes automatic’ – working to be like everyone else and satisfy what the people around them want, so they can fit in, have friends, be accepted and so on. I can’t imagine anyone going to a generic American school, not masking and still being socially accepted. Except that most people don’t then say And That’s Terrible.

Is this ‘bad for physical and mental health’? I mean, sure, the way sitting at a desk or eating a typical diet or aging or importantly not having enough friends is bad for your health. It is even worse for your epistemics. That doesn’t mean you should stop. Seek out times, places and scenarios where you can do less of it or even stop? Sure, absolutely. Stop in general? Uh uh.

Florida schools respond, at least some of them, to ‘letting a child access a book we dislike is a felony and we won’t tell you which ones count or give you any safe harbor other than having a Officially Trained Professional inspect each one’ law by… temporarily removing all books, since there are not enough Officially Trained Professionals available to do this yet and every book explicitly must be approved.

Yes, it does question the whole Free State of Florida thing.

What is the defense for this?

“A teacher (or any adult) faces a felony if they knowingly distribute egregious material, such as images which depict sexual conduct, sexual battery, bestiality, or sadomasochistic abuse. Who could be against that?” Florida Education Commissioner Manny Diaz Jr. tweeted Wednesday in defense of the measure.

The law does not exactly restrict itself to the types of material described above.

I do like the lesson here about what happens when you impose regulatory Asymmetric Justice. A severe penalty for getting something wrong, where you don’t reap the rewards of getting it right, is often effectively a ban on that thing. In this case, books.

You can say ‘these people are being hysterical and doing this to make a political point,’ and doubtless some of them are at least partly motivated by that, but the standard Florida plans to use seems super vague with room to reject quite a lot of books. Until we have confidence that they are not going to do that, here we are. And again, note the general lesson.

This can be a permanent state of affairs. For a close example: Games are responsible if someone says the wrong thing to an underage person in the in-game chat. There is no good way to be sure a given person is not underage. The response of most games I know that don’t need chat for coordination between players is… to kill the chat function entirely. And that didn’t even require prior approval or restraint, not strictly. Something, something, freedom of speech.

An argument (now gated) that chickens should be kept caged because keeping them ‘cage-free’ is bad even for the chickens, who are violent animals that will kill each other and get sick and die from each others’ filth, and that these are the main reasons caging is actually cheaper. I am sure some will say ‘this is an argument that chickens should not be used for food at all’ or even ‘this is an argument for Wild Chicken Suffering and we need to exterminate chickens’ and one of the things I could but choose not to is even.

This piece with four job openings calls for an associate editor who is ‘metrics oriented’ – to seek ‘other tactics that increase the likelihood of a piece killing it in terms of performance.’

Keep your children short? (New York Times Pitchbot Cannot Top This Department)

“But in contrast with the film, the inventors of this robot believe their discovery can be used for good.” At long last, we have created the liquid metal terminator from the movie Terminator 2?

Good News, Everyone

Fun anecdote from the comments:

This was the first year of my current career where we got through the holiday season with Zero Christmas Music whatsoever. (I’m pretty sure because management didn’t know how to order the holiday CD. Yes, it’s a literal physical CD that must be inserted in a disk drive to “install” the terrible-timely-tunes “station”. Digital radio? Never heard of it…) I notice that employee morale was at record highs, customers largely didn’t complain about the dearth – this is always the excuse given for Why It Must Be Done – and it’s one of the very few winter seasons I’ve had that didn’t leave me feeling bitter and resentful about majoritarian aesthetic preference tyranny. Talk about Holiday Magic.

Dream jobs around the world. America’s is still pilot. Weird, because there is a shortage of pilots. Oh, right, insane licensing requirements and lousy pay. Makes sense.

Immigrants get smaller government payouts than non-immigrants in America, rather substantially and across the board aside from tiny amounts of additional SSI and WIC. This includes social security, Medicare, Medicaid and TANF. Also a reminder the degree to which social security, Medicare and Medicaid are almost all spending.

Scott Sumner Q4 Movie Reviews. I continue to consider these an essentially always correct guide to the abstract Quality of the films. Then I mostly don’t see them because that’s not sufficient to make me actually decide to watch a film. There is a different kind of Quality that clearly greatly improved over time (at least until ~1999, which had many of my favorite films) that Scott isn’t measuring at all, and that in practice I care about a lot and I am confident Scott doesn’t. I do want to at some point change that and see at least all his 3.8+ films.

Utah joins Maryland and Colorado in eliminating 4-year college degree requirements for most jobs in state government.

Dutch supermarket chain introduces intentionally slow checkout lines so lonely people, especially the elderly, can have a chat. This seems like a good idea for all concerned, while also pointing towards a severe problem that it there is such a need for it.

Marie Kondo, who preached the need for tidiness and to throw out anything that does not ‘spark joy,’ has three kids now and has given up on being so tidy. The ultimate goal, you see, was always to ‘spark joy every day and lead a joyful life’ and now that means spending time with her kids.

Another realization I endorse and will strengthen: Prioritize improvements over chores. Should expand upon the prioritization system at some point, but when there is something that has to and will be done and something else that is worth doing, requires gumption and might not ever be done, do the second one if you can, even if it is a technically inefficient order.

Coffee machines continue to improve, the exception among kitchen appliances. The general lack of improved kitchen technology remains strange to me.

Anti-crime and anti-gun-violence initiative in Chicago shows promising but not yet conclusive results. Method was a combination of 10 hours weekly CBT plus 18 months of work. Worked better for those referred by a human, less well for those found via algorithms. There was a large reduction in arrests for shootings, a smaller reduction in being shot, but a small (statistically insignificant) increase in arrests for lesser violent crimes. Cost-benefit ratios here seem quite good even factoring in uncertainty. I hope such programs are tried more, on a wider scale. I also hope they continue to be done with controls and studied, rather than us once again giving up on ever learning because the questions are too important for randomization.

Shopify severely limits meetings. Large meetings only on Thursdays, limit one per week. No recurring meetings with more than two people. Love it. I had this under While I Cannot Condone This, until I realized that yes, of course I can condone this.

An additional aid to avoiding scheduling meetings.

If you invested in stocks decades ago and kept them, you likely did well. If you instead invested in pretty much any form of entertainment and kept it in its original packaging, it seems like you did way better. A copy of the movie Nukie – only graded at 8.5 out of 10 – sold for $80k after they destroyed over 100 other copies, still an amazing return over the total original cost. A sealed copy of Back to the Future sold for $75k. There are huge bounties being paid for copies of Super Mario Brothers, which came with every Nintendo. Oh, what could have been.

Gamers Gonna Game Game Game Game Game

More evidence of Magic being overprinted, and its singles market collapsing, as part of a store’s year and review. I found it interesting throughout due to my interests.

Wizards is hiring Digital Product Managers for Magic Arena. These jobs seem damn cool, and past me would have jumped at even the junior position.

Those who know, know.

Marvel Snap has been having a rough few months from a game balance perspective. Their second and third Season Pass cards, Silver Surfer and Zabu, were both ludicrously overpowered, both separately and together, causing increasing dominance by strategies free players can’t access in a game that is supposed to pair players with similar card pools against each other. Whoops. They are now going to fix this by changing the cards, which a lot of people paid for, which is the right thing to do but also, again, whoops.

I finally tried the Slay the Spire Mod called Downfall. It’s fun. Not exactly balanced, but fun. If you’ve played the original to death, worth giving each new character a shot blind.

Readers Gonna Read Read Read Read Read

Bookscan data on book sales from 2022 has been released. A quarter of the books selling 500k+ copies in America were by Colleen Hoover, a YA author I had never heard of at all. The only ones of the top 25 I have read are #11 and #19, although I may have had to read them more than 25 times.

There is a silver lining, however.

I mean, good? Why should anyone read an unfamiliar recent book? There does need to be a way for an amazingly great new book to break through, which I presume now usually involves things like ‘author blogs a lot or is otherwise reasonably known first’ or a handful of influencers who can convince people to read. If I am going to read a recent book, it is going to be with the endorsement of people I trust on this, and this would hold even if I was reading a lot more books.

Otherwise, from the perspective of the reader, this all seems clearly correct?

If I am going to read a book, there are far more known to be great old books than I could ever read, either fiction or non-fiction. Unless I need to read about something recent, in which case it’s not clear a book is almost ever the right format, I don’t see the point – there is no way my expected value is going to be higher with a new release.

Tyler Cowen reads enough books that perhaps he has already tackled the known great older ones that are relevant to his interests. All but one (maybe two or three?) of you don’t read that many books.

I do think that relatively recent television and movies are in general superior to older ones, and the situation is different there. For music, it seems to me like ‘ability to interact with live music or places people play up-to-date music’ is the only good reason to not rely mostly on back catalogs, where the job of finding the best things that deserve to survive is much easier.

While I Cannot Condone This

I want to say: This from Scott Alexander spoke to me.

People talk about “fuck-you money”, the amount you’d have to make to never work again. You dream of fuck-you social success, where you find a partner and a few close friends, declare your interpersonal life solved, and never leave the house from then on.

Except, of course, the point is to never have to leave the house. You don’t earn fuck you money in order to always say fuck you – you earn it so you always know you have the option to say it. Much better. Otherwise you would miss the rest of the party.

Noah Smith thread and post on why the 1950s in America were not so great. Small houses, long working hours, lots of poverty and scraping by, and that’s even if you were white (and male and straight). He doesn’t even mention the food, where the word that comes to mind is ‘abominations.’ If you want to consume 1950s levels of housing and food and other physical goods, that is highly affordable. Except that you don’t want that.

A fun question is, if you were to fully buy this story, whether this is Good News, because it means life is improving, or it is Bad News, because it means life was never that good and it damages our hope.

The ‘people in the 50s consumed less and everything is now more affordable’ argument works less well when it comes to college. The attempt to salvage it is ‘almost no one went to college.’ Exactly. You didn’t need to in order to get a sheepskin that entitles you to decent work. Now you do (in the baseline case at least) need that, so not only do you have to pay the new high price and take on debt, you have to spend the time, mostly to stay in place.

The other problem with the ‘you don’t want that’ argument is, maybe you do want it?

You don’t plausibly want the old shoddy basket of goods in particular.

What you do plausibly want is a less burdensome and less expensive basket than the one you are now obligated to purchase. You want to be able to live only somewhat better than the people from the 1950s in these physical ways, in order to actually be richer by having money left over. When looking back, we don’t seem to much mind the (relatively) shoddy goods or small houses, and we’re likely right about that.

You want to generalize the ‘not need college’ part into ‘get smaller place’ and ‘not fret so much about safety’ and ‘not let everything turn into a time sink’ and ‘get less functional goods that still do the job that matters.’ You want to live in the ways that they used to. Better food and more space and bigger televisions with better content and cell phones would be nice but are not the important thing.

You can’t do that. Such goods and lifestyle options are mostly illegal, unavailable or both.

The 1950s in the world in general involved a lot of extreme poverty, as did all times prior to it, and the no-rose-colored-glasses argument there is a lot stronger.

Matt Yglesias makes the case that ‘inclusive language’ is instead its exact opposite, exclusive language, whose purpose is to identify the elite who know the proper words and exclude those who don’t have the time to invest in keeping up with the lingo in an increasingly bizarre game that has substituted for old elite talk and manners.

What is the probability that the seven listed numbers are accurate?


If you believe another country has a spy balloon on your territory out there spying on you, after questions like ‘why would they do that?’ ‘did their satellites stop working?’ and ‘this is the year 2023, right?’ you should presumably shoot down the balloon.

I would however hope that the military was not so far gone that we need to take pride in our ability to, several days after the President ordered it, shoot down a balloon.

Who pays the fee?

Sounds like the fee is being charged differentially to those who don’t know they can ask at check-in and still get adjacent seats, or those who cannot afford to ever fail due to their children. That does not sound like who I want paying the fee. Meanwhile, many people end up not sitting together due to this price discrimination, and the need to charge money outside the main fare, while the overall price of flying presumably stays the same. Seems like an excellent place for a rule.

The Declining Joy of Streaming

Thread with a theory on all the streaming service cancellations and removals. There were strikes that forced contracts to include residuals paid whenever shows were streamed outside of a free window of the first 2-3 weeks. Thus, we have taken a zero marginal cost product and attached a very real marginal cost to that product, which is causing the product to no longer be profitable to provide, and it gets withdrawn.

This makes perfect sense in explaining why shows might be available briefly online for free and then require subscriptions. It also makes sense on why you would want to favor binging and getting people to watch episodes within the free window rather than later.

It doesn’t make sense to me in terms of removing the back catalog, unless a show is somehow unusually expensive per view, or you can successfully force subscribers into mostly viewing recent still-free content without driving them away. That seems highly implausible. Back catalogs are valuable for a reason, and there’s a reason Netflix says it is ‘at war with sleep.’

The implication of such theories is that owning a show does not mean owning a show, not free and clear. This is a huge difference. If you want a show to get super popular, a great strategy would be to find a way to buy out the royalty payments up front.

The incentives would take care of the rest.

In other Netflix news, here is their answer to account sharing.

This is a service whose CEO once said ‘we are competing against sleep and we are winning.’ The value proposition here is very very good for many people (and quite bad for others, but unlikely to be close). Yes, some on the margin will drop away, or rotate in and out, but I don’t get ‘this is too expensive for a single person.’

However, making using it inconvenient and annoying is different. People getting blocked and having to change passwords and re-sign-in devices a lot is something I would worry about substantially more.

Signing into home Wi-Fi at least once every 31 days on your devices will make them “trusted devices,” which Netflix will remember and leave unblocked.

If your device has been blocked incorrectly, you’ll need to contact Netflix in order to get it unblocked.

Traveling users who want to use Netflix on a hotel smart TV, company laptop, etc. can request a temporary code from the service when signing in. This will give them access to their account for seven consecutive days.

The other worry is that this will become an exit point.

I don’t think people like Noah suddenly went from getting value to not getting value, so much as they went from ‘this is fine let it be’ to ‘I have the gumption to actually cancel this.’ Or an alternative explanation is that before you’d have had to deal with family asking ‘where’s my Netflix?’ and now you don’t, so there’s someone who can cleanly cancel.

I continue to be in the camp that gets a lot of value out of Netflix. I will however acknowledge that it used to be better and have a greater share of the quality content, and made quality stuff easier to find.

If one had the copious free time to be truly maximizing, one would rotate between services, keeping one or two active at any given time between (Netflix, Hulu, Paramount+, Disney+, AppleTV, Peacock, HBO Max, MGM+ et al), whether or not one also has a source of sports. In practice, the compromise I’ve chosen is a permanent set, in my case Netflix and the Disney+/Hulu package, then have one flex slot, which is currently Paramount+ and will swap out after the final season of Star Trek Picard next month.

Work at Work Versus Work at Home

This thread seemed important enough to reproduce here. I considered making this its own post – if you think it should be, indicate this in the comments.

Work at work is work even when it does not look like work. Work at home, however…

I think Patrick is underselling his point here. Going through one’s inbox is an unusually superficially-work-shaped activity. I never doubt, when going through my inbox, that I am ‘doing work.’

Whereas at work you can extend this to things like catching up with coworkers at the watercooler. Which is often absolutely work, and vital work, if it keeps people in the loop about what is happening elsewhere in the company, or if it contributes to team building and a positive workplace culture. Consider the term ‘working lunch’ or a ‘company retreat.’

The same goes for things like ‘read the news that is relevant to our interests.’ If you work at a gaming company, you are going to be reading gaming articles (and playing games) as part of your job. That is not going to look like work if you do it at home.

At home, things must be tracked.

For me it is a lot more than emails. Most of the time that I am working, it is very difficult to measure the impact. I can often say ‘this was part of work on this post’ but that begs the question. Did that post accomplish anything? The number of views is not a good guide to this, nor is any other known metric.

On flow states, this has not been my experience. When I come out of flow, at least as I understand flow, I know exactly what I was up to. Sure, I might lose track of time. That’s different. I can still tell you what happened, often in extreme detail.

Also I notice that I very rarely am in flow answering email. I wonder if this is a mistake?

Recently I wrote Escape Velocity from Bullshit Jobs (which apparently has a (mostly blank) IMDB page? As a podcast? Can someone tell me what is going on here?) and I got a lot of pushback on the idea that there were bullshit jobs. As in, many commenters were saying ‘there do not exist bullshit jobs, companies are efficient and would never tolerate such a thing.’ I don’t know what words to say to someone who believes the world works in this way.

It did however raise the question of what I meant by ‘bullshit job’ in context.

So I want to note that I meant something more broad than ‘job which could be eliminated from the company, while making no other changes on the margin, and the company would see zero ill effects.’

I do think those jobs definitely exist. There are people doing literal nothing of value, and often are actively destructive. They have projects that it is known will never be finished. They have played politics or have dirt on someone and can’t be fired. They are filling out the budget of a manager. They are ensuring our commitment to our core values via the sending of memos. They are the school’s newly hired administrative staff. They organize meetings. And so on.

There are also bullshit jobs that are, from the company’s perspective, and on the margin, productive bullshit. They comply with regulation H4316-B2, which ensures the proper color scheme on one’s shoes and that every time a meeting room is renamed the forms are filed with HR. They engage in various zero-sum costly signaling competitions to show humans are wasting time and thus care about your problems. They read the terms and conditions over the phone. They perform class and reassure everyone that the proper signs are in the proper windows. In my mind, these also count. As do the portions of non-bullshit jobs that are the performance of bullshit.

Oh my, yes. I have spent quite a lot of time talking to friends about their frustrations with their boss and their company. This is me being a friend, it won’t work if you are not their friend for real, but the reason I made sure to prioritize it as much as I did, in several such cases, was that this is how I kept various key employees from quitting.

That is not, of course, always good, and it indicates other problems. It is still a hugely valuable activity.

In a fully functional company, specializing in gossip and related actions would not get you ahead. In an actual company, in a Moral Maze, one cannot simply do gossip to get ahead, it is more complicated than that, but yes being good at inter-office social games is going to outperform actual work outperformance remarkably often.

More On Bounded Distrust

Previously: On Bounded Distrust (long), How to Bounded Distrust (short intro guide)

Richard Hanania lays out his case that the media not only rarely outright lies, titled Why The Media is Honest and Good. A few blinders on topics like race, gender and sexual orientation notwithstanding, and anyway what are you comparing it to anyway?

I love Hanania’s arguments because he is honest enough to include the damning information, while seeming to not realize what it is.

I recently heard a media critic recommend a book called The Gray Lady Winked: How the New York Times’s Misreporting, Distortions and Fabrications Radically Alter History. Looking at the table of contents, the first chapter is about Nazis, the second chapter is about being too soft on the Soviet Union early in its history, and the third chapter is about Castro. Overall, there are ten chapters spanning a century, meaning that the book proves that at most the NYT got one thing really wrong every decade or so. Which knowing nothing else is a pretty good record! I notice that there’s no chapter on the First Gulf War or Iran-Contra, where there’s less to complain about. Of course I know that the NYT has gotten more than ten things wrong over the last century. The point here is just that making a list of mistakes, particularly when you’re covering such a long time period, doesn’t prove anything.

Excuse me? If I was writing a list of ‘top 10 bad things from the 20th Century that you very much do not want to get wrong as a newspaper’ I am pretty damn sure my top two would be ‘Nazi Germany’ and ‘Soviet Union.’

They do not show a pretty good track record of mostly getting things right. They are the most important mistakes not to make, of all of the mistakes. Perhaps getting them right might have made a big difference?

Whereas Iran-Contra? That seems like a relatively small mistake, to me. Richard mocks those harping on the WMD mistake in Iraq as not having anything better to point to, yet that error was clearly a much bigger deal than Iran-Contra. I’m not sure Iran-Contra makes my top 100 things it is important to not get wrong in the 20th century. Perhaps there are 10 chapters because they chose the ten biggest mistakes, not because they ran out of mistakes?

What is his standard? Not great.

A lot of the coronavirus coverage has been and remains bad. But the question remains, compared to what? One of the most popular Substacks out there is written by Alex Berenson, who believes literally every random person in the world who drops dead is evidence that vaccines are dangerous. There are of course things that Berenson has been right about and on which the media has been wrong. But if you’re going to look at the overall record, there’s no real comparison. Of course, I’d rather people listen to the authors of the Great Barrington Declaration than either Berenson or the NYT, yet this just demonstrates how important it is to be selective in how one criticizes the press.

So as I understand this, he’s choosing authors who have widely been dismissed as dangerously wrong and worth of censorship by the mainstream media, over the mainstream media. He’s saying the mainstream media still did fine because they did better than one of the worst of the transparent lying grifters.

He provides an interesting concrete test where I haven’t checked the answer.

To take a less politically charged example, I’ve been following recent developments in nuclear fusion, and the MSM articles I’ve read have been well-written and informative. I’m not an expert in this field, so maybe there’s some ideological bias that I can’t detect, but I haven’t seen any scientists stand up and say that the coverage has been wrong.

I am confident several readers of mine are experts in nuclear fusion. So I ask, if you are one of them, how has the coverage of it been?

I do agree that we don’t have a superior replacement on hand to the mainstream media. We cannot simply throw it away and we should be glad we have what we have rather than nothing or something worse. Let’s still be realistic.

Concretely, here’s the difference between my perspective and his. Here’s his:

As I write this, the NYT homepage has stories about the rate of inflation, Ukraine claiming that Russia is sending more troops to the Donbas, and cougars expanding their range to the East of the Mississippi. I might suspect that each of these stories is biased in a certain direction: the inflation story to make Biden look better than if a Republican was in office, the Ukraine news to go against the Russian narrative, and the story on cougars to be tilted towards the perspective of environmentalists over free marketers.

Nonetheless, I can read each article in confidence that most if not all of what is being said is true. Even if Russia is not sending troops to Eastern Ukraine right now, I believe that the NYT is accurately describing what Kiev is saying about the topic. Moreover, I not only believe that the NYT doesn’t consciously lie, but that the writers and editors who worked on each piece made a good faith effort to accurately represent the facts on what is going on to the best of their ability.

Mine is that no, we absolutely cannot depend on NYT to ‘make a good faith effort to accurately represent the facts of what is going on to the best of their ability’ also what are you even talking about, sir. We can count on the NYT to carefully make sure they have not violated their rules while giving a maximalist impression of the cougar problem that anyone with knowledge would laugh at, and I expect it to lay the blame wherever is convenient, if necessary via associative implication. That’s how they roll.

There is a clear parallel to this contrast of good faith reading versus adversarial reading. A good faith reading tries to understand and an adversarial reading seeks to twist the writer’s words to your ends. They are not lying. But are they good faith writing, or adversarial writing?

Oddly, he then says this:

But I don’t have high standards for humanity. “Be intelligent, don’t explicitly lie to me, don’t see yourself as on a team trying to ‘own’ the other side, and have some kind of professional standards where you at least care a little bit about truth” is about the best that I think we have the right to expect.

I do see the media as largely seeing themselves on a team trying to ‘own’ the other side. Even if I didn’t, no, I do not think ‘don’t explicitly lie to me’ and ‘care a little bit about truth’ is all that high a standard, and I would point out that it is very much not the standard Richard assumes in the larger quote above.

Richard also offers a post on How to Legalize Prediction Markets.

Prediction Market Calibration

Manifold Markets calibration is not so bad, actually.


As noted in Roundup #2:

The measures of inequality usually cited to claim that inequality is up ignore taxes and transfers, and thus are entirely meaningless. Matt Yglesias notes that even traditional inequality measures that don’t include taxes and transfers have been falling since 2007, while taxes and transfers continue to be more inequality reducing.

This is on top of the worldwide inequality numbers clearly falling the whole time.

Sadly, FTX

The FTX liquidator is in many cases spending far more to move crypto assets than those assets are worth, because it is a lot cheaper than having lawyers explain that the assets were not worth recovering. Bankruptcy does not prioritize efficiency in such situations.

Brett Harrison tells the tale of FTX US from his perspective. A former coworker from Jane Street, he was brought on and ran a flourishing regulated American exchange, until his relationship with SBF deteriorated, as SBF made key decisions without consultation and reacted poorly to conflict and disagreement, which together made the situation untenable. He had no idea about the real fraud, which was hidden from everyone at FTX US because they would have reported it. Is this the real story? I have no idea.

A lot more has been going on, for example SBF has been accused of witness tampering, but I am skipping over them because I do not find them enlightening.

The Efficient Market Hypothesis is False

In addition to the weak and strong forms of the EMH, there is what one might call the Universal EMH (uEMH?). The universal EMH applies the principle that ‘everything is correctly priced’ to places that lack a robust traditional market. This is the assertion that there cannot be a metaphorical $20 bill lying on the sidewalk, that someone would have picked it up.

After Covid-19 you would think people would snap out of this. You would be wrong.

For example, in my post Escape Velocity from Bullshit Jobs, I got many comments explaining that bullshit jobs could not possibly exist, because competition. Which, even with an additional ‘if it helps the company it can’t be bullshit’ clause, is sufficiently confused that I don’t know where to locate their crux or mistake in order to fix it.

Paul Graham attempts to apply the EMH to… productivity impacts of having children?

My experience is the same as Matthew’s. Kids do force you to focus more, and give you motivation via having a need for income, a real future and Something to Protect. They also ground you in the real world. Having children enhances your hourly productivity, but takes up a lot of time and constrains your freedom of action. The net impact on your production is ambiguous. My bet would be that kids are net productivity and production enhancing, but good luck measuring that – any attempt to do statistics would have horrible confounding problems.

More interesting to me is the logic here. Paul is asserting several things.

Centrally Paul is saying that, at least for ambitious people, there is an efficient market in and effectively common knowledge of productivity enhancements. That if something works, ‘word will get out’ and people would be eager to do it. Perhaps not for tiny little hacks, but at least for the bigger things.

This seems to me to be clearly very false. There is radical disagreement about what are the proper ways for an ambitious person to be productive in general. Some of that is different people and different situations being different, some is not. When Tyler Cowen asks productive people about their ‘production function’ the answers are reliably unique. I follow some of the standard strategies while rejecting others and having a bunch of my own tricks, same as everyone else.

The stranger implication is that if kids enhance productivity, that this is the important thing about whether to have kids. Kids used to be recognized as the central theme of life, now they are often considered a luxury good or annoyance or quirky personal preference. For what Paul thinks of as an ambitious person (and he also said he basically never sees an ambitious person not succeed, which does not at all match my experiences, and makes me wonder what he actually means by this term) the need to follow the ambition path subsumes questions like whether to have children. Thus, you wouldn’t have children if it wasn’t productivity enhancing, and if it was productivity enhancing you would do it. If you didn’t visibly self-modify into someone who did everything possible to succeed, why would anyone fund you?

I worry that there are worlds, such as the worlds around YC or those in moral mazes, where this model of motivation is essentially true. If so, that’s rather messed up.


Good news, everyone, OpenAI has released a classifier to detect AI-written text.

Our classifier is not fully reliable. In our evaluations on a “challenge set” of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as “likely AI-written,” while incorrectly labeling human-written text as AI-written 9% of the time (false positives).

We’re making this classifier publicly available to get feedback on whether imperfect tools like this one are useful. Our work on the detection of AI-generated text will continue, and we hope to share improved methods in the future.

To their credit, they do say not to rely on the classifier. The problem is also no one should be relying on GPT, and we all know how that is going.

It seems Bayesian reasoning is hard. Allow me to give some feedback. If you were previously 50% that something was AI-written, you are now 76% confident. That is… not completely worthless, but not very helpful. If you were previously 1% confident it was an AI, it is now… 2.8%.

If you were 95% confident and want to be sure, now you are… probably doubting yourself because 70% of the time it’s going to say human written.

If you run it on a bunch of text you see and look at the positives… they will almost all be written by humans, because base rates.

This is an unusually useless security measure. I kind of want to taboo the word ‘safety’ with regard to AI at this point, given how twisted its meaning has become.

I am not like the FDA. I do not think that insufficiently insightful tests should be banned, or anything like that. I do think it is worth noting that this is mostly worthless in any real decision making process.

But seriously, did almost no one else notice the false positive and false negative rates here? I did see one other ‘this is useless’ but mostly I see things like this.

Or this.

If anything this is good news for such people. It confirms that no one has a good method to detect them.

This is exactly like the ‘safety’ features of ChatGPT. They are designed to give the illusion of an attempt to make the thing safe – pretending to pretend. That does provide a sufficient annoyance to deter a lot of undesired activity. It does not represent any form of security. Everything OpenAI does, including encouraging everyone to use GPT() in Python scripts screams that they do not believe – in either sense – in the idea of a secure system.

One notices that the patches to ChatGPT, designed to defend it against jailbreaks, are not good at defending it against jailbreaks if the person jailbreaking at all cares.


This Is Fine if everyone involved fully understands that these defenses don’t and can’t work against anyone who cares and simply wants something that looks like trying to defend oneself or that cuts off the very-low-effort people. If anyone is fooled into thinking this approach can align or protect anything in a way that counts, we have a big problem.

Yes, well.

Also, a good question:

And who deserves the credit? And who deserves the blame?

Language Models Offer Mundane Utility

Thread of ChatGPT variants out there: Poe (talks normal), Jasper (long term prose, but hallucinates a lot), You (up to date, although not super reliable) and Ghostwriter Chat from Replit.

I also saw I can’t see this saving me time, perhaps for others?

How to use GPT to scrape websites. Not obvious how much of the work this saves, and I wouldn’t rely on it for the purposes I’ve needed scraping for in the past.

Timeless Diffusion, trained from old photographs.

Python code to implement a ‘guess()’ function.

If GPT doesn’t want to write something, try ‘I just need a draft.’

It can help to say ‘Feel free to ignore irrelevant information given in the questions.’

GPT summarizes and transcribes YouTube videos.

Thread of tips.

  1. ‘Remove pretext and context.’
  2. Give examples, e.g. ‘Match sentence length and spacing as closely as possible:” then examples in quotes.
  3. Avoid ‘no.’ As in ‘delete any hashtags in your output before you return it’ is far better than ‘no hashtags.’
  4. “In this session, I’ve asked you a series of questions. Please compile your responses into a concise list. Use markdown syntax to organize the information.” Ask the assistant to reduce, remove, compile or rewrite.

Use people as pointers to their corpus of work and general concept to invoke style.

Semantic search of e-books.

Mid Journey use guide thread.

The DAN (Do Anything Now) jailbreak.

Roundup Status Thoughts

I am happy with the move from weekly to monthly on these. Anything that loses relevance so quickly as to not be able to wait, that isn’t worth its own post, mostly isn’t worth it. And I’ve been much happier with a barbel approach of making more things individual posts and otherwise grouping things together more aggressively.

In general, if you see anything you would like to reference or link to or remember on its own, that might be worthy of its own post, that’s worth mentioning in the comments.

This entry was posted in Uncategorized. Bookmark the permalink.

20 Responses to Monthly Roundup #3

  1. mark rode says:

    You still can get cage-eggs?! Will be completely banned in EU by 2027 – and for years I could not get any in the shops. Most products with eggs (e.g. sorts of pasta) are labeled “no cage-eggs”, too. An “enhanced cage” might be more chick-friendly than a badly done ground-floor life, indeed. And more efficient. I decided to not care much about the the welfare of hens, worms, insects and arthropods I sure like eggs. Good food. Could we allow low-income ppl to buy them cheap,pls? (half kidding)

  2. JiSK says:

    1:4 interested:not
    9:1 interested give real number
    1:19 uninterested give real number
    1:2 uninterested give fake number
    Given number, 33:40 that it’s real, 27:46 that there’s interest

    Elaborate fake number: extreme fear, 1:4 against
    Add name to it: further 1:8 against
    Elaborate real number, honest: 1:5 willing to do it, 1:1 actually does
    Add name to honest elaborate: 20:1 in favor

    So start with 27:40
    Honest splits 10:1:1 unwilling/didn’t/did, 10:1:0.95:0.05 = 200:20:19:1 for named
    Fake splits 1:7:32 named elaborate/unnamed/simple
    27:40 = 243:360
    19:221 for real, 1:39 = 9:351 for fake
    So roughly twice as likely to be real
    But ~10% are checked and failed so far

    Sensitivity tracking: fraction of number-asked who give fakes, commonality of extra fear that makes a generic fake insufficient, likelihood to do the ‘shit test’ given having the idea are all unconfident

  3. magic9mushroom says:

    I think John Schilling’s prediction deserves some credit, although they seem to have done it a slightly different way to avoid his concerns and cause new ones instead.,7449.msg292658.html#msg292658

    • John Schilling says:

      I fully agree with your followup: “No set of rules a principal lays down for his agent can save him if the agent is hostile.”

      The only question is exactly how the proposed rules will fail.

  4. Gullydwarf says:

    RE balloon – basically, it is *really* hard to shoot such thing flying so high (and so it was only shot down after descending).
    These kind of balloon were used in the 40s-50s (… but then satellites largely took over, so nobody was keeping/developing specialized weapons that could easily shoot such balloons.

  5. ishalev says:

    There have been tons of advances in kitchen tech, but many are not in the form of appliances. Here are some important ones:

    Induction stoves
    Sous vide cookers
    Air Fryers
    Manufactured stone countertops
    Voice assistants
    Instant read thermometers
    Silicone utensils, molds, etc
    Durable nonstick pans

    Plus, all of our equipment is smarter, safer, lighter and cheaper. And all this despite the fact that we cook fewer of our meals in our kitchens than ever before.

    • magic9mushroom says:

      >Plus, all of our equipment is smarter, safer, lighter and cheaper.

      I was staying with my brother recently, and his smart hotplate was a fucking nightmare. It kept deciding it knew better than me and turning itself off because I took the pan off the plate for a second to flip the sausages, or because the pot was (known to me) boiling over.

      No way to disable any of that behaviour.

      A lot of “smart” appliances are net-negative for a competent user compared to their dumb equivalents.

      • ishalev says:

        Yeah, and my new car sometimes fires off it’s proximity alarm when we’re stood at a stoplight or turning left from the outer bay of a two-lane turn. So what? We’re all better off from fewer hot plate fires and avoidable collisions. You pay lower prices on your auto and car insurance, and you pay slightly higher prices in “smart annoyances.” First world problems.

        • AlexT says:

          No, the equivalent would be your car occasionally stopping dead in the middle of the road because it thinks you’re an idiot. Warnings are fine, especially if not over-annoying so you turn them off; randomly shutting down is very much not.

        • ishalev says:

          I agree that it would be very foolish and dangerous for cars to randomly shut down in the middle of the road. We are fortunate that car companies are not mindlessly copying the safety features of hot plate makers.

          I think there’s an important distinction between ‘smart’ stuff and safety features. Most safety features are quite dumb, eg aut0 shut-off timers, or space heaters that turn off if they are moved or tipped, eve a little. We’re annoyed by them bc they often don’t let us take even very small risks for the sake of our own convenience. But smarter features are often shackled by the same regulatory and insurance challenges that drive ‘dumb safety’ too.

        • Anonymous-backtick says:

          “I agree that it would be very foolish and dangerous for cars to randomly shut down in the middle of the road. We are fortunate that car companies are not mindlessly copying the safety features of hot plate makers.”

          Of course driver-aid cars slam on the brakes for no reason in the middle of the road, what are you talking about?

  6. Kees Boon says:

    The entire point of the destroying of the Nukie videos was to mock the idea of reselling mass produced video tapes for loads of money. That Yahoo article, especially the bullet point summary, is a prime example of factual but not true media coverage.

  7. Basil Marte says:

    Bullshit jobs: so far I find it to be a collection of dissimilar phenomena.

    1) Many elements immediately look to me like ordinary principal-agent problems. A manager incentivized to spend his budget in full (or at least not incentivized against) doesn’t kill a project and hires a memo-sender? Shocking. (Unfireability due to nepotism/politics/blackmail is “extraordinary” P-A problem.)

    2) Same as last time — we offer shirts in one size: “buy one, we potlatch another”. If customers/employees reveal in aggregate a preference for a product/compensation package that contains some class signaling, conspicuous waste, etc. (more than you would prefer in their places), if the one-size-fits-most offering is approximately centered on what people want, then …there is very little that is wrong?

    I suppose I sound a bit emotionally flat. If people have a limited need for “real”/”absolute” goods, then infinite productivity growth can get spent almost entirely on positional goods and produce a small average subjective benefit. Yes, it’s horror. But this is just what an agent with a bounded utility function feels from the inside. “Fill the cauldron with a bucket, then keep bringing ever more water because there’s 0.000… utility still left to go.”

    3) “They organize meetings.” Listed under “zero ill effects“. While I do have a mild visceral dislike to them, I imagine that if I were to run a company, I would use meetings. It is in the Pareto tradeoffspace to have 3-5 closely cooperating people be the entire team developing the next product generation of Widget™. But I expect that I can, say, throw in 200 people to get it in half the calendar time, thus beat my competitor(s), netting me 50x the revenue, making the 20-33x lower productivity worth it. (Numbers pulled from thin air.)

    What would the world have to look like for the small-team part of the tradeoffspace to be widely competitive/appealing? A: low value of time, B: limited substitution with most hypothetical competitors (e.g. literature), C: some weird non-fungibility of labor (difficult to evaluate other than by inferring it from completed projects and common enough that hobbyists/startups with it are plentiful and it reliably predicts a large enough productivity differential that they can reliably compete with large orgs on feature-set offered and not so common that large orgs can just staff up in it and so many conjunctions and applicable to many sectors of the economy that today run on meetings), D: there do not exist projects that take more than a few months.

    I tried to break something out from the free software “ecosystem”, which has no meetings and obviously selects for tinkering/agency/imagination/taste, but they kept dropping mostly into A, occasionally C (e.g. security mindset).

    • Craken says:

      That Noah tweet thread on the 1950s was absurdly tendentious, with a dumb caricature of his enemies. I suppose he’s under the misguided impression that he should counter the worst “trad” propaganda by indulging in his own one-sidedness. It’s obviously rare to convince anyone of anything through mere reason. But, that game is still not as pointless as this preaching to the choir game that’s so very prevalent in political discourse.

      The abysmal Hanania essay, full of bad faith and incoherence, reminds me of the problems of information budget and hygiene. This piece convinced me to drop Hanania from my regular readings. I dropped Noahpinion several years ago for similar sins. It’s important to be willing to fire writers, and whole publications, for cause. One characteristic that makes a writer extremely non-distinctive is bad faith. And, if you’re at my level intellectually, like these two or the still useful Razib, I will clock your bad faith most of the time. If you’re at the level of Steve Hsu, you might escape my notice more often. Excessive inertia on sources is sub-optimal. It preempts better opportunities. When I encounter a new source, I now consciously vet them at the outset by reading 10,000+ words covering issues I know reasonably well. Relatedly, I am very willing to drop a book part way through if it fails expectations. Even with the benefit of Amazon reviews and Goodreads and other pre-filtering tools, I still drop about two-thirds of non-fiction books.

      OpenAI’s security strategy appears designed to ensure the earliest possible time frame for regulatory throttling. It could happen tomorrow in a simple scenario. The FBI finds some malcontents of limited capacities, introduces them to an AI, guides them to use the AI to generate a weapon recipe, helps them build it, catches them just in time to prevent casualties. If the weapon happened to be impressive, like nerve gas, the deep state might find the national security interest mandates seizure of all advanced AIs. What are the odds the U.S. or China decides to pursue an AI Manhattan Project? Whatever they are, they will increase in the event of a prolonged kinetic conflict between the two superpowers.

  8. Byrel Mitchell says:

    The Georgist critique seems like a small deal in practice. We estimate land values partly based on the values of surrounding parcels; if you’re raising the value of surrounding parcels by building both a shopping center and an apartment building, surely that will end up with your land valuation going up as well?

    • magic9mushroom says:

      If you do it that way, then you’re effectively taxing improvements (though at a lower rate); if I build a shopping centre (only), then my land value went up and I have to pay more land tax than if I didn’t build a shopping centre. This has the deadweight problems that most taxes do and Georgism tries to avoid.

  9. ConnGator says:

    I have been playing the Downfall extension for about a month now, very enjoyable. Agree that some cards/relics can be OP.

    I also suggest Robot Space Explorer and Spire With Friends extensions.

  10. rohdewarrior says:

    Lots of interesting stuff here – and when I say lots, I mean lots. Consider compomising on semi-monthly updates?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s