Apologies for the delay. It’s been a lot, this month. At the start of it, I had

Chana Messinger has a good post on emotions about money and how it should be spent, if you’ve been following the latest EA internal debate round on it.

We’ve since had some very different discussions about money, prompted by changing events. I don’t have a lot to say here. Some of these people were friends of mine. I wish they hadn’t done it. I hope they’re OK, I hope justice is served. It won’t be. I hope we don’t all die. I wish I had gotten paid1. I want things to be good instead of bad, and the world today has many bad things, and figuring out what to do about them is hard. That’s why I’m an EA: I think we have a good set of intellectual tools.

But speaking of pieces that were better when I wrote them, let’s talk about


Me: Who cares about no-press diplomacy? No-press removes almost all of the difficulty!

Meta AI: Hold my caution and watch this.

Yeah, everyone’s talking about ChatGPT. It’s scary. It’s not the most impressive model I’ve seen, but it’s public and it’s communicating that yes, this is real. Of course, that’s not all. Stable Diffusion is good enough that people doing personal projects won’t necessarily search for art: they can just make it themselves, like in this example. We’re approaching an interesting age. Terrifying, but interesting.

We’ve done it. You can now have an AI flirt for you. I’m looking forward to the first (mildly unethical) experiment where someone runs this, sticks a photo of an attractive woman, and sees what happens.

In lighter news, the dismal science

Fantasy Econ league! Because saying that academic publishing is about as important as the NFL is a great way to anger everyone.

Here’s a piece I enjoyed on the benefits of intensive parenting. It’s as well argued as the pieces I preferred arguing that they don’t exist, at least to the level of (minimal) investigation I subjected them to, so I feel obligated to pass it on as I would Caplan.

Urbanization is (still) really good.

Best piece of journalism I’ve read in at least a few months is this piece by Dylan Matthews on Robert Greenstein is about a particular person, and a broader approach to lobbying and politics, and the American welfare state. Worth reading, particularly for the second piece: Greenstein was extremely effective, and he did it mostly through research and people, not massive sums of money. Of course, the critique is to fault him for all the bad things that happened to the American welfare state on his watch, which isn’t wrong.

How Credit Cards Make Money is fascinating.

The primary component of profitability is net credit margin (NCM), which is the profitability of revolving balances.12 Credit card lenders receive revenues in the form of finance charges borrowers pay and fund the revolving balances with interest expense. On average, the credit function of credit cards—that is, NCM multiplied by the share of balances that are revolving balances—makes up around 80 percent of aggregate credit card profitability.

The share of revolving balances is highly seasonal, as revolving balances rise around the winter holidays and subsequently subside.

Spending on your credit card is now net costly for credit card companies because of all the rewards, and this is a recent development.

But see also Who Pays for Your Rewards, which shows that only 13% of credit card users make a net profit from their cards. The numbers are compatible, of course, even assuming identical data-sets and data cleaning and processing decision-making: making net money off of rewards, to a first approximation, requires that you pay no fees or interest.


In case you have somehow been laboring under the delusion that we can rescue science with meta-analyses, here’s Data Colada. It’s the second in a series, and it’s not looking good for finding the average effect size of interventions in a category as a meaningful value. So if you want to make a guess of how much impact some change will have, you can’t read individual papers and you also can’t trust the average meta-analysis. We’re still a long way from good science. Relatedly, though it really shouldn’t be a surprise, the effects of air pollution on health are exaggerated. Not zero, but substantially lower than previous (published) research estimated.

It turns out that people in the American West have been murdering federal agents a lot longer than I realized, going back to 1907. Congress was absolutely furious that the investigators…were investigating members of Congress for land fraud. The Congressional attitude that they should be immune to criminal punishment and laws isn’t new either.

It’s probably worth keeping track of how easy it is for an individual to make a machine gun in their house out of 3D printed + easily commercially available parts.

An interview with a Russian asset who, after being captured in Estonia and returned to Russia in a prisoner swap, defected to Estonia.

Thanks for reading! If you want more like this, you can subscribe with this button.


Don’t worry about me, I’m fine. Others were relying on full salaries and actually need support, for me it was one project of several.

New to LessWrong?

New Comment