Seeing like an outsider, with Yatharth

Capping off our experiments for the holidays, my Twitter mutual Yatharth reverse interviewed me. Some of it is professionally relevant (how did I fall backwards into knowing a Bit about Money), some of it is personal (how I met my wife), and some is application of lenses and scripts that I've accumulated over the years.

In particular, we come back several times to the theme that outsiders are forced to explicitly study cultures to be successful within them, and perhaps might understand them on levels that insiders do not. Or perhaps that is just a story I like to tell, as someone who has found himself the insider outsider for most of his career.

Sponsors

Check is the leading payroll infrastructure provider and pioneer of embedded payroll. Check makes it easy for any SaaS platform to build a payroll business, and already powers 60+ popular platforms. Head to checkhq.com/complex and tell them patio11 sent you.

Timestamps

(00:00) Intro
(01:30) Patrick's early interest in credit cards
(03:53) Navigating financial challenges
(05:10) Becoming a financial advisor
(10:35) Cultural and educational insights
(18:57) Sponsor: Check
(22:14) Personal stories and reflections
(25:48) Parenting and cultural integration
(33:26) Writing and storytelling journey
(37:35) Translating financial systems for a new audience
(39:08) The role of AI in writing and style transfer
(40:46) The concept of alpha in writing
(44:48) Legal advice and AI's role
(52:12) The impact of AI on bureaucratic systems
(01:04:22) Reflections on government software and policy
(01:13:49) Sabbatical insights and future interests
(01:18:52) Wrap

Transcript

Patrick McKenzie: Welcome to Complex Systems, where we discuss the technical, organizational, and human factors underpinning why the world works the way it does. Hi everybody. My name is Patrick McKenzie, better known as Patio11 on the internet, and I'm here with my Twitter mutual, Yatharth.

Yatharth: Hi everyone. It's good to be on the show, Patrick.

I spent the last couple of days asking myself a question. It wasn't very salubrious in nature, but it was: if Patrick died in a plane crash tomorrow, what would I really regret not having asked him in this hour? That's where I'm coming from today, and I'd love to dive straight in with some of those things that came up for me.

Patrick: Sure. Just to give context for people who weren't flies on the wall when we were arranging this - we follow each other on Twitter, we think we mutually have interesting ideas about a funny potpourri of things, and this'll be something of a reverse interview.

Just so people know where you're coming from, what do you do when you're not on Twitter?

Yatharth: Currently I'm in graduate school doing research on screen time addiction and AI capabilities. I'm also a software engineer at a small health tech startup in LA. On the side, I volunteer at a preschool, draw at an anatomy lab, and write a lot on Twitter.

Patrick: Alright, I have no idea what's coming next, but hit me with your questions.

Yatharth: The first one - I've read some of your articles on debanking and Know Your Customer requirements, and all these things.

[Patrick notes: The debanking piece is probably the best thing I’ve written in the last 18 months or so, and very worth reading if you’ve been following pmarca et al’s discourse on debanking as it relates to crypto or U.S. politics in the last few months.]

Yatharth: There's a magic to them. You seem to actually care about how the system works. In all the discourse I encounter about these topics online, it's mostly focused on stories that sound very plausible, with little concern for what actually exists. I understand you were working at Stripe and are now an advisor there, but I'd love for you to take me back to a moment before you understood credit card exceptions well enough to talk about them for 20 minutes, and how that experience came to be built up.

Patrick's early interest in credit cards

Patrick: Obligatory disclaimer upfront: I am currently an advisor at Stripe after leaving their full-time employment. Stripe does not necessarily endorse what I say in my personal spaces. (You can tell I have that one macroed out.)

When did I first start caring an irrational amount about credit cards? It was a long time ago. For those who don't know me, I had a fairly standard upbringing in a Chicago lower-to-middle-class American family. Happy family live, some hangups about money, yadda yadda. I got to university the same way many people do - by studying a lot and signing some student loans.

[Patrick notes: As long as we’re doing deep lore here: I paid for university in small part by making a list of every essay writing competition attached to a $500 or $2,000 scholarship I could find, and won a lot of them, partly because I wrote better than the average entrant and partially because the typical organization which throws a $500 essay competition will be lucky if they get 5 essays to pick from. (As a certain ethnically affiliated organization told my sister, when they proposed giving her a fairly sizable scholarship which any rational person would assume was extremely unlikely to go to anyone who looks related to me. We’ve got to distribute some grant money and you have the winning essay, on the strength of there being no other essays, so please help us out by cashing this check quickly.)

Patrick: If I had been more organized about this with a spreadsheet, templated essays, and similar, I probably could have simply paid for university that way.]

Partly due to one of the very few parenting choices from my parents that I will intentionally not replicate, I was protected from the mechanics of money until I got to university and had to figure it out for myself.

As soon as I arrived on campus, the marketing operation of the American financial industry descended upon me saying, "Hello, student who we rationally assume will have money in a few years because you are going to a research university in the United States. Would you like a credit card?"

I reacted as if someone had proposed a deal with the devil, because that's not something you're encouraged to get into if you're from a lower-middle-class American background. I gradually gained a somewhat more sophisticated understanding. One factor was simply attempting to read and give myself an education on consumer finance, because I thought, "This is going to be relevant for my interests. I'm going to become an adult. I will have to do adult things like getting a mortgage. Apparently getting a mortgage requires you to have a credit score." So I read up about that sort of stuff.

Navigating financial challenges

Then one incident ended up being really important: in the course of finding out what a credit score is, I pulled my credit report and discovered something surprising: I had over $100,000 of various forms of consumer debt. In many cases, those consumer debts dated back to before I was born. I did not remember having bought cars before I was born.

It felt like: one, I'm in a bit of an emergency here. It appears that the United States financial industry thinks that I owe them $100,000, which is approximately $100,000 more than I have. And two, this is obviously crazy. This was largely dawning on me after I graduated university and pulled my credit report for the first time, because who pulls it as a university student?

Anyhow, I get to Japan, I pull my credit report - oh my - and then I spend essentially a six-month adventure working with the credit reporting agencies (CRAs), a panoply of debt collectors, and financial institutions in resolving the various issues that the industry had with me. I can tell you how that crazy situation happened, but let's just say that I managed to get past it.

Becoming a financial advisor

In getting past it, I did what anyone does and Googled around a lot. I found myself on The Motley Fool's discussion boards and participated on the Credit Cards and Consumer Debt discussion board.

[Patrick notes: My pseudonym was laughably transparent to people who knew me, but I didn’t expect anyone to ever read those posts, so please don’t take unmasking 24 year old me as a challenge, Internets.] 

Patrick: After a couple of months, the questions that were asked by people coming in for the first time were very familiar to me because I'd gone through this process myself.

The answer was always the same: in this circumstance, you need to assert your rights under the Fair Debt Collection Practices Act. You do that by writing a short letter to the responsible person.

The typical person who ends up on the Internet with severe debt problems and beseeches a random forum to solve their debt problems cannot immediately action this advice. "You just need to write a short professional letter to the vice president of customer affairs at your bank, cite your knowledge of the Fair Debt Collection Practices Act, maybe mention Regulation E a few times, and then this will be solved for you." They find that rather difficult.

Because I had a lot of time on my hands and was on the Internet - and have this genetic abnormality that any question asked above a text box on the Internet demands my immediate reply - I would just ghost write letters for people. I think I ghost wrote on the order of a couple hundred letters over the course of the next few years, just walking people through the mechanics of this.

Meanwhile, I did a lot of random reading on the Internet. I think I read every Supreme Court opinion - I'm not exactly sure that I kept a spreadsheet of that, but I had lots of time on my hands. And while running a small software business, I did something that very few small software business owners do: I actually read all the contracts about credit card acceptance, because I was fascinated by how PayPal pays me money. It appears that money magically teleports from my customers to me over in Japan. This implies something cool in the middle. I wanted to understand it.

As a result, by the time I actually worked in the financial services industry for Stripe, they did training in the first week and the trainer asks, "Okay, who can diagram out credit card acceptance on the blackboard?" That's a common gambit that teachers make, expecting to say "Of course it's nobody, but I will do it for you." I asked, "How much detail do you want?" And the trainer said "You tell me how much detail you can give." I replied, "Do you want the case in which we tokenize the credit cards or the case where we don't tokenize the credit cards?"

She was like, "You understand what credit card tokenization is? Have you worked in the financial industry before?" I said, "No, I was a Stripe customer before. And tokenization is a pretty important thing in Stripe.js. It's for reasons related to PCI DSS." I got a lot of mileage out of my weird extracurricular interests over the course of working at Stripe for a number of years. And go figure, part of my job became explaining these weird extracurricular interests to various audiences.

That's my origin story of how I came to know a few things about various financial fields. And then there is a feedback loop here in any sort of Internet participation. The more you do anything, both you will get directly better at doing the thing, and when you do the thing in public - to use a very evocative phrase from my buddy Kevin Kwok - it's like tapping a tuning fork and seeing who resonates.

People who have maybe more connection to that thing than you do, maybe they've made it the entire focus of their career rather than simply having it as one of your diverse intellectual interests, they will say, "Wow, it's great that someone who can write actually cares about Regulation E. As a matter of fact, I have worked for a financial regulator for the last couple of years, and it really grinds my gears that people don't understand that Regulation E has cross-applicability to Zelle specifically. Can I take you out to coffee and just dump on you about Reg E?" Being willing to listen to people about that sort of thing exposes you to all sorts of tidbits. And this snowballs, if you allow it to snowball.

So that's my villain origin story.

Yatharth: That is more satisfying of a story than I could have come up with myself.

Patrick: By the way, people often ask if this was planned - did I have the 20-year arc of my career planned out, that I was going to decamp for Japan and then deliberately write my way into the financial services industry? Nooooo.

For a variety of class-inflected reasons, I was at an American research university, Washington University in St. Louis, with many people who took the more direct path into consulting or investment banking.

I told people contemporaneously, "I don't think I'm exactly the kind of person that goes into investment banking." And they said, "What do you mean? You're in the engineering school, top of the math classes, etc. It seems like you could probably do this. Have you seen the people that show up to the interviews?"

I said, "How would one even go about that?" And the true answer was probably the open casting call interview they were having on campus. Perhaps I should have arranged to go to that. But that did not occur to me as an undergrad who knew little about anything in the world. 

That is one mechanism I try to explain to people - "go through the door that is open" is a viable strategy in a lot of circumstances.

Cultural and educational insights

Yatharth: There's this phenomenon of people who generate a lot of value for others writing on the Internet where some misfortune happened to them. They had to figure out a lot of stuff that normal people didn't have to figure out and they're somewhat autistic about writing about it. In your case, it's this massive fraud and the fact that you actually didn't come from that professional-managerial class culture. It's very interesting and maybe reflects the way you talk about it.

In my own case, I have had to figure out a lot of things about just how emotions work and feelings - things that I now articulate in painful detail to some, but that people regularly message me saying "this changed my life," just because I had to figure it out from scratch and write about it.

Patrick: Yeah, the intersection of class and education and professional capabilities/responsibilities is a deep well. I think part of the story is that as one grows into adulthood, one starts to master things that different people encounter at different rates.

You mentioned a particular word that I have a complicated relationship with. I like to phrase it as "being wired differently than the other kids." Growing up, I was very obviously wired differently than the other kids, but there was less of an understanding that that specific thing could be medicalized. Although they did try. There was a school psychologist at one point who wrote down in some formal fashion that I had a witch's brew of issues. They would prevent me from ever achieving academic success if I didn’t receive medical attention.

I went to my mother and said, "This is clearly nonsense. I've had nothing but A-pluses for my entire life. Clearly we will just bring in the report card and people will understand that I am not, in fact, disabled in any medically relevant fashion."

My mother, who had been a teacher, understood something very important about the education system: there is reality and there's a piece of paper, and where the piece of paper disagrees with reality, the piece of paper wins. My mom used a very specific word in that conversation: "We will unmake that piece of paper."

I only gradually as an adult started to imbibe some of those lessons. Oh, people lie as a considered strategy sometimes. Sometimes it's more of an accidental thing, but it still matters. And institutions are not necessarily truth-seeking by nature.

Anyhow, class is complicated. One of the things that fuels my weird interest here: I got to university, majored in East Asian Studies - you can round it to majoring in Japanese - in addition to Computer Science. There are performances that one must do as a student of the Japanese language, for example, being able to perform the difference between in-group and out-group. This is explained to you explicitly because you literally cannot produce the Japanese language if you don't see in-group versus out-group, just like you can't produce Spanish if you don't reliably differentiate between genders.

They explain to you the difference between in-group and out-group, and then there's an aside: "By the way, your culture has an in-group and out-group too, it's just probably not been explained to you." That was a bit of a moment for me - oooooooooooh, I wish someone would have given me that memo 15 years ago. And the notion of needing to manage up to one's superiors and treat them differently than one's peers, which I encountered gradually in my twenties. Oh wow, they wrote this in a book. There's probably a book like this about America. I wish I had a copy of that book when I was seven years old. That would have made life so much easier. So partly this is why I keep writing these things: be the book that you didn't have when you were seven.

Yatharth: Oh my God. I remember what that book was for me. It was this book - I'm ashamed to admit this - but it's called "How to Succeed in Business" and something about how alpha men work in business, a guide for women who want to succeed. I kept reading it, and page after page, it was incredibly cringe and incredibly life-changing too. It was talking about the way people would apologize, and I'm like, "Oh, that's the thing I was doing, that's the thing they were doing." I didn't understand. All of these books that are written for outsiders that you just sometimes wish that you'd found.

Patrick: Yeah, not endorsing, but by way of explaining - this weird stuff when I attempted to get my head around the concept of dating in one's 20s, which I was encountering in my 20s and not teens as a somewhat delayed individual, what can I say. The literature around dating for men in their 20s was talking about things like status games, and the notion of a status game explicitly connected to the Japanese language and performing professionalism in a Japanese workplace. That was new to me.

Now it turns out you can learn about that sort of stuff if you take improv classes. I did not take improv classes - I took math team classes, and status games don't come up in that curriculum. But it ended up being useful.

I think some of the general reaction and the general understanding of things as "cringe" is one of the ways the system protects itself - to describe as deviant the behavior of describing the system.

That is a recurring thing that comes up in these circumstances. Mentioning this out loud because it is important.

There are a variety of systems that are extremely relevant to people's lives, say the financial system. Many of those systems have formal, written canons of behavior, but they distribute those formal canons of behavior widely. There's an informal shadow canon of behavior, and it is either implicitly or sometimes explicitly deviant to document the informal canon of behavior.

One somewhat mind-blowing fact is that there are lists of red flags, high-risk activity for going through anti-money laundering processes at banks. One thing which is on many of those red flag lists is: "the customer evinces an understanding of anti-money laundering law and regulation." And that isn't on that list irrationally. You're thinking: if they haven't been in the financial industry and haven't been through this complying with AML regulation course, very few customers understand AML. And very many money launderers do, because that's professionally relevant to them. Therefore, on Bayesian evidence, if you understand the regulation we are enforcing against you, you're probably evil.

One can understand how the system arrives at that equilibrium and also think, "But wait, from our basic understanding of civics and democratic governance and how systems should work in some Platonic sense, that sounds incredibly perverse." And you need to be able to hold both of those ideas in your head at once.

Yatharth: Yeah, like one of the things I had to learn was - what's a better word than "autistically"? That's a term I picked up from Twitter's casual usage of it, and I don't like it and I wish there was another one, but using it nevertheless. One could autistically spam the truth button. "This is true, this is true." It's obviously a quick route to getting a cold shoulder in a lot of contexts, and it has taken a lot of work to go from there to actually speaking with contextual sensitivity to audience and waiting to talk about something publicly until all of the resentment and anger around it being a shadow canon in the first place has been processed. I'm curious if that was ever a factor for you when you started encountering these things.

Patrick: Yeah, and again, this was helped out by textbooks which explicitly said things like: when you work in a Japanese company, your superior [Patrick notes: was “subordinate” in the audio, that version is also true but was me misspeaking] will very likely be extremely annoyed if you tell them in the course of a meeting that they are wrong. The proper ritual for doing that is to wait until you're getting drinks together. Your drunkenness is not a requirement to engage in this ritual - everyone understands there is a social context for it. You need to have an alcoholic beverage in your hand, and then you are allowed to say absolutely anything, and their part of the ritual will be mostly ignoring you in the moment and not bringing up this conversation the following day, but successfully processing the information that you gave them during the uncomfortable truth-passing ritual.

And you start to reflect on previous interactions with teachers where the teachers were annoyed when you pointed out math errors. Why would a teacher be annoyed for having an error pointed out? Obviously math works for 14-year-olds the same way it does for math teachers. And it's like, "Oh, things are falling into place a little bit."

Another book - this book has extreme partisan political valence, and I will say that I do not endorse the political valence of the author or their broader project, but it's very useful to read to understand how power works - "Rules for Radicals" by Saul Alinsky. They describe things like: bureaucratic institutions have a written canon of behavior which they're simultaneously incapable of upholding. If you force a bureaucratic institution to operate under its stated rules of behavior, that will cause breakdowns within the bureaucratic institution, which is useful if you are an opponent to the bureaucratic institution.

I read that when it became politically relevant to the United States for various reasons. 

[Patrick notes: Long story short, the author and a future U.S. president were quite buddy buddy. His opponents made quite a bit of hay out of that; this was unsuccessful in derailing his political career, but probably sold a lot of copies of Rules for Radicals, including one to me.]

Patrick: I did not read it like "okay, this is how to cause the collapse of democracy and replacement with democratic communism" that the author actually wants. Rather, "Wow, this is really useful for understanding how a credit reporting agency works" and getting the things you want out of the credit reporting agencies.

Just to continue in a for-loop, asking them "Was that compliant with your responsibilities under the Fair Credit Reporting Act?" and continue forcing them to document their compliance or noncompliance with things until they give you what you want to get you off the phone. Not the phone - do it all in writing. The phone is a place where consumers get abused when talking to credit reporting agencies.

These are somewhat hidden facts about how the world works.

Playing back my memories of formal education in the United States - we are capable of telling children as they're coming up that you write differently for different audiences, you adopt a different voice for those audiences. But we don't talk about the status ladder explicitly. We largely don't talk about the functioning of adult institutions, partly because we don't trust kids to understand them, partly because teachers are heavily socialized into a particular class and tend to have blinders on with regards to that fact.

Filling in the gaps in one's education is a project for many people in early-to-middle adulthood, and hopefully one continues to learn for the rest of your life too, but you're not operationalizing them by the time you get to middle age… there is room for improvement.

Personal stories and reflections

Yatharth: I'm curious - when you did end up encountering the literature that made some of these status games explicit, in dating and so on, how did that end up going for you? And how did you eventually end up meeting your wife?

Patrick: As some people might be able to tell from my general presentation, I have at no point been a suave Casanova.

I dated a couple of times during my twenties, not to a great deal of success. Didn't understand at the moment why I was probably not a great boyfriend. Did come to that understanding in clarity of hindsight - I was, as many salaryman do, putting a huge number of my character points into the mechanics of being a salaryman, and was probably emotionally unavailable except during the three-hour date window every Saturday. I did not have a sufficient understanding of theory of mind as to why that might be other than "just the thing that everyone does."

How did I meet my wife?

The literal truth is: I was coming off three years of being a salaryman at a large Japanese company, spent all my time in salt mines, etc. I quit, for various reasons. My best friend Keith took notice - "Hey, almost no one has seen you for about three years now because you've been hitched to the wagon. We would like to see you again. So will you come out to a barbecue?"

I said, "Oh yeah, sure, I'll come out to the barbecue." Day of the barbecue rolled around and I texted Keith to say I'm just not really feeling it right now - I'm dealing with a little bit of depression, even though I probably didn't phrase it that way - so I'll skip it.

Keith has long since had a key to my apartment in case of emergency because we live in Japan and earthquakes are a thing. So he let himself into my apartment and physically threw me in his car and took me out to the barbecue. I was Mr. Sourpuss about this decision. I said, "We're going to go to the barbecue, there's going to be some new English teachers that have just arrived here in central Japan. And we all know how this game goes - they're going to be here for a year, then most of them will go home. There's no point in talking to any of these people. They have very different social situations than we do."

True for most of the attendees of the barbecue. However, Keith's wife had invited a friend of hers from baby English classes, and that mom did not feel exactly comfortable being one of the only people not primarily an English speaker at a barbecue, so she decided to invite a friend of hers from work. That friend's name was Ruriko.

I took one look at her and was smitten.

We're getting very far off the usual beat for this podcast - but whatever, run with it. I thought "how does one chat up a young lady?" Don't really have so much experience with that. I think the question that one usually asks in these circumstances is ‘What sort of music do you like?’"

She said, "Actually, my favorite artist is Beyoncé." And I said, "Wow, how convenient - my favorite artist is also Beyoncé!" Which is not in fact the case. "What is your favorite Beyoncé song?" And I thought when saying that, it better be "Single Ladies" because that is the only one I know. And she said, "It is 'Single Ladies'!"

I said, "What a coincidence! I love 'Single Ladies'!" And then I got up in front of 200 Japanese people at a river in Gifu Prefecture and a cappella'd "Single Ladies" in full diva mode. Some time later, after proposing, I asked her, "When did you know this relationship had legs?" And she said, "You had me at 'Single Ladies.'"

And we've been married happily for 12 years now.

Parenting and cultural integration

Yatharth: This is the second time in this podcast where I am receiving an answer far more satisfying than anything I could have come up with myself. It's very delightful to ask you questions. I am curious also how it has been for you, this process of having children and watching them grow, and how old they are, and your sort of observations there as someone inclined to notice bits of culture and articulate what things people otherwise miss.

Patrick: Sure. My daughter Lillian is ten now. My son Liam is seven. I have many observations on parenting. Where to begin?

[Patrick notes: It’s worth saying explicitly at this point that one of the exercises I immediately go through when answering a question like this is not merely calculating “What do I think is true?” but also “What do I think which makes a good story that is novel to my audience?” and “What will make a good story which is socially acceptable to say broadly?”] 

We'll begin with one observation. I think the phrase used for this is "third culture," where people are not entirely in cluster X and not entirely in cluster Y. They have to deal with being at the intersection or some sort of fusion or integration of X and Y, and this often affects the "third culture kids."

My children are central examples of that concept. Let's say I have some cultural background as an American and then spent my entire adult life in Japan and ended up in some superposition of those cultures. I don't exactly know that synthesis or blending or similar is the best model for it. I tell people I'm a recovering Japanese salaryman, and sometimes they take that as a joke. It's not a joke. There are parts of that value system that I very much vibe with, as the kids say.

My kids attended a public elementary school in Japan and did not, at the point of being second grade-ish, have the experience of ever meeting the American educational system. Then we moved to America, and now they are in a local Catholic school here in Chicago, and there's a bit of difference in how a Catholic school runs things versus how a public school in Japan runs things. For one, to use an old shop-worn joke, if you don't do well in math classes, we nail you to a plus sign.

They're adjusting to this new reality. They're adjusting to this language - they've certainly encountered English before. The family rule since forever, from their perspective, is that I only speak English in front of the children. My wife only speaks Japanese. When the two of us were interacting together, probably 90-ish percent of our conversations were Japanese prior to the kids being born, simply because it's easier for me to speak Japanese than it is for my wife to speak English. They're in this bilingual, bicultural - it's not really bicultural, the situation is much weirder than that - environment at home and then dealing with this at school.

One of the reasons motivating the move to America was - there are many things to love about Japanese society. However, many people in Japan who are not viewed by many members of Japanese society as being 100 percent Japanese have a certain amount of friction or frustration associated with that fact. I seemed to be less annoyed by this than many people, for a complicated variety of reasons. But be that as it may, I definitely had a choice and have continued making that choice for 20 years in my adult life - nobody kidnapped me and sent me over to Japan, and there's a plane ticket back anytime I want it. Yet I continued re-upping, and this being an outsider thing - at no point in my life did my kids get a choice on being outsiders. They just ended up there by default.

That choice was starting to have consequences, and they are unpleasant consequences to report. If your daughter comes home one day and reports a conversation at school that starts with "So I'm Japanese, what the hell are you?" - that isn't a fun thing to hear. And it's a lot less fun to experience. So that's one of the things driving our current American adventure.

All cultures change over time. Japan has a complicated collection of people who have diverse views with respect to how foreigners integrate in society. That cultural cluster has moved very perceptibly on this issue in the 20 years that I've been there, and it will probably continue moving over the course of the next 20 years. Who knows how much it will change before Lillian gets to university or before she gets her first job, if she gets to university or gets her first job and feels "I am just done with dealing with this nonsense."

I thought it would be nice if she had the experience of living in America as an American, where she will deal with far less nonsense on that score. Then she will have some amount of choice and agency: "Do I want to live in Japan and then put up with this one thing that will periodically be a thorn in my side? Or do I want to live in the United States of America and get a different collection of pluses and minuses from that?"

Any thoughts aside from the cultural thing - if you have a good memory of what happened to you as a child, and then you are raising children, you will realize that many of the things that happened to you as a child did not happen for the reasons that your child brain thought they happened. You start to appreciate things like, "Oh, my aunt wasn't around the house frequently by accident - that was to help out my mother in raising us, because mom needed to do some things." The reality of those things is not frequently exposed to me as a five or seven or 12-year-old, because one does not necessarily expose one's five or seven or 12-year-old to all of the facts relevant to their situation.

I'm somewhat backfilling models that I had of the world as a child, which are not good models of the world and do not generate predictions that will routinely turn out to be true. One of the relatively few reactions to my old childhood experience that I'm intentionally being reactionary on is: my father basically taught me to read with The Wall Street Journal on his lap, and he would explain the mechanics of this stuff that was happening in far-away finance land, but was borderline religious about never speaking about finance at the family level.

I've had directionally similar conversations regarding "Okay, here's what a mortgage is - it's a loan between the bank and you, it has an interest rate, etc. By the way, the 't' is silent. No, I don't know why." My kids have been extremely aware for a while that we have a mortgage and I pay every month, and if we don't pay every month, there will be consequences associated with that. That's one reason we work at trying to give them a well-rounded understanding of things relevant to their interest.

Yatharth: Yeah, I find it interesting to hear about some of these details, like The Wall Street Journal, because more and more, it makes a lot of your story seem - it obviously came from somewhere. It came from a bunch of places. It didn't have to go like this, but those things were there.

Patrick: Can I make a disclaimer on that point? I'm a storyteller by nature. Hindsight is 20/20, we have rose-tinted glasses, etc. Did it actually happen this way, or is this a narrative that I'm capable of constructing retrospectively and telling in a very pleasing fashion? I don't know.

I think most of the major points that I make in this narrative probably mostly actually happened. I'm not inventing convenient details most of the time. The nature of being a storyteller is one is naturally compressing the procedural record of what happened. One kind of amps up the importance of certain things that are narratively relevant or resonant, and deemphasizes some other things that happened. One tends to tell the story in a way which makes sense to oneself, which is possibly not correctly drawing the inferences.

Be that as it may, that is my disclaimer: you should trust that I think that I'm telling the truth, but you might not trust that I'm actually describing reality as it happened with the appropriate amount of emphasis in all places.

Writing and storytelling journey

Yatharth: True. I'm glad you're at the point where you seem to have fairness and no hesitation about describing yourself as a good storyteller because you are, and I've met a lot of people who are and seem to struggle with that word. I'm actually curious - because even the way you speak obviously matches the way you write, and there's something that's very engaging about it. It's very wry, it's very often indirect, and I'm curious if you have thoughts about how your writing style or speaking style changed over time, being the one who'd be most familiar with that.

Patrick: On one hand, people sometimes describe me as quite indirect and elliptical in my writing and speaking style. On the other hand, I'm really a Japanese salaryman here. A number of things that Americans think are wry and indirect feel like me straight up punching somebody in the face.

There is a bit of an expectations gap there. I think I regret that I don't have copies of my writing from high school or college because those were the pre-Dropbox days. I was much more direct in writing, also more emotional, more raw in those days than I am currently. Partly style evolves over time, partly there really is a reset button in the center of my life where I do a massive flip between cultures and start adjusting to the new reality. That wasn't the only reset button in my life, but it was a big one.

Partly one grows and matures over time as everybody does. As every generation instructs their children: you will not be the same individual when you grow up as you are right now, be ready for that. Every generation of children says "No, we are the ones that will definitely never grow out of our current thoughts." That's a tale as old as recorded history, I think, but it has to be so.

Part of it is simply having more experience with what works for me and what resonates with readers. I've written somewhere on the order of 5 million words - which for context, the Harry Potter series is 1.08 million. In the process of writing those 5 million words, I've had the opportunity to introspect on when I touch on this topic, when I write in this style or voice. At the end of the day, I get to read my own writing and say, "Oh yeah, I'm pretty happy about that one" or "Oh, really?"

When I look back at things that I wrote in 2006, in the early days of my blog when I was still finding my way - I was a pretty good writer in 2006. My teachers in university had recommended my work for publication in journals. I was extremely good at pleasing the teachers in high school too. It turns out that there are ways to be a good writer that do not strictly map to pleasing authority figures.

In the early days of my blog, I was floundering around with regards to topics, mostly writing about what was top of mind for me and my work with software business on any given day. Then eventually I refined to what I can talk about better on that subject. It seems when I talk about marketing and sales stuff, that it resonates more with readers. There is a gap of people on the internet who can describe marketing and sales in such a way that engineers don't think it is evil and/or witchcraft. So I focused more on that.

The center of my writing or my beat - where has that changed over time? Sometimes in response to whatever I'm working on, the thing obsessing me on a day-to-day basis is usually a thing that causes me to obsess for the hours or days that it requires to get a piece out the door. Sometimes I have "Bits about Money" as a publication distinct from my blog, because I wanted a space where I can just talk about financial infrastructure. People have been reading my blog for 15 years might or might not be here for my thoughts on financial infrastructure, but if I have this dedicated space, then I can award myself permission to geek out on that one subject that is relevant to my interests.

Complex Systems, similarly, is awarding myself permission to do a slightly different form factor, a slightly different broader look at infrastructure systems, societal and cultural factors connected to them. In my other spaces, I tend to avoid writing just about philosophy or just about anthropology when it's not connected to concrete financial systems or similar.

Translating financial systems for a new audience

Yatharth: A lot of the really important writing happens as translation. It's interesting to think of your blog as someone who writes with a certain eye of an engineer or a cultural explorer, but taking these topics like financial infrastructure and PMC negotiation that are otherwise very loathsome or repugnant to the kind of person who would write in that style.

Patrick: There isn't zero value created by humanities departments - happy to say that, got a degree from one. And yet the amount of signal that you will get in reading the collected works of every humanities department ever with respect to class - you'll get some signal on how the financial systems work, but it's a pretty inefficient way to get that signal. And fill in the hole in the internet.

Yatharth: ChatGPT and Claude are quite aware of your writing style, and can somewhat produce it terrifyingly well.

The role of AI in writing and style transfer

Patrick: It's interesting how you use the word "vibes" - we love that word on Twitter. They are just amazing bits of technology with respect to doing a style transfer or vibe transfer. To cook eggplant parmesan except in the style of a Shakespearean sonnet - that was something they were able to do extremely well, almost out of the box years ago, which certainly sounds like it requires some amount of intelligence.

Similar style transfers on "okay, sound like patio11" except on topics that I have never written about, they do better than one would expect for that. And certainly cheaper than one can buy it from any other place on the internet. If you ask a class of high schoolers who are all literate in the English language, "Okay, I'm going to give you this guy's blog, now I would like you to successfully predict how he would write about dating" - topics that I have never and will never cover on my blog - the high schoolers will not do a great job of style transfer. Professional writers will not do a great job of style-transferring Matt Levine to a topic in finance that Matt Levine has never written about, and ChatGPT does an amazing job.

[Patrick notes: Relative to expectations! Particularly expectations calibrated around, say, 2019. Here's a low-effort version of prompting ChatGPT. This is not better than what I'd write given an hour on this assignment, but it is better than what most writers could deliver in an hour, and cost ~nothing.]

They don't get everything right - the last 20 percent is really rough for them at the moment when writing by experts for experts on a variety of topics, but oh wow, will it pass your inspection for the first few seconds? Yes. That's a somewhat amazing fact about the world.

The concept of alpha in writing

I think it should probably inform our - there's a variety of words that I wish we had, which we do not yet have. One word is we have the concept of "alpha" in finance, and alpha - one Greek letter smuggles in a huge amount of understanding about how the world works. I would love to be able to describe someone's alpha above the LLM baseline in discussing a topic. Because there are a lot of human writers in the world who have no alpha above the LLM baseline, and that's been true since before LLMs were a thing.

The Twitterism is sometimes "this person is an NPC" - there is no intellectual content here, the performance of class and similar can allow one to pretend that there is intellectual content, but there is no intellectual content. Now that we have an LLM that can spit out infinite amounts of minimal intellectual content on demand, I would love to be able to describe "okay, if you were reading about topic X, writer Y is someone that has substantial amounts of alpha above the LLM baseline." And writers Z, Q, and P who are often recommended to people by university syllabi and by high-status journals and name-checked by politicians actually have nothing going on there.

Hopefully someone will develop the Greek symbol that everyone understands to mean "is someone distinguishable from an LLM." And that will start to be a moving target because the LLMs have been getting rather substantially better every year to 18 months, whereas the replacement-rate writer on a topic - I don't know, financial services for a well-respected journal of news and opinion in the United States - has not seen a great increase in capabilities in the last 18 months, nor did they in the 18 months before that, nor did they in the 18 months before that.

It's a challenging set of facts about the world, but they are both facts. I don't have very strong opinions as to whether we will continue seeing a great increase in capabilities over the course of the next 18 months then the 18 months past that for LLMs. I do have a pretty strong default hypothesis on whether the average person writing in a high-status journal of news and opinion will understand the financial system much better 18 months from now - to the extent that they are not simply letting an LLM write their columns for them.

Yatharth: I think there's a distinction between writing styles and thinking styles. One of the things LLMs can already do wonderfully, as I think many readers have reported to you, is use the dangerous professional voice, imitate that pretty well. Cool style transfer, no problem.

Something interesting I find is unlocking thinking styles, especially now that we have more compute in context, more chain of thought, all of that stuff. In particular, to connect to our culture of PMC ritual objects and unlocking behavior - like unlocking the non-NPC behavior from doctors and nurses and so on in conversations - I found that putting a sentence like "Hey, talk like we're in a Slate Star Codex article" or "talk like we're in a Gwern article" suddenly makes the LLM not dumb. I don't think it's because it magically gave it more intelligence - it just made it stop giving me the normal answers that it assumes everyone wants, or that it's maybe trained on during the RLHF process, and just makes it more willing to be rigorous yet speculative with what it knows and continue thinking in those lines.

Legal advice and AI's role

Patrick: It's amazing to think that just someone's name is a pointer to some sort of conceptual space, and that LLMs can follow pointers like that. It's somewhat wild to me - this technology is bordering on magic. I think there are probably, descriptively speaking, other ways to coerce similar behavior out of LLMs.

If you are floundering around in a space where you are not yourself an expert, there are certain words that are more effective than other words in getting the right advice out of LLMs. For example, because certain regulated professions have a very complicated relationship with the word "advice" specifically - asking for advice is a great way to get the LLM to say "Of course, I cannot give you legal advice."

However, if you learn the sort of circumlocutions around asking for advice that you would get - say this is not hypothetical - if one had spent a substantial amount of billable hours speaking to lawyers who had made the unlicensed practice of law a center of their legal practice, and like, where does the boundary start? You know how to articulate it in such a way that the LLMs will provide interesting information.

Let me save everybody a very large legal bill. If you talk to people who are actually experts - they're registered with the local bar - about the unlicensed practice of law, one of the things they will tell you is try to avoid using the word "you" and particularizing things to the situation of someone you are talking to. That sounds more like advice, which you are not allowed to give. Instead, make general observations about the legal system itself - that is information rather than advice. Don't use the word "you," "your company," etc. "A company in the United States will be taxed by..." etc.

Then one might hypothetically have a follow-up question to that lawyer and say "Hey, this sounds really silly, does it matter?" And they will tell you "Oh, the reason you are paying me for this is to tell you what does and does not matter." And this is a thing where I will be like, "Huh, now I have learned more of this crazy system that is the law that we all operate under."

One of my other shake-fist-at-society moments was discovering that - I seriously considered getting a CPA at one point because I was annoyed that when I wrote about accountancy on the internet, I had to routinely make the disclaimer "I am not an accountant. I'm not your accountant. However, I do know a little bit about how taxes are performed in the United States."

And I got told good news, bad news: there is a path to you becoming a CPA. However, you will not be able to say as a CPA "I understand how this is going" in internet threads because you will not have a professional relationship with the person you are addressing in internet threads. So you will still have to do circumlocutions like the ones you do right now. Even an accountant can't play an accountant on the internet. This is madness.

Yatharth: Okay. Only your accountant can play an accountant on the internet.

Patrick: Essentially, yes. And there's reasons for it. Complex systems didn't arrive out of nowhere.

Yatharth: So one of the amazing things is that your name or your handle is now one of those LLM unlock words, in the sense that when I'm frustrated, sometimes asking the LLM to "produce an article as if patio11 was talking about the internals of this topic" and boom, suddenly it's very willing to talk about this thing. I love that you have this space in style and thinking that is now going to be encoded in this tool that will increase and predominate our lives.

Patrick: I would like to explicitly say - because many writers believe the opposite - I love this too. I don't think that LLMs are a replacement for me in economic life. I think they are a force multiplier for work that in many cases I did more than a decade ago and wish I still had time to do every day but do not.

Back in the day when I was extremely underemployed and had a lot of time on my hands and no wife and no children, I could ghostwrite letters to banks. And now an LLM told to ghostwrite a letter in the persona of patio11 to a bank will come up with a very acceptable output often on the first try. In my few experiments with this, it's required two to three tweet-length suggestions as to "Oh no, take out that paragraph, reword it," etc.

This is a free extension available to essentially everybody in the world on work that I did many years ago. I don't necessarily know that it's a necessary extension - I haven't rigorously tested "write a letter in the style of patio11" versus "write a letter in the style of a lawyer expert in consumer law." But I do know that people keep filling my inbox with "Hey, I have this problem with the bank. I went to an LLM and asked them for you to help me on it, and now I no longer have a problem with the bank" and that makes me extremely happy.

Feeding this back - who was I writing for in 2006? Or who was I writing for when I was ghostwriting these letters back on the Motley Fool discussion boards? I have some notion of that. And at no point during that notion was it "Ah, clearly there will be rapid advances in AI technology over the course of 18 years, and an important audience to always consider when writing is future AIs that will be using this to inform many more letters than I am actually physically writing."

Now in 2024, when I write things, part of what I am thinking about is "Will this inform future intelligences?" I don't have a strong point of view on whether to count it as intelligence or not. I actually have an increasingly out-of-date undergraduate concentration in AI from that other degree that I have, computer science.

You're very familiar with this, perhaps many listeners have heard about it as well - there's this extremely unproductive argument that has been had in computer science departments since forever on whether a computer system can count as intelligent or not. One part of this argument was Searle's Chinese Room hypothesis, which in brief says that if you put a man in a box with a sufficiently detailed set of written rules for the Chinese language, but he doesn't actually read Chinese, and requests go into the box and he successfully produces output to pass outside of the box that is Chinese, but can't actually read the things, just doing simple manipulation - is the system as a whole competent in Chinese? You would say yes. And believe it or not, that counted as intellectual contribution 20 years ago, folks. We had very little going on in the AI departments.

One other thread of similar arguments is Chomsky's perspective - wildly overrated, but he said some things that are true. One of the things that he said: "Do boats swim?" Is that a characteristic of boats, or is that a characteristic of the language that you use to describe boats? In Russian, submarines swim - that's just the way it is. Do you and the Russians have a different point of view on the capabilities of your submarines? Are your submarines different technical artifacts, or do you simply speak a different language?

That's pretty much where I am with regards to "are LLMs intelligent?" I don't know, I think that kind of matters where exactly you're drawing the contours of the word "intelligent" within your language. I will observe that they write persuasive letters to banks given a tweet-length description of what that letter needs to be, and that those letters cost approximately one-thousandth of what it would take to have me write the same letter. But I will tell you as something of a published expert in this topic, that letter will have the same outcome when sent to a bank as if I had written the OG organic free-range intelligence version of that letter. We are where we are.

Yatharth: One of the things that, given the amount of writing you do about SaaS businesses and marketing and sales, that I haven't quite seen as much of is the impact of AI agents, LLMs, and so on. I'm wondering if this is bubbling in you and some massive article that's going to come out, or if you've been tracking that at all.

Patrick: Ooh. So theoretically, one of my goals for the sabbatical that I currently find myself on was throwing myself into AI in all the ways. I’ve spent a lot of time on Factorio. I've spent quite a bit of time decompressing. I haven't spent as much time on LLM related projects as projects in the last year and a half now, as I expected to.

The impact of AI on bureaucratic systems

Man, there are so many unanticipated consequences of LLMs existing as a technology. And if you are listening to this podcast, you might not appreciate how much more you appreciate those unanticipated consequences than people in positions of power do. In a previous podcast, Dave Kasten told me that when he listens to folks in Washington talking about the great AI national security threat, they talk about things like "Obviously, there's election interference and misinformation, and the fact that you will be able to conjure up misinformation much cheaper than having freelancers hired by adversarial nations - that is the national security threat."

There are some people listening to this podcast who said "Sure. How about the end of the world?" And Washington might not have in their understanding that serious people think end of the world is a thing to be worried about right now, despite Washington having many people in various offices who are in charge of literally ending the world.

Let me make it more concrete and less doomy: there exists a variety of systems in the world that have resolution paths built into the system, where those resolution paths guarantee a certain amount of human attention on behalf of the system to a person who attempts to engage the resolution path. And often what counts as "attempting to engage the resolution path" - there is a defined tripwire for that. And that defined tripwire might be writing a facially plausible letter.

And if you are a bank, and you have a hundred million customers, which at least one bank in the United States does, you are institutionally aware of how many facially plausible letters your customer base is capable of writing every year. And there is a number there. And based on that number, you have a certain number of people sitting in an office, reading facially plausible letters and engaging the resolution path.

What if facially plausible two-page letters were no longer scarce in the world? What if anybody could conjure up a hundred million of them and all of them would be, under your own rules which you must follow, facially plausible? How quickly can you staff up your office to deal with infinite facially plausible letters? Is the person in charge of staffing for that office - who is a real person who has a real address, this is really their job - do they understand they need to start staffing up right now?

They don't. On the last episode of Complex Systems, we were talking about protests for real estate appraisal, which drives real estate property taxes. There's a defined procedure for protest in basically every county in the United States. Do the people who are in charge of running that procedure for protest understand that the amount of human cognition available to do a protest rate-limits the number of protests that they receive, and equivalent cognition is available from the market at arbitrarily low prices right now?

They can expect to get many more protests in 2025 than they got in 2022 as default. And then very shortly after that, you will be able to Google "how does one protest one's property taxes" and get someone who built a tool that has an LLM under the hood that will generate a protest for five bucks. The people who are in charge of those - to use Dave Guarino’s very evocative phrase - "adversarial touchpoints" absolutely do not understand they're about to get smacked by a tsunami in the face.

Yatharth: We've seen this with driving ticket apps that can contest it for you. And I don't quite know how that system handled that - probably has a five-year headstart.

Patrick: There are a number of things that have that start. Like for property tax appraisal disputes, there was this prior art on having computers generate those disputes, but they're very low effort and often can be templated as low effort - you understand on reading it very trivially, "Oh, this is a garbage dispute. I will put it through a procedural tripwire. The system that is interacting with me will not be able to walk over that procedural tripwire, and then I'll be able to discard it and talk to the actual informed humans."

Agents are going to be able to sail over those procedural tripwires very soon. Back in the day when we were doing this - myself and other assorted people in civil society were exercising the rights granted under consumer protection legislation on behalf of people that had issues with the financial system - one of the funny things that happened in the United States political economy was that the credit reporting agencies convinced Congress to put a carve-out in the Fair Credit Reporting Act for abusive templated communications.

Their point of view is that okay, customers have rights, but if you assert that right in a templated document, we should be able to throw out the templated document immediately. That's not the way that many rights work in the United States - if you're a government agency and you violate the First Amendment, merely the fact that your violation of the First Amendment is adequately described by a piece of paper that a number of people could use does not excuse your violation of the First Amendment.

For the Fair Credit Reporting Act, things being templated did excuse the violations. Part of what consumer advocates like myself were doing in ghostwriting those letters was: you have to be sufficiently aware of what is going on to cause the state machine to make the next transition that you need for this customer. But you don't want to be too explicit about what particular paragraph of the legislation you're referencing because someone reading it will be like "Oh, templated letter, RIP."

That sort of tripwire exists in prior art that is attempting to protect these adversarial touchpoints from scaled use by advocates. However, if I was good about reading responses and saying "Oh, you've asserted that I'm using a templated letter, but I am not, in fact, using a template - I am a human, and I reiterate that you must engage the next process here" - the LLMs are arbitrarily resourced to do that.

The naive version of "Okay, don't allow the LLM to lie that it is a human" - oh, the list of rules is written right now and the list of rules doesn't have a rule that you can ignore it if it's computer facilitated. If you write the rule that you can ignore it if it's computer facilitated, I can put a human in the loop there. And in fact, we do that in a variety of places in various systems and allow the human to use the system as leverage, but have the human be accountable for the deliverable.

So again, I'm very interested in how this shakes out and it will shake out in different ways in different places at different timelines. It's a little bit alarming to me that people in various positions of authority and power still consider LLMs to be like that weird new toy that the geeks are interested in. Oh boy, folks, this is relevant to your interests and it's going to hit like a freight train.

Yatharth: We've just broken the social blockchain that had required proof of human work, and we can now just DDoS government infrastructure arbitrarily.

Patrick: And not even - there are certainly abusive ways to do it and to DDoS etc. And maybe you can go to a lab or similar or to RLHF and say "Okay, clearly you wouldn't knowingly participate in directed concerted action designed to degrade the behavior of a subsystem of the government." And I think you would get widespread agreement that yeah, this is not why we're building this tool, etc.

Okay. If I phrase it as "Clearly, you would not help a Kansan grandmother in avoiding an improper foreclosure on her house" - I think a lot of people would say "Wait, no, that is in fact not against my morality or my intended use of this tool." "What would she be doing?" "Oh, she wants to write a letter to a bank explaining that they are improperly foreclosing upon her house."

Few people understand that those are the same picture - that the limited availability of cognition that can present as the American professional-managerial class has been load-bearing for a while. And now we have cognition that can present as that class without being effectively limited.

Yatharth: Yeah. In my field of healthcare, there are some quite onerous licensing restrictions and paperwork restrictions and shell companies you set up to do something. And there are starting to be tools that are the Stripe Atlas of government touchpoints where they will do the physical mailing for you and track it in one place and have the templating of what to do. I don't know how well they've integrated LLMs, but it's very exciting. It reminds me - like you, I really liked when I first found Twilio, I was like "this is the greatest thing ever invented on earth." Similarly, it's like "Oh, something just got abstracted away. I don't have to touch the thing. I can press a button on my phone and a pizza will come." This was another one of those things.

Patrick: Computers can achieve physical results in the world. One of the least believed things about computers, despite being one of the most obviously true ones. Oh man, you could go for infinite depth on that observation specifically.

One thing that it implies is that the status awarded to people who understand how computers work - and probably the power awarded to them - will be up over the course of the next couple of years. And that greatly discomfits people.

Yatharth: You mean you're talking about like programmers...

Patrick: A frequent observation that has been made to me about the operations of the government, for example - and I think this generalizes to many other respected institutions in society - is that the movers and shakers who decide policy consider implementation to be beneath their notice. They don't talk to the customer directly. They don't talk to the implementers directly. They design the ideal policy in consultation with various stakeholders who are at similar levels at other institutions.

They write up the document and then hand it over, and then eventually it gets passed to grungy little software people. That is not the typical way that software development works at well-operated companies in places like Silicon Valley these days. The programmer spends a lot of time talking to the user directly, and the physical operation of the artifact produced at the end of the day will in large part be downstream of decisions made by the programmer.

Is it a 22-year-old legislative aide at a random member of Congress, or is it someone with an engineering degree who will determine the behavior of edge cases in the United States welfare system for food stamps in 2035? One could bet that it is the 22-year-old. I don't think I'd bet that way. If I'm right, that implies a relatively seismic realignment of where power exists in the United States political system - and the United States would still be a democracy, of course.

Reflections on government software and policy

Yatharth: That makes me think of the impediments, because from my understanding, a part of how a lot of this stuff works is like expensive, complicated procurement processes and whole production and certain companies that are able to get around that. And by the time you're an engineer, there's layers and layers of compliance in your product - might not work, but you are largely not responsible for it.

Patrick: Yep. Grant that all of that is true in the status quo. If one assumes that the operation of the United States government is downstream of laws passed by Congress, then it's important to understand that the production function for Congress passing laws is not that we have 535 people who we sequester in a room to get all their great thoughts about the world.

In fact, each of them has staff and the staff members are descriptively often 24-year-olds, and they have a particular educational background that gets them selected as likely prospects. They write the laws, and eventually those laws get ratified by people who are - the average age is north of 60 these days - who have actually been voted in. But the physical hands on the physical paper is the 24-year-old who just got out of Brown. I won't say they don’t know anything about anything, but that's who physically writes most of the laws.

Then - oh man, we should do an entire episode on software contracting and software procurement at some point, but it won't be today. Yes, the laws eventually cascade down into the regulations that are affecting various departments, which cascades down into how they write the requests for proposals, which cascades down into the marching orders that are given to a consultancy, which cascades down into the implementation team at the consultancy, which cascades down into the Java code that implements the system. That is the United States of America as interacted with by an individual user.

If we decide, "Oh wait, the actual system is pretty borked" and we make some things that will seem like common sense changes - "We have bad software from government, we have good software from Silicon Valley, maybe we should imbibe some of the good software genes, the good software vibes, and the good software cultural practices" - that's going to change the balance of power between people with engineering degrees and 24-year-olds who graduated from Brown and are currently working in an office in Washington, DC.

I think that is relatively well appreciated. And this is why - not all of the animus between American centers of power, but a non-trivial amount of the animus between the centers of American power is downstream of some people seeing the handwriting on the wall that "Oh shoot, the geeks either have their hands on the levers of power directly, or their hands are micrometers away from those levers. Maybe they don't even understand that, but I certainly do. And I do not want them touching that because that limits mine.”

This was the backstory to the Cambridge Analytica affair - blown entirely out of proportion by the security state and the media, but that is what it is. 

[Patrick notes: If one took Cambridge Analytica seriously, it’s an argument that with a team of half a dozen smart people and a few hundred thousand dollars you could suborn a national election. (And further, that we have an existence proof. That part of the story is just made up, and many smart people not only believe it, they believe that anyone who disbelieves it is a threat to democracy. Those people are fools.)

Anyhow, on the theoretical level… look, we have a number of professionals who have run advertising campaigns over their time, right? All of them will tell you: this is nonsense on stilts.]

There's a number of observations that have been made by a variety of people in a variety of places that "Oh shoot, it looks like the tech industry is a concentration of power."

The tech industry absolutely is a concentration of power. It's extremely competent at achieving its aims. That's why we go to work every day, but not everyone loves that fact. [Patrick notes: We spend an awful lot of effort trying to strategically dissimulate about this fact, in part because other centers of power find it threatening and, when it is brought to their attention, attempt to extralegally punish us.] 

And to the extent that we have choices in a democratic system and we can choose to have systems that routinely generate competent outcomes or have systems that routinely generate incompetent outcomes - it is not a law of nature that the incompetent system always wins. Don't know where we go with that one.

Yatharth: I feel like I'm missing some of the mechanisms of how that would happen. I think I just sound like trust and safety platforms and very social media are effectively...

Patrick: Can I give you one very straightforward one? Two people want to work in the upper levels of an American political campaign. Keep the candidate constant for the purpose of this. One of them has a PhD in political science from a well-regarded institution, legitimately an expert in understanding how American demographics interact with regards to the policies of the parties, can talk to you in arbitrary level of depth about Spiro Agnew. One of them is really good at doing Facebook ads. Which of these people should get one position in a modern political campaign, if you only have one position to give?

Yatharth: Oh God, anyone worth their salt takes the Facebook ads guy. Okay.

Patrick: Now people in Washington who are more like the first individual than the second understand that is the rational choice these days and they hate it. And so there are consultants - there's an entire built economy around Washington which will attempt to tell you that "No, the geeks don't understand what they're talking about. No, we just entertained the people who won the presidency for Barack Obama because they are great storytellers, but we didn't learn any lessons there. No, we have all the received wisdom for how one runs a great ground game," etc. They won't frame it exactly that fashion.

The people are going to get ambushed by reality repeatedly on that score. And they have been so ambushed, repeatedly. It's an open question as to to what degree the system accommodates that new reality and hires their own geeks, acculturates them, etc., and to what degree they are simply out-competed in places that have a very brutal fitness function.

One thing that - excerpting a Twitter thread, and I previously made this observation a number of years ago - someone said the geeks will never matter in Washington because the geeks don't understand how power really works. It's very possible that I don't understand how power works. However, I know one thing about elections, which is that if you win the election, you take power. And if you don't win the election, you don't. So capability at winning elections is really important. You can have any level of theory and institutional support etc., and then lose elections and you don't take power. And I think we've seen that a time or two in the recent past.

But be that as it may.

Yatharth: Yeah, the getting people elected thing makes sense. The contribution to what actually gets legislated - I wonder if you have any bit of specific information about that. As an engineer, it's obviously horrifying to me to imagine a codebase that keeps growing and doesn't get cut down. And I don't really understand the equivalent of pull request reviews and who actually is red-teaming all of the little edge cases in all the laws. I don't know, it seems like maybe that just concerningly doesn't happen, at least, or maybe the lobbyists do it. I just don't know what that process looks like.

Patrick: Yeah. Long, complicated ball of wax there, and I don't know if we have enough time to get into it. "Literally no one knows" happens much more than you would expect. And then sometimes there is a very ponderous process.

I love the phrase "loophole" - the popular conception of a loophole is this is designed behavior where some evil outgroup - the lobbyists or those bastards in Congress etc. - they met in a smoky back room and they engineered behavior of the system to be pathological.

What not too infrequently happens is the system of behavior is engineered, and then people realize what the system of behavior is. Code is publicly inspectable and saying "Oh, there is part of the system of behavior" - whether it was intended or not perhaps up for debate, whether it exists or not far less up for debate - like here's the sentence, and then they drive a truck through that.

Then frequently the reaction to driving the truck through that takes years to catalyze itself because these systems are often not very monitored in real time. And then we have an iterated version of the game where we say "Okay, we're going to pass a new law, write a new regulation or similar" which will not retrospectively make illegal the thing that happened (because retrospective application of law is a thing that our legal and political system are extremely allergic to due to potential for abuse and a history of actual abuse).

They will say going forward, we don't want that happening again. So here we go. And then who does that? Lobbyists, actors in society, legislators, courts, etc. The system of government that we live under is complicated and perhaps not fully described by anything other than itself.

But anyhow, we have been talking for quite some time now. Do you have maybe a wrap-up kind of question for me?

Yatharth: I guess what I was going to say is I would love a topic on this idea of just red-teaming the law, or just zero-day security vulnerabilities in the law - all of this stuff. From my experience in crypto, it's like "Oh, you let a vulnerability sit for like two days and then people are going to exploit the heck out of it." How come that doesn't happen with law? I'm just understanding that whole process, and it probably does happen.

Because from, like you said, the layman understanding is like telling these stories about lawmakers being really dumb or malicious or whatever. And the shift that I found for myself in reading your blog about finance is like, these processes that seem magical have become less magical. It lets me live by the whole "God, give me the power to give a fuck about the things I can unfuck, not care about the things that are completely fucked, and tell the fucking difference."

Patrick: Not the usual way the Prayer of St. Francis is phrased, but…

[Patrick notes: It appears that the prayer, while popularly attributed to St. Francis of Assissi, was actually written by Reinhold Niebuhr, a Protestant theologian. Like many prayers in both faith traditions, the exact wording depends on the source or speaker; the version I am most familiar with is “Grant me the will to change what can be changed, the serenity to accept what cannot be changed, and the wisdom to know the difference.”] 

Yatharth: The one that's stuck in my brain. But your writing on the finance system helps make that attitude with respect to my interactions with the financial institutions possible. I'm very calm when I don't know what to do - I know what I can do and it's just most pleasurable. And I'd love something like that to just reduce the amount of inflammation with the government. So that's just calling it out as a feature episode.

Sabbatical insights and future interests

And otherwise to wrap up, I'm just curious in the sabbatical of yours, if you have had other things you've wanted to poke at, kinds of people you want to talk to, surprises you've encountered, or changes to that plan thus far.

Patrick: Sounds like a joke - not a joke. "Factorio." I've spent a surprising amount of my sabbatical on both Factorio Space Exploration, the mod, and Factorio Space Age, the new spiritual sequel.

Yatharth: You mean surprising to yourself, right?

Patrick: Surprising to myself. It would probably also surprise many people who think I'm an adult professional when I say that I've probably played 800 hours of a particular video game over the course of the last year and a half. Very edifying experiences for systems thinkers.

So if you've never played Factorio - it has my qualified endorsement. On one hand, incredibly edifying for systems thinkers; on the other hand, somewhat dangerous. It's a bit of an info hazard.

The non-joking joke out of the way - what would I love to understand more than I do? We were talking a little bit before the recording started on medical insurance. That is an issue as newly becoming relevant to my interests. The American medical insurance system is very different than it is in Japan, and so I'm now a user of it. And on one hand, complicated political economy around it - we want it to do many things. On the other hand, it's an actual system that exists that had to be designed, and why did we design it that way, and why do we find difficulty in pushing it into potentially other designs? Something I would love to explore.

I would love to talk to an expert about government software procurement at some point, which I think is downstream of a lot of the infelicities around the systems that we use to interact with the government at retail touchpoints, because those are increasingly software-mediated and you saw a lot of that during the VaccinateCA experience.

And broadly interested in talking to people who find these subjects interesting. So if that's you, drop me a line.

Yatharth: Awesome. Thanks so much. And if I could drop another tidbit - I would love to hear more about not just the American financial system, but like the Indian one, which does a lot of fascinating things around credit cards and UPI and payment rails and various other counterfactuals.

Patrick: This is sad. Something that I attempt to do in my writing is to keep less of a blinkered perspective on it. There is some amount of difficulty for me in writing about UPI directly, partly for "the usual reasons," but partly I simply don't have the level of understanding of the UPI or the political economy of the Reserve Bank of India, or for that matter, the political economy of the larger Indian political system.

I generally need to interview an expert before I can write about those subjects in depth. But since there are plenty of experts on those things, since the international comparison is a huge source of advantage for those subjects and many others, I would be very happy to chat with folks.

I think, by the way, that taking things that are well understood in X and exporting those understandings to Y is underappreciated as a source of value by many people in the world, inclusive of many intellectuals. Which is funny because it's received wisdom in the American PMC - "diversity is wonderful, you travel to learn things about other cultures," etc. But our actual actualization of that is pretty poor.

If you yourself listening to this podcast are wondering what is a repeatable way to find things to write or speak about - if there is something in a particular community that is well understood that's not understood in another community that you belong to, simply writing up what the first community understands in a way that the second community can appreciate is basically an infinitely repeatable script for value generation.

For a number of years for me, this was - I was the guy who did basically nothing but write "Okay, I'm going to take a 100-year-old research result from the field of marketing and write it such that an engineer can understand it and not think it's evil." That's it. And that was basically career-making for me, repeating that one trick repeatedly for about 10 years and no one ever caught on that was what I was doing. [Patrick notes: I’m somewhat tongue in cheek here. Somewhat.] Even when I said at the top "I'm the engineer who understands marketing. Here's the thing that the catalog marketers understood back in the 1920s, it's called A/B testing."

There exists blank space in the world, even with the availability of magic style transfer machines, for someone to take what India has done in RBI and what Brazil has done in PIX, etc., and say "Hey, these are very interesting payment systems that are relevant to the United States of America, particularly as it attempts to roll out FedNow and similar." And there is a different substrate with respect to consumer protection legislation in India and the operation of UPI would probably encounter some challenges under Regulation E, etc.

I'm not the person to write that article. Someone out there is - consider writing it or consider talking to me about it. Anyhow, there are a variety of sources of knowledge and wisdom in the world. Importing and exporting between those pools of knowledge and wisdom remains underrated, probably always will be.

So feel free to take advantage of that.

Yatharth: Cool. If you're those people, please send an email to Patrick at any of his domains because I would certainly love to hear about this sometime.

Patrick: All right. Thanks very much for the reverse interview. And for the rest of you, thanks very much and we'll see you on the internet probably next week.