I'm an academic economist and I've been bullish on crypto for a few years. The thesis is simple. Crypto is a tool almost exclusively for criminals: it's basically useless to anyone in the developed world. But it's a really good tool for criminals
Crypto does, however, link to a very fundamental question: finance, as we know it, is a derivative of legal systems. Finance is the trading of promises, and promises only work when people think they'll be enforced
Finance has historically lived in hubs like NYC, London, Hong Kong, because these places have the state capacity to enforce promises. Finance doesn't meaningfully happen in the developing world, largely because there isn't the state capacity to consistently enforce promises
What is deeply interesting about crypto is that crypto is an alternative promise enforcement system. Anyone with a Starlink in Nigeria or Vietnam or South America has the exact same guarantee of BTC wallet promise enforcement as someone in NYC or London or Hong Kong
I discuss this in a short blog post: should be easy to find on Google, but DM me if you want a link. I ended the post with a really dumb, but potentially interesting, prediction about where crypto - and finance more generally - will go in the next few decades
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I feel like a substantial amount of creativity comes literally just from knowing a lot of stuff. If you know like, three paintings, history looks like a linear progression from painting 1 to 2 to 3. If you see like 3 million paintings, you have a richer understanding of history
You see not only the famous things, but also all the smaller ones not taken, that could have been famous, but somehow missed the bus, you draw connections that haven't been drawn, revive forgotten stuff, and so on
My basis for this is that (IMO) creativity among PhD students in our field is surprisingly predictable. The people with richer life experiences tend to be more creative. I don't think this is psychological, really. I think they just have more material to draw connections from
Conjecture: econ academia is very far from the optimal architecture for data research. The optimal architecture is much more centralized. Half the researchers work together, or in large teams, to build clean and efficient datasets. The rest work in smaller teams to analyze data
All empirical research is based on data. In the current architecture, building and maintaining datasets is under-rewarded. Some of this work is done by government workers, who are generally (and increasingly perhaps) underpaid and under-appreciated
Some of it is done by private teams. But because ultimately findings are published and rewarded - not data - the incentive is to keep your data private, writing at least a few good papers with the data, before making it public
IMO, a large driver of how productive econ/finance PhD students are is work habits. Basically, can you set and maintain realistic goals that interpolate between having nothing, and having an R&R'd or published paper. This is not taught, and it's surprisingly hard
The issue is, if you go directly from school to PhD, maybe with an RA-ship in between, you're used to having tasks handed to you in like 1-2 week increments: psets, exams, RA work, so on. You've never really had more autonomy than this
PhD is very different! Your goal is to produce papers in 3-ish years, from year 2 to when you're on the job market. The right strategy is to:
1. Divide this into 1-2 week tasks 2. Do the tasks
The thing I'm realizing is lots of people don't get how revolutionary the o3/o4 class of models are, for (econ) research. 4o and the like models are borderline useless for econ research-level math. Deepseek is borderline useful. o3/o4 are at the level of solid PhD students
o3, o4 represent a step change in how I use them. Prior to o3/o4, I used them occasionally for tasks here and then. After o3/o4, I use them in general as basically the first thing I go to, for more or less everything research related I'm doing: coding, model-solving, writing
Literature review, evaluating paper/idea impact, brainstorming, so on and so forth. There is nothing a human student/coauthor can do, that it cannot do a pretty good shot on. It functions basically like a very good generalist coauthor on everything
Putting kids through too many status games makes them risk-averse. They go through middle school, high school, college, every step doing the "right" thing. They go into a world with a deep fear of ever accidentally stepping off the "golden road"
They see peers, who they see as "less successful" than them - less badges and accolades - make choices they would never have made; and judge them, and think how horrific it would be to ever fall so low in life
Some of these peers will end up overwhelmingly, outstandingly successful. They attribute this essentially to luck; otherwise, how could someone so much "less good" than me do so much better than me?
I'm reading Holmstrom 1979, and I'm a lazy student who doesn't want to actually do the work, so I'm making ChatGPT do it for me. It is very simple, I just give o4-mini screenshots of parts and ask, wtf is going on. It tells me exactly what Holmstrom is saying, much more clearly
I can give it literally the laziest, most notation-insensitive questions of "wtf is going on" and it just puts up with my shit. My TA would kill me for even asking a question this way over email. But, until the AI overlords come to punish me for my sins, o4 just cooperates
Derivations skipped? No worries! Ask o4-mini to do it for you, intuitions and all