HODL'ing Bitcoin is a Human Action.

...and AI can't do it.

While attending the Bitcoin 2023 conference in Miami, I was asked a provocative question by a newly formed friend I'd met and eaten dinner with on Saturday evening.  He asked what I thought about whether Artificial Intelligence would end up accumulating all the bitcoin once they figure out what an obviously valuable asset it is.  I was intrigued by the question.  My initial reaction was that HODL'ing bitcoin is a uniquely human action.  The thing is, even amongst the humans, the vast majority of bitcoin that has ever been acquired has changed hands several times over.  If humans have a hard time hanging on to their bitcoin, why would AI be any different?

Human action cannot be modeled

The question, to me, isn't actually about AI, but about praxeology, the study of human action .  Praxeology, like AI, is concerned with predicting what actions a human being will take on an "a priori" basis, or before the fact.  The great Austrian thinkers have dumped gallons of ink disabusing people of the fiction that human action can be predicted before the fact.  Human action is based on an individual's unique set of circumstances at a given moment and, at that moment, the largest source of dissatisfaction that can be removed at the lowest cost.  It might be someone in a desert that turns down all the diamonds in the world for his first glass of water, but would clearly not have the same value system for a second glass of water.  Perhaps AI can make this simple deduction about what a human being might value, but to think that AI could model every single human value decision, or even a "good enough for government work" version very much strains credulity. AI is far more likely to influence people to make the decisions it predicts they will make simply through the power of suggestion than it is to predict human actions with any accuracy.

There are fun science fiction scenarios like the one where Satoshi is actually an advanced AI from the future that travels back in time to create bitcoin only to go back to his time after getting it off the ground.  Science fiction aside, AI may get better and better at emulating a human being.  But they will more likely mold humans into acting the way AI suggests they should.  People with low resistance to psychological operations (psy-ops) will become the most predictable.  These are the people who pride themselves on being able to command their Siri or Alexa.  People who HODL bitcoin are on the opposite end of the spectrum and are highly resistant to psy-ops.  I assert that these will be among the last group of people that AI will learn to emulate.  Not only will HODL'ers be difficult for AI to predict and understand in general, their propensity to HODL bitcoin will be particularly difficult for AI to understand.  AI will have very little credible data with which to train their prediction models, as Bitcoiners tend to have far better Operational Secutiry (OpSec) than the general population and leave much less of a digital footprint for a model to learn from.  The majority of data that AI will have to learn about Bitcoiners will come from people who willingly provide it, which will give quite a false impression into the nature of the actions of Bitcoiners.  AI's falsely attributed success in predicting the actions of the weak when they actually trained them will reinforce its ability to predict all people, and will make it even worse at predicting the actions of the free-thinking and independent.

Of all the reasons why AI will struggle to understand and predict the actions of Bitcoiners, perhaps the most difficult of those actions to understand will be that of HODL'ing itself.  AI might well understand why some people buy bitcoin, and why they sell bitcoin, but I would guess that it has no way of understanding why people HODL.  Human fiduciaries are required to sell bitcoin on behalf of the trusts that they represent, unless explicitly specified by a will.  It is recognized as an asset that is too volatile to hold prudently.  The conventional interpretation of what human beings do with bitcoin is to sell it.  AI may never be able to truly willfully HODL because HODL'ing is a uniquely human action.  On the internet, we get asked to prove we're human by submitting to a variety of actions like checking boxes and clicking images, but it may be that all we need to do to prove we're human is to post some bitcoin, proving that we've held it.

Institutions can't and don't act like humans

Extending the fact that HODL'ing bitcoin is a uniquely human action comes a difficult fact, that institutions can never hold bitcoin with the conviction that an individual can.  Many bitcoiners are waiting for Apple or Facebook to put a bunch of bitcoin on their balance sheet and then watch the price moon.  Some think it will be governments who will get into the game.  Others are waiting for pensions and endowments to commit some of their asset allocation to bitcoin.  They are all misguided.  Even bitcoin institutions like publicly traded mining companies eventually succumb to the pressure of selling their bitcoin.

What about Microstrategy and Tesla?  What about El Salvador and Bhutan?  What about all of that pension money? (and this recommendation?). Or endowments?

Companies

We all know the story of Microstrategy and Michael Saylor and their 130000 bitcoin on their balance sheet.  Michael Saylor understands bitcoin like no other human being alive and the move will likely be remembered as the smartest thing a company has ever done.  We also remember Tesla adding a significant bitcoin position in 2020, only to dump it in 2022 after we found out that Elon Musk was a shitcoiner who never actually understood bitcoin for a second.  It turns out that Tesla is more the rule and Microstrategy the exception when it comes to Bitcoin.  Microstrategy is an institution that is actually an extension of a human being, Michael Saylor, who founded the company and is the majority shareholder.  In essence, he is Microstrategy so the institution has the characteristics and abilities of that individual.  In a more typical company, there is a CEO but they have very limited power to pull off what Michael Saylor did.  They report to a board and are far from the majority shareholder.  Companies have to report quarterly earnings to their shareholders under a GAAP accounting standard that ignores inflation prospectively.  (The present value of all their assets and liabilities use forward rate curves that reflect current market prices, ignoring future inflation).  What it comes down to is that a company will always have a higher time preference than an individual.  Even if the individual has the power and courage to try what Elon Musk did at Tesla, the pressure to eventually sell is just too great and the high time preference will always lead to letting go of the bitcoin.  Even Michael Saylor had to step down as CEO and reduce his overall role in order to be able to focus more on Bitcoin.  Don't hold your breath waiting for Google or Microsoft to put bitcoin on their balance sheet, and if they do, don't count on them keeping it there.

Governments

El Salvador is the Microstrategy of governments when it comes to Bitcoin.  Nayib Bukele has a superior understanding of both Bitcoin and his government and his popularity enables him to unilaterally extend himself upon his government.  Even with all of this in place, we don't know what El Salvador will look line in a few years when Bukele is no longer the President.  Bhutan is another country that seems to have been accumulating bitcoin for a while.  While these have been rare exceptions, I would expect to see more governments accumulating bitcoin than companies.  Countries can have a lower time preference from companies.  Most do not and end up endlessly borrowing from the IMF and World Bank in a forever-cycle of defaults.  Western democracies have generally high time preferences as the leaders are elected every few years and have little incentive to encourage or seek out long term investments.  

Pensions and Endowments

For the same reason that companies won't hold bitcoin long term on their balance sheets, pensions won't hold bitcoin in their trusts.  A pension has even less of a connection to an individual actor than a company does.  It's decisions are made by a Board or Committee that may or may not even be participants in the pension.  At best, Bitcoin shows up as an investment opportunity but as fiduciaries, pensions open themselves up to lawsuits from participants if they allocate assets to bitcoin and it doesn't perform well.

I recommended that pensions replace their bond positions with Bitcoin on August 7, 2022 in the embedded tweet.  We all know in the middle of 2023 that this would have been a Godsend for pension funding as Bitcoin is up over 20% in that time while bonds have cratered in value.  The reality is that anyone who followed this recommendation would have been under extreme pressure in November of 2022 when Bitcoin's price in US dollars fell to as low as 15K, which represented a 30% loss.  That pressure just makes it untenable for institutions to hold their bitcoin position long term.  

I had a conversation about Bitcoin around August of last year with the Dean of my childrens school.  We both agreed that the school should hold Bitcoin in their endowment but when it came time to think about talking to the Board about it, I pointed out how difficult it would be to hold if it goes up 2x, 5x or 10x.  The pressure to sell is just too much for most institutions to bear unless there is a really motivated human behind it.

Unions

As I write this, the Santa Monica Fire Fighters have just announced that they are buying Bitcoin and holding it in self custody.  While its hard for me to get details, this section will be a placeholder in the event that Unions demonstrate an ability not shown by the other types of institutions.  More to come on that as more details emerge.

Conclusion

Artificial Intelligence can not predictively model the behavior of Bitcoiners and will likely only get worse at it as time goes by.  The false signal of accuracy that comes from the people who have altered their behavior at AI's suggestion will feedback into AI overfitting that type of person in their model.  Bitcoiners are resistant to AI suggestion, and they don't provide a dataset that can be trained for a robust predictive model.

Institutions also will do a poor job of emulating human beings when it comes to HODL'ing bitcoin.  Institutions have an inherently higher time preference and will always fail to resist the selling pressure before an individual.  There are exceptions where institutions are extensions of individuals like Microstrategy and El Salvador, but by and large, institutions will be inferior at holding bitcoin relative to individuals.

Holding bitcoin may likely become the singular way that people prove they are human in the digital world as only humans will be able to hold through the times that are coming ahead.