The economy is made up of people, none of us are perfectly informed, and given that we can't even agree on what behaviors are rational it would be very surprising if we were rational all the time. So it feels like a good explanation for any economic phenomenon that doesn't make immediate, obvious sense is that whoever is responsible is behaving irrationally. But in practice, as with inefficient pricing, this doesn't work as a null hypothesis that must be believed by default. Instead, the real game is to find some instance of bounded rationality where people persistently following their incentives will be led in a predictably wrong direction. And the reason that's the case that needs to be made is that the economy is incredibly complex, and abounds with local knowledge that isn't available to outsiders. Think hard about any business, make a list of everything they do that doesn't make sense, and there's a good chance that what you've inadvertently done is made a list of everything the company does that only makes sense to experts.

There's a free market maximalist view that the economy is actually either perfectly efficient, or close enough that there aren't any realistic opportunities to improve it. But even if that's directionally right, the strong version of it falls apart the same way the strong form of the efficient markets hypothesis does: if that's the case, there's a whole lot less to do—market research doesn't make sense if consumers' spending is already perfectly optimized, and in fact any kind of research presupposes the existence of unexploited opportunities for gain. There are clearly frictional costs involved.

And, in fact, the importance of these frictional costs was highlighted by economists who were working within a free market tradition: they used models as a rough descriptor of the world that, crucially, enumerate its assumptions, so you have a limited search space for why the model doesn't perfectly describe reality. In other words, they started with a model specifically so they'd have a good starting point for critiquing that model.

One of the paradoxes of perfectly-efficient-market thinking is that many of those same successful companies got that way by bucking some industry trend. Today's experts are the people who did something differently from the previous generation of experts. Cloud computing exists because the on-prem model has limitations; search is good today because it was bad in the past; early social networks weren't good at figuring out what their use case was—for some people, MySpace was a place to be themselves, and for others, it was a place to be someone else entirely, so the overall user experience is like being invited to a party and finding out that half of the invitations said "cocktail party attire" and the other half said "wear the zaniest costume you've got."

But MySpace also did something right and performed a public service even though they didn't make their investors all that much money.¹ They demonstrated that there were still open questions in social. The obvious bet based on time spent online and on how quickly the Internet was devouring business communications was that eventually, some fraction of our social lives would be digitally-mediated. But different companies had different visions of that, and different taxonomies for describing the world: for LinkedIn, you're embedded in a network of connections to people, but also connections to institutions and skills. For Twitter, "you" are disconnected; in Twitter's vision of social media, the atomic unit is the message rather than the user.

Even when a company is motivated by an idea that isn't a direct critique of a competitor, framing themselves in those terms is inevitable. Reddit's founders say that they weren't aware of Digg when they started, even though both companies were in the business of vote-based rankings of popular news stories. But once they were aware, each company had to make product and moderation decisions in reference to the other. Reddit won by being more freewheeling than its competitor, which meant that the site spent a decade-plus keeping its fractious civil libertarian userbase in line as its scale forced it to adopt more conventional decisions about what kinds of content it would host.

Sometimes companies end up in a sort of battle of the framings: there was a time when Chrome looked like it was competing with Internet Explorer, but it turned out to be competing with Windows instead.

And over the course of all of these different developments, everyone involved had to have a working theory of what role their product played, because that told them what users wanted and how to monetize those users' behavior. And all of these theories eventually went by the wayside; the time any product is perfected is also the moment that it's being replaced by a different framing for what the boundary between different products ought to be. Right at this moment, for example, AI agents tend to operate within the browser, and their pitch is "do what your browser did, but through natural language." But that pitch is going to be as durable as the paradigm that TV was a way to listen to your favorite radio stars and also see their facial expressions. Bounce through random bits of this 1948 episode of what was then the fifth most popular show on earth and you'll almost never see the camera or the actors move. Just porting the old paradigm into the new one is a great way to get started, because everyone knows what to expect, but if the new model is really new then what it's capable of is a superset of what the old one does, and some of the old model's functions will simply disappear.

You can think of a thesis as a way to ensoul assorted facts and start making them dance in the right direction. Even if the thesis is wrong, acting on it means testing it, and that's the only way to get it right.

The Diff is, in one sense, an ongoing effort to throw out a bunch of hypotheses, refine the incorrect ones, and double down on the important ones. Some posts that cover this:

• You And Your Investment Research

• Investing as a Rationality Dojo

• Idea Velocity

• Bet-the-Company Ideas

• Predicting, and Knowing What to Predict

Join the discussion!

1  At least in the long run; at one point, an analyst pegged MySpace's value at $20bn, the year after News Corp. acquired them for $580m.