Collapsing prices and valuations are just a symptom of an innovation misallocation problem: fixing the wrong thing. The good news is that fixing the right thing is weird and difficult, and therefore profitable.

The wrong problem

Finance, like every other social technology, is made of protocols: combinations of regulations, institutions, customs, formal standards, and one-off decisions now sunk into the landscape itself that make it possible —in some senses give meaning— to things like "having money in an account," "owning a stock," "generating a money-like asset," "telling your bank what to do," or "executing an option." The main value of these protocols isn't that they are sophisticated or efficient. They rarely are. What they are is boring: they have been performed so often and for so long that their failure modes have been mostly exposed and patched up, they are relatively well documented, and there are large tooling and service ecosystems around them. A SWIFT transfer is something akin to a TCP connection: it's not the best imaginable pattern for two entities to interact with each other, but it's good enough and everybody uses it, and those are the two necessary and often sufficient conditions you want in a protocol.

A quick way to summarize the past wave of fintech innovation is that technically minded people took a look at financial protocols as a whole, and found them antiquated and inefficient. And they were right: many basic contemporary financial protocols are old not just by computer standards but by sociological ones. So they went and build more efficient, or at least less human-intensive ones:

• Mobile apps with simpler protocols to pay for things, send money to other people, or trade stock options.
• Distributed cryptographic ledgers to record ownership changes without having to use stodgy legal protocols.
• Smart contracts, stablecoins, tokenized-everything, etc, with new protocols around concepts like "asset" and "money."

Some of these new protocols were immediately taken over by the traditional system - cloning being the sincerest form of flattery, bank apps became, if not great, at least immensely better. Robinhood bloomed. Other advances, though, like Bitcoin, they dipped a toe into and bet some chips on, but generally not in any structural way, which seems wise in these difficult days that some feel are the Twilight of the Coins.

What made some new protocols catch up in a sustainable way and others fail to was, essentially, whether they improved a bottleneck. As conceptually fascinating as distributed ledgers or stablecoins can be, the way we track ownership and handle monetary engineering is, by and large and with some pathological corners, good enough that although you could make money on the enthusiasm behind rebuilding them, the financial system wasn't particularly tempted to do more that gesture towards implementing them in a fundamental way, and they haven't provided enough of an advantage to replace them by sheer force of competition (in fact, when their tires were kicked, the new protocols proved to need more of the "legacy" legal system as fallback infrastructure than it had sold itself as needing).

This became clear a while enough, and with the promise of cannibalizing the existing system deeply devalued, external shocks were enough to push down most of the ecosystem and most of the enthusiasm outside of the core support group.

To return to and justify large forward-looking valuations and interest, fintech needs to focus not on drastic improvements to protocols that work well enough but on drastic improvements to protocols that are bottlenecks to systemic growth.

Bottlenecks being specific to goals, I'm going to focus on finance as an enabler of other long term systemic innovations. In other words: How can fintech fix bottlenecks in the way the most significant large-scale innovations are conceived, implemented, and profited from? If we are betting on, if we're counting on a world of fast and accelerating technological development, fintech needs to shift from benefiting from the halo effect of these expectations to being a practical driver of their implementation. Smart contracts vaguely related to speculative improvements on relatively functional protocols won't cut it.

The bottleneck is obvious, though, when we zoom out slightly from the finance platform itself.

The right problem

Let's look at anti-aging technology as an example —it could be electric batteries, neural interfaces, or anything along those lines— in a deliberately over-simplified way: this isn't an article about any emerging technology field, but on how how we interact with them, which begins with how we think about them and starts going bad pretty much there.

You log into your state-of-the-art financial platform and find out that Acme started developing an "anti-aging" Compound X that in laboratory experiments activates mechanisms associated with exercise. You have at hand vast amounts of information about their price history, market signals, etc., but the most important fact about Acme is how Compound X will work: your investment is a bet on biochemistry through the protocol of finance, and the purely financial data in your platform isn't strongly informative about it.All the AI in the world can't change this. In a fight between Machine Learning and Information Theory, always bet on math.

So what do you do next? You look at blogs. You check the news. You browse papers. You dig into forums. Essentially, the platform handles for you most of the cognitive load of financial analysis, finding counterparts, handling payments, and so on, and, except for maybe showing you a tab with relevant news you could have searched for elsewhere in seconds and perhaps one or two in-house writers, leaves you alone with the problem of evaluating the very thing that makes your investment potentially valuable for you and for society.

It's important to repeat the point because it's such a normal-seeming aspect of the process: we have very powerful and, when we want them to be, efficient financial platforms that help us with everything except understanding the underpinnings of what we are investing in. Not, at least, with the sort of sophisticated cognitive tools they offer about the financial protocol itself.

In short: you can, if you want, use everything up to AI models to understand a time series of prices, but for biochemistry it's up to you to separate the Twitter hashtag from the meta-analysis published in a scientific journal (not that those don't require their own form of critical reading).

You could spend months reading social blog posts and not properly understand, or not give it enough structural importance, that the Acme bet isn't Compound X works but rather, oversimplifying, Compound X activates a certain biological mechanism associated with exercise and this mechanism, activated in realistic conditions, has a positive and significant marginal impact on aging-related outcomes (and besides the side effects aren't worse).

That would be a pedantic rephrasing if not for the fact that there are relevant information about those each of those dependency links that can only be found by a potential investor by looking through the the literature and tracking them down, but that are as essential to the business case as any cash flow projection, so should be just as visible in a properly structured way.

The well-known stories about Acme's main investor, Willy EC, are likely to be more salient in the online material that the more informative fact that Compound X is meant to be an activator of a biological mechanism, so if it doesn't activate it correctly in the right cells in the right way in the right context, etc, it won't work; all the tests in animals can't help you much there, because most of what works in, say, mice, works poorly or not at all in humans (an extremely well-known thing to everybody involved even remotely in biology, but how many billions are invested betting against equally basic constraints?).

Or maybe the activator works, but "associated with exercise" can mean "if you turn it off exercise doesn't help," not "if you activate it you get the benefits of exercise." Or maybe it does provide 5% of the benefits of exercise in some aspect of biology, and the impact on aging is negligible. Or it works in the way it was expected to but multiplies your chances of cancer. Or, most commonly, it mimics some of the effects of light exercise but it doesn't add anything if you were already doing it.

The point is, there's a complex web of scientific and technological dependencies for any speculative investment and

1. There's usually a technical consensus for what the links are.
2. There's (possibly partial or uncertain) information about the links themselves.

Any mildly innovative investment will have a complex structure behind it, and each node may not be on itself well-known, but they will nonetheless have a narrower and more robust range of uncertainties. More interestingly, investments in the same areas will share many dependencies, so if you have an specific view on anything from clean room technologies (a key to scaling up cultured meat) to the practical energy efficiency limits of large-scale carbon capture (possibly problematic), it has an immediate impact on how you would look at any number of potential investments.

It may be best to explore this idea through objections to it; after all, if it's almost never done, then the objections are the default state, and it's up to this text to try to convince you that a new protocol would not just be better but also better enough.

So that's it, just show dependency diagrams behind investments? "Yes" in that structuring a financial platform in this way would be a huge step forward in terms of helping users make informed trading decisions, "No" in that what I described is essentially a Bayesian network, which means you can apply computational power not to extract predictive water from time series stones or to listen to the social media deluge for whispers of future changes, but instead to get the best possible conclusions from whatever information and guesses the relevant experts and more specialized markets have come up with (or the hypothesis the user has asked you to consider). The idea is to transform financial platforms from a mixture of quantitative financial data and textual analysis, marketing, and gossip into a more powerful tool to analyze technological futures.

Why should this be the job of a financial platform? It doesn't need to be, no. Maybe news organizations will get first to it. Maybe universities will take a stab. Probably Google. It's not who has the obligation to do it, it's who gets there first and who gets pushed further down the stack. With the cost of building a financial platform having fallen through the floor (why else would the valuation of so many of them collapse?), differentiators are becoming necessary, and knowledge —and more so knowledge expressed in non-textual structural ways that inform investor behavior— is a powerful one. The opportunity view: Being a financial platform with scalable AI-first domain expertise is a better business than being one without.

Isn't this just investment advice with different math? Very much no… well, almost entirely no. There's no true emerging technology investment for which the science and engineering are settled enough that such a dependency graph won't have huge ranges of uncertainty on many nodes: that's what makes them emerging technologies. What finance platforms can provide is a way to structure this uncertainty so a potential investor has a clearer idea of what is known and what they are betting on, and if they have an information or analysis advantage, how to leverage it best.

Is that sort of thing enough to improve decision-making? In most fields, yes. In a field rife with hype and obfuscation, an emphatic yes.

If you can do it, why not just do an investment fund? It's one of the ways in which they are built. If one selling point of traditional fintech is giving everybody access to the same data that professional investors use, you can reframe this idea as giving everybody access to the same meta-cognitive tools (some) professional investors use.

How can a platform offer access to a wide array of instruments if it has to build and update domain models for every group of them? I admit it, it can't - not without huge resources. But "Trade All The Things" is by now almost a baseline feature; "Trade All The Things, and Some In A Smarter Way" or even "Trade These Things In A Smarter Way" is an under-exploited competitive advantage (in practice if not always in marketing). Hard-to-build information integration models about specific emerging technologies are, indeed, hard to build, but that makes them a competitive advantage once you have them. It's impossible for anybody to compete on what users can trade with, but there's still an early mover advantage in competing on what users can think better with.

Won't this mean much less volume on the very bad investments? Perhaps; not because a platform of this sort would forbid anybody from doing any trade, but rather because trading, and talking about trading, is for some a social activity not an analytical one. The opportunity view, on the other hand, is that there's already plenty of competition for the Robinhood/Reddit crowd money, but there's a deeper, almost unlimited pool of money looking for ways to benefit from still more radical technological transformations (and some of it actively looking to push them forward). What's failing right now in our financing of the future isn't the protocols integrating supply and demand information, but the protocols integrating all the rest of it.

If you think I've stacked the deck by ignoring a certain objection (I haven't, I just didn't think of it), let's talk about it.

The takeaway, in three points:

1. The hard problem in finance is not how to invest in something but what to invest in to push for and/or benefit from changes in the world.
2. Most financial platforms have very late 20th-century information architectures to help customers, entrepreneurs, or even themselves, figure it out, and don't really have the sort of mindset to attempt to do so.
3. Attacking hard problems most competitors don't have the mindset to attempt is a high-risk/high-reward strategy, and in bad times it's not riskier than playing it safe.

Post-credits scene

• For an example of the sort of detailed analysis that should be baked into any financial platform geared towards emerging technologies, see this (skeptical) analysis of the technological dependency path of cultured meat. You could summarize this article as saying that your financial platform should show you these dependencies and assumptions without you having had to read it.
• I was vague about the process, mathematics, and technology of building something like this not because it's a mysterious proprietary black box, but because the less understood part isn't the how but the why. In an entirely unintended pun, for an introduction to the field of causal modeling, one of the basic conceptual underpinnings of what I did talk about, you can do much worse than reading Judea Pearl's The Book of Why.
• If you're doing something along these lines, drop me a note to chat about it.

None

None

None