Blockchain has come a long way since its 2008 debut as the ingenious technology underlying Bitcoin. Today companies and governments the world over are exploring new applications in which blockchains track, not the flow of digital currency, but things and events in the real world. But this tether to reality is enterprise blockchain’s Achilles heel.

To see why, let’s go back to where it all began. When bitcoin first appeared, a common reaction was: Why would anyone exchange their hard-earned dollars for that? The answer, it’s become clear, is that bitcoin is valuable, and that is a good reason to own some – or, better put, to hold some. How much bitcoin you and everyone else holds is recorded in the Bitcoin blockchain.

But this is a funny kind of “record.” Most records memorialize real facts, facts that are independent of the records themselves. Take your birth certificate. It records a real event, your birth, which exists apart from the certificate. This is why it is possible for your birth certificate to be inaccurate. It could simply get the facts wrong about when and where you were born.

Bitcoin is different. The blockchain records everyone’s bitcoin holdings, but those holdings are not independent of the blockchain. Quite the opposite: How much BTC you hold depends entirely on what the blockchain says. (Philosophers call this kind of dependence “metaphysical ground.”)

So calling the blockchain a record of people’s bitcoin holdings is misleading. There are no real facts, existing independently of the blockchain, about how much bitcoin everyone holds. Instead, it’s better to say the blockchain just dictates how much bitcoin everyone holds. What it says, goes.

This means the blockchain has a remarkable property: It is guaranteed to be an accurate guide to people’s bitcoin holdings. Because the blockchain dictates how much bitcoin you hold, it can’t fail to be right in what it says about that.

See also: Brady Dale - Don’t Follow the Suits, Follow the Weirdos

I’ve mostly been talking so far about blockchain’s application to cryptocurrency. It’s in those applications that the blockchain is guaranteed to be accurate (bracketing so-called double-spends or other rare events). But there are many other applications that lack any such guarantee.

Take a case that has already come up: birth certificates. Births are real events in the world. We can design a blockchain to record those events, but the events themselves are independent of the blockchain. Unlike your bitcoin holdings, no blockchain can dictate when and where you were born. And that means it is possible for the blockchain to be inaccurate. Nothing prevents the blockchain from just being wrong about your birth.

Of course, just as it is very hard for anyone to tamper with the transaction statements that make up the Bitcoin blockchain, it will be very hard for anyone to alter your birth records once they have been added to the birth certificate blockchain. But that does nothing to ensure those records were accurate in the first place.

The same goes for many other enterprise applications of blockchain. Consider supply chain management. We can design a blockchain to track diamonds from mines to jewelers, but that does not mean the diamonds are where the blockchain says they are.

Or consider health care. We can record on a blockchain that you have been vaccinated against COVID-19, but that doesn’t mean the shot has really gone into your arm.

So we have two big groups of applications of blockchain technology: the ones in which the blockchain is nearly guaranteed to be accurate, like Bitcoin; and the ones in which it is not, like birth certificates. Call the applications in the first group accurate and those in the second group non-accurate.

I don’t want to say that we should only ever pursue accurate applications of blockchain. Some non-accurate applications can still deliver accuracy, if they’re supplemented by something else. For example, if we have some way to ensure that birth certificates are accurate when they’re added to the blockchain, we can be sure they will remain so – even though the blockchain on its own doesn’t guarantee their accuracy. And maybe there are non-accurate applications where accuracy is unimportant or where the concept of accuracy just doesn’t apply.

But the distinctive potential of blockchain technology lies in accurate applications – applications where the blockchain contains data, not about some independently existing reality like births but about a reality the blockchain itself dictates, such as bitcoin holdings. In these cases, the blockchain (almost) can’t be wrong.

See also: Paul Brody - Public Blockchains Are Set to Reshape Global Commerce

For a long time, the only accurate applications were to cryptocurrency. But that is changing. The current non-fungible token (NFT) boom is a prime example. The fact that MetaKovan paid over $69 million for a NFT of Beeple’s digital artwork “Everydays” is not independent of the blockchain. Quite the contrary. His holding that NFT depends entirely on the blockchain saying he does. The blockchain dictates who holds which NFTs.

What other accurate applications are there? We might try to extend the example of NFTs to the ownership of intellectual property like copyrights and trademarks, or even real property like land and houses. Of course, a house exists independently of any blockchain, but not necessarily from the fact that you own the house.

A reworking of the foundations of property law seems unlikely anytime soon. But perhaps a more viable, accurate application is to voting. A jurisdiction (say, West Virginia) could recognize a blockchain, as opposed to a paper ballot, as the final word on what votes were cast. It could be guaranteed to give an accurate vote count.

The price of accuracy is unreality. The Bitcoin blockchain is accurate only because what it “records” isn’t real: Bitcoin holdings are shadows cast by the blockchain itself. Enterprise blockchains reach beyond these shadows into the real world, but in doing so they abandon the surety that is the technology’s distinctive promise.