Blockchain after the hype: what actually shipped

Our first production ledger system shipped in 2016. Token platforms and a self-custody wallet followed, and in 2018 APAC CIO Outlook listed us among the region's top ten blockchain solution providers. Looking back, about half of what the industry built in that period had a defensible reason to exist. Some of ours was in the other half, and it is worth being specific about which.

What held up

Key custody, which nobody found exciting at the time. DigieCash is an Ethereum wallet where users hold their own private keys and transfer tokens peer to peer. Issuing the token was a week of careful work. Making self-custody survivable for a person who is not an engineer took far longer: device loss, backup, recovery, an address book so nobody pastes hexadecimal strings by hand. That problem has not gone away, and it is what institutional digital asset custody is solving today with more lawyers attached.

The second thing that held up is narrower than the marketing of the period suggested. An append-only record with a published finality rule earns its cost when several organisations write to it and none of them is an operator the others will accept. That situation is rarer than 2017 believed. It is entirely real when it occurs.

Third, and least expected: the discipline. Code that cannot be patched after deployment forces a review standard most enterprise teams never reach, because there is no Friday hotfix available. Auditing the CRBT contract for Cruisebit was slower and more adversarial than the review we then ran on ordinary server code. That habit transferred to work that had nothing to do with ledgers.

What was a solution looking for a problem

Wherever a trusted operator already existed, consensus was a cost with no buyer. A great deal of that era's enterprise blockchain work was a distributed database run by one organisation, which is a database with extra steps and a worse query planner. Signed, append-only tables with a decent audit trail deliver the same evidential properties for a fraction of the operating cost.

The phrase "removing the intermediaries" also did a lot of unexamined work. Cruisebit's platform settles travel payments through smart contracts and does compress a long clearing chain into something close to instant. That chain also carries dispute handling, chargebacks and somebody legally answerable when a supplier fails to deliver. Instant settlement leaves no window in which any of that happens. For counterparties who want the trade it is a reasonable one, and it has to be a decision made in the open rather than a property discovered afterwards.

And a number of projects in that period used the token primarily as a funding mechanism and secondarily as a product. That is a financing structure wearing an architecture's clothes, and it aged accordingly. We built some of them competently. Competence was not the issue.

The test we apply now

• More than one organisation writes to the record.
• There is no operator that all of them will accept.
• Someone outside has to verify the record without trusting any writer.
• The cost of running consensus is lower than the cost of the disagreement it prevents.

If all four hold, a ledger is the right answer and we will build one. If any fails, the honest recommendation is a database and a well-designed audit log, and we have talked clients out of blockchain programmes on that basis. It is a poor way to sell a large piece of work and a good way to still be working with somebody in five years.

Where the interesting work went

It went to institutions and got slower and more boring, which is usually the sign that something has become real. Asset tokenisation, central bank digital currency architecture and digital asset custody all force the questions the retail token era could avoid. A CBDC design has to answer what happens when a citizen loses their phone, and no acceptable answer begins with the words "seed phrase".

The other direction is post-quantum cryptography, where a good deal of our current work in this practice now sits. Every ledger, wallet and custody arrangement rests on signature schemes with a shelf life, and migrating off them is unglamorous, expensive and dated by other people's timetables. It is the least exciting form this subject has taken and probably the most consequential.

Being on a top ten list in 2018 was pleasant. Being sceptical in 2018 would have been more useful, and we managed both only some of the time.