Decentralization is  a spectrum. On on end, you have the likes of bank servers that are controlled by a single entity with read-write power. On the other end you have something like Bitcoin that is run and secured by a globally distributed set of nodes and miners.

The more centralized a service is, the more that single entity and its authorized personnel can impose their will on others. The more decentralized a service is, the more censorship resistant it is. But there are trade-offs across this entire spectrum — just as there is with, well, virtually everything.

Take the example of total self-custody. It’s a cool concept, because it massively enhances your ability to evade censorship and having someone else decide what you can do with your assets. The biggest benefit of self-custody is the user’s ability to significantly cut down counterparty risk. Everyone in finance is always trying to find ways to minimize counterparty risk.

But the downside is the ability for someone to irrevocably lock themselves out of their wallets, forever losing whatever assets were held on those blockchain addresses. That wouldn’t happen with Stripe or HSBC. Because of their centralized nature, they are able to re-authenticate access to their customers when they forget a password. Self-custodial apps are, by design, not as forgiving.

It’s clear that there’s trade-offs across the decentralization spectrum. So the trillion dollar question — that I believe ultimately unlocks mass adoption — is which layers need to be decentralized and which can be centralized?

Some degree of centralization is probably ideal on the customer-interfacing layer. Maximal decentralization is ideal on the infrastructure layers. Note the plural “layers”, because there are several of them.

At the bottom you have the actual network. Be that Ethereum (an L1) or Arbitrum (an L2), decentralization of this layer should be non-negotiable. On top of that, you have specific infrastructure. Uniswap, Aave, and Opensea are liquidity infrastructure. Things like Manifold (NFTs) and Gnosis Safe (wallet) are creation infrastructure (or something of that nature). You want this layer to also be maximally decentralized. But at minimum, it should be sufficiently decentralized so as to allow people to exit before a change is enacted.

As Ceteris points out in his fantastic report on wallets, the industry is making significant headway towards self-custodial experiences that minimize the negative trade-offs associated with self-custody. That said, the risk of getting locked out is never zero. And thus, I think there is a large subset that would not be comfortable with that trade-off and would prefer to bear counterparty risk of a legally incorporated entity.

In my mind there’s a big distinction between specific infrastructure and dApps. I would not call Uniswap a dApp. But I would call something like Instadapp a dapp (duh).

And so, in the grand scheme of things, the things I call dApps will likely compete with centralized apps. Something that looks, feels, and smells like Binance but uses decentralized liquidity infrastructure behind the scenes. They would have their own in-house DEX router and a “smart algo” to determine where the cheapest loan is (and the ability to do one click re-finances) among several other things.

Centralization on this layer is not bad. You are not forced to use this Binance-esque app that uses decentralized liquidity infra. If you, like me, enjoy pain then you can continue using hardware wallets with browser plugins to directly interact with front-ends. But having options with centralization on the customer-facing layer opens the door to many, many people who are otherwise not comfortable with self-custody. And it does so in a way that does not compromise the integrity of our industry and the core infra that powers it.