Decentralization from first principles - polynya

The title is a bit of a misnomer, because decentralization is merely a means to an end. Here are the actual goals of public blockchains, as I see it:

Corruption-resistance: no double-spending or otherwise changing the results of transactions
Censorship-resistance: the network cannot censor anyone’s transactions
Permissionless access: anyone can run a node and get their transaction included and very other transactions

The solution to corruption-resistance are twofold:

Economic security: run mechanisms like proof-of-work or proof-of-stake where the attacker has to spend money. The more they have to spend, the more difficult an attack is. Of course, you don’t actually have to spend money, you could also rent, collude or bribe - however, the more distributed your validator set is, the harder it is to collude.
Decentralization: for times when the above mechanism fails, and a collusion gains 51% of the economic security, you need many participants on the network to run their own nodes and change the rules to thwart attackers. To achieve this, you keep the cost of running nodes low so there are enough independent and unsubsidized nodes. It’s also important to note that just increasing the number of nodes isn’t as important as their quality - how geographically distributed they are, how socially aligned & attentive their noderunners are etc. It’s all about quality over quantity. 1,000 truly independent nodes run by enthusiasts around the world is far more valuable than 10,000 nodes mostly subsidized by the chain’s foundation.
Permissionless access: kind of the same as above - keep your barrier-to-entry low.

Needless to say, these methods are extremely inefficient, often requiring thousands of nodes to verify everything, significantly slowing things down for everyone on the network. So, what are the solutions to improve things?

Gain high economic security: Conjecture - if the network has enough of an economic security, decentralization becomes less important. Let’s say if there’s $1 trillion in ETH staked, it’s just not feasible to attack Ethereum, given the high risk of social slashing. Of course, the challenge is for ETH to become a multi-trillion dollar asset first, and the only way to achieve these heights is by becoming a bonafide reserve asset for the internet - an entirely different can of worms.
Implement PBS and proofs: Make it so not everyone has to reprocess everything. Implement validity proofs, state proofs, data availability proofs etc. so the people running validators and full nodes only need to verify a small proof rather than the whole computation and data. Someone has to actually do the full computation and such of course, but make this an honest-minority assumption so they cannot censor or corrupt. In essence, you replace everything honest-majority with honest-minority + succinct proofs for honest-majority. I believe all public blockchain L1s will tend towards this.
Novel cryptography and mechanisms: Justin Drake has an interesting comment about one-shot signatures - this method may be decades away, but the tech is constantly evolving, and let’s hope we see more innovations. There can also be new mechanisms that make corruption-resistance or censorship-resistance harder, e.g. smarter slashing deterrents, enshrined PBS etc.
Reputation: Requiring people to contribute stake or power or whatever is extremely inefficient and also purely plutocratic. Worse still, a vast majority of participants are well-behaved and non-malicious, so going through all this trouble is very inefficient. Communities like Lido prove it’s possible to have much fewer trustworthy participants who can represent a larger group.

Unfortunately, this is a hard problem because the key USP of a network like Bitcoin or Ethereum is objectivity and neutrality, and bringing any kind of reputation system brings in subjectivity. So, the extremely inefficient and unfair plutocratic methods are the only way. It’s possible the above trade-off is acceptable for some chains - particularly Ethereum L2s who have a fallback objective layer, and perhaps even some app-specific or community-oriented L1s. It’s also possible we see a breakthrough innovation that enables mostly objective reputation systems. We shall see.

To summarize, then, decentralization is not the goal, it’s merely a means to an end. The goal should be to increase censorship and corruption resistance, yes, but not through greater decentralization, which adds inefficiencies, but using smarter cryptography and other mechanisms.