Economic security is neither a meme, nor the cost that it takes to produce a safety fault of the finality gadget, in a model that considers more attacks than simply buying up enough stake to force the safety fault. Economic security is the ability to say "If a safety fault happens, at least X dollars will be lost", for some X that depends on the slashing rule. It doesn't say who loses X, or how much it *really* costs an attacker to lead the network to this outcome. Blacklisting a validator is a form of slashing, where the cost incurred is a cost of capital. It may be perfectly sufficient to consider the network secure. In Ethereum Proof-of-Stake, the choice was made to set the smallest possible X at the level of 1/3 of the total active stake. This does not mean that the true principals of this 1/3 of stake are malicious, and it could very well be that the agents they delegated this stake to are the malicious parties. Delegations are inevitable in Proof-of-Stake, so the question is how do we ensure that the principal-agent problem at play is healthy enough to consider the network secure? By raising the penalty a principal may incur by delegating to a possibly malicious agent (as Ethereum does), one should expect that, all else equal, the principal will be more careful picking their agent. This also means agents must level up their defences against being trivially compromised. This is the spherical cow world version. In practice, I think it is legitimate to ask whether such a large X is the optimal choice. - First, this raises the bar for agents, meaning that we should want more sophisticated agents and thus the barrier to entry to become an effective operator (credible enough to their delegator) is higher. Being less punitive with faults means more permissiveness for agents. On the other hand, this forces more innovation in making delegates credible, such as bonds, DVs, dual governance and other mechanisms. - Second, the credibility of the threat is in question. I believe we should go through with slashing X (i.e., at least a third of the stake) if Ethereum ever gets to this point. But if it does because of a consensus bug due to a faulty majority client, this is sure to cause a lot of discussion and pain. In this case, the large X actually imposes a lot of discipline on the system as a whole, including maintaining pressure for client diversity to be effective. Generally, we've been considering directions where the possibly very heavy consequences of slashing are contained to specific services, instead of to the validator role in its entirety. This is the gist of Rainbow staking The post mentions that given the (small, but nonzero) risk of a principal losing all or a significant part of their assets, we expect to see professionalisation and chains of intermediation for the services that are backed by full slashing commitments. In this light, it is probably a good idea to limit their possible fallout to as small a share of the network as possible, e.g., by targeting smaller quantities of ETH locked up in these services. This has of course other effects, such as compressing margins for less capital-efficient operators, such as home operators or solo stakers. The issuance curve discussions relate to these issues. It may be fine to do so if we can keep the participation of home operators effectively on other services that are not slashable, such as censorship-resistance. The problem is whether these services can or should be incentivised, and how. Different networks will make different decisions, so I appreciate Celestia proposing this route. Their light node sampling is very similar to a non-incentivised light FOCIL mechanism, albeit for different purposes, but as a service light enough that it can silently happen at the back while providing resilience for the network as a whole. If this resilience is strong enough, it is an argument for relying less on broad participation at the validating layer, and being more opinionated there. Excited to see this play out, if the proposal goes through!