Honestly, "DeFi" fucking blew it and isn't decentralized. We need to reset to first principles and use formal methods to build contracts from spec, prove the lemmas of the intermediate states, prove the invariants, and then build the code from those specs. Then test the resulting bytecode against the specs, then unit test, fuzz, and invariant/stateful fuzz everything.
Also, agentic security auditing is the new proof-of-work. Your protocol should have auditable compute from the latest frontier models red teaming the entire thing. If you can’t outspend your potential attackers, you lose.
Finally, no humans in the final deployed products, just protocols and the promise of lego building blocks. No fucking multisigs, no governance, no trust based human failure modes. Just permissionless, immutable, decentralized code the way God intended.
first - Cowswap was one of the first to actually serve its frontend via cowswap.eth
However - it got so little traction that it was deprioritized.
But more importantly - in the current state I would also not recommend it: The list of actors that can get compromised gets longer - not shorter:
1) whoever is running the gateway (https://t.co/8YCvizMirY https://t.co/2teCNWXfxn)
2) the DNS provider of that gateway
3) the RPC node the gateway is using to resolve cowswap.eth
4) potentially an IPFS gateway as well
So unless the user actually has a proper Ethereum light-client und IPFS node and verifies everything locally I think using cowswap.eth is not a security improvement.
I have to say, I need to say that with the decentralized thoughts blog (https://t.co/okwrsNB5Od), @kartik1507, @ittaia and others have done incredible job building one central resource that is extremely helpful for parsing through the dense web of distributed systems.
To the extent that is humanely possible, I admire a lot in their effort to cross link old relevant blogposts in the newer blogposts, linking to the twitter discussion for each blogpost, and updates to old blogposts as new findings come up. Speaks a lot about their effort, integrity and thoughtfulness.
I have found it extremely useful and I would highly recommend others to also use this resource as an anchor for navigating the world of distributed systems.
@bajpaiharsh244@uttam_singhk Exactly, with a validator set of size 20 with advanced hardware and concentrated geographical positioning, TPS shouldn’t be the right metric to judge an L1.