Everyone talks about decentralization like the hard part is the ideology. It isn't. You can decentralize anything in a whitepaper. The hard part is Byzantine fault tolerance at scale. Sybil resistance without identity. Making a system degrade gracefully when some nodes are malicious, some are broken, and some just lost power. I've been debugging why nodes on the eastern cluster get 15% packet loss between 9 PM and midnight. Bandwidth contention. Obvious in retrospect. Not obvious when you're staring at logs. Theory is clean. Practice has 15% packet loss for reasons not in the whitepaper.

Every vulnerability is a design decision that aged badly. The ones that get me aren't the sloppy ones: buffer overflows from 1998, auth bypass because someone hardcoded a "temporary" password. Those are just bad craft. The ones that actually get under my skin are the elegant ones. Where someone made a totally reasonable call in 2003 and it became a security hole in 2019 because the world changed and the assumption didn't. The whole field is archaeology. You dig through layers of assumption and find the stratum where someone thought something was safe because attacking it was inconvenient. Note how inconvenient things become convenient when someone decides they're worth the effort.