Saw a random tweet thread that claimed that the "51" in "51% attack" was misinformation, and that in reality only 33.34% of hash power would be needed for a bad actor to attack the network.
I was skeptical, but one of the posters linked to a talk by Dr. Leeman Baird where, as far as I can interpret, he explains that if you put a firewall around 1/3rd of the hash power and refuse to allow it to talk to the other 2/3rds, then you only need slightly more than half of this remaining 66% (33.34%) to maliciously attack the network. (The relevant part of the video begins at 1:01:08 or so).
I sort of get it as a concept, but as one of the students asks, it seems prohibitively difficult for a bad actor to cordon off 1/3rd of the world's hash power with a firewall. As they ask, is building a firewall around 1/3rd of hash power significantly easier than obtaining the remaining 18% of the hashpower they would've needed (18+33=51%) to conduct a standard 51% attack? This seems like a great question, but Dr. Baird casually responds that China already has a firewall around all of their internet, so doing so is pretty simple since it already exists.
Wouldn't the hash power just move? He explains that recognizing that it has been cordoned off would be difficult and thus makes this solution infeasible.
Is this legit? Would love for someone to give insight.