I think the battle for free speech is going to seriously heat up. So far in the West they are making a few famous examples out of people, deporting them, or famously cancelling them somehow. This has a chilling effect on everybody else who becomes afraid to speak openly. But the technology to cancel certain ideas could easily become far greater with AI and automation. The CCP in China puts a lot of effort into controlling what people say, and they are pretty successful at it. And that is prior to AI and automation AFAIK.

So I am of the belief that we need to use this short time that we have to leapfrog them, to become seriously and deeply uncensorable, not just "censorship resistant". And to me that means things like (1) linking to previous events so you can at least detect if an event was censored, like scuttlebutt did, (2) avoiding placing trust in DNS and CAs, (3) avoiding any kind of centralization, even the "where should I look for your relay list" semi-centralization of NIP-65 (which I meant as a temporary stop-gap measure). And of course ensuring IP privacy techniques are usable.

I think nostr would stand up against a lame censorship attack (taking out a single relay, then another, then another), but not against a sophisticated one.