I looked into his first two claims since they're somewhat in my area of concern. I didn't find anything about the peer-to-peer network thing. Without seeing more, it's could mean almost literally anything and I can't really comment on it.
I did find
this (
archive) regarding the first question. TL;DR: Intel produced a hardware RNG but there's almost no info about how it works. Bardfinn "called out" the Intel CEO by... posting a politely worded question about it in an AMA with the Intel CEO. Wow. Hardcore. And late to the party, considering people had already solved this problem when it first arose. He accuses Intel of trying to insert a backdoor, and frankly he's probably right. (Although they already have IME, so what-fucking-ever.)
So what's the answer? How do you trust an RNG without the specs? Any tech with two brain cells to knock together will come up with the same answer in about 3 seconds: You fucking don't. Bardfinn seems to admit this in his post, but doesn't like how people were able to come up with a solution without consulting his genius. Quoth: "Linus Torvald's response to this, while incredibly tone-deaf, is technically correct." No fucking shit Steve. Next you're gonna tell me that it's "technically correct" for wheels to be round. Idiot. (Rate me MATI for this part, but... Jesus. Some people's kids...)
I have no earthly idea what he's going on about regarding PKI. Trustworthiness and management of PKI is a live problem, but it has absolutely nothing to do with hardware RNG. Either I missed something major here, or he's literally retarded. It's just so completely out of left field compared to everything else.
The final point he makes is about how it's hard to be sure hardware is built to spec, and hasn't been tampered with in the supply chain by the time it gets to the user. His analysis of the problem is basically correct, but his proposed solution is pants on head retarded (a discrete crypto processor with a transparent casing so it can be visually analyzed, there are about a billion reasons this is infeasible). Much smarter people than him have tried to tackle this, but nobody's really figured this one out yet.
In his defense, he's not really WRONG about most of this. But he's being an assbarrel if he think he's got some stunning and brave original insight, rather than chronic and well known set of problems in infosec circles that trickled down to him from smarter people. The things he's right about are the things that are obvious and well known. The things he's wrong about is... everything else.
Steve isn't nearly the big brained badass he thinks he is. He's far from stupid, but he's not Albert Einstein and nobody is clapping. He's a midwit. He thinks his IQ is 150 when it's really 105.