I read about udp holepunching, and protocols for opening up ports in the NAT/firewall, (uPNP, PCP, etc.). Seems like using a good library would be the way to go here, and there are several to choose from. For bootstrapping, I want to take advantage of bittorrent trackers to prevent a single point of failure. There are about 250 of them listed here: https://www.desmos.com/calculator/lrttqfca8b which would do quite nicely. I'm pretty sure I can get a browser to pretend to be a bittorrent peer and get a valid response. Then I make a fake captain-dirgo torrent infohash, that actually points to the captain-dirgo network, and thats how nodes will find each other. The only problem is that the only thing a bittorrent tracker will store is an ip and port. So, I will need some people running daemons that can expose a TCP port on the public internet, which I think I can make easier with uPNP et al. I was concerned about my sampling idea for PoW because I hadn't heard it being done before. So I researched it further, and found this paper on what's called "Proof of Sequential Work". The thing that made me relieved was that they are actually doing the same technique that I was thinking of: https://eprint.iacr.org/2018/183.pdf You can see this under section 1.2 PoSW Definition under Challenges (this is just another name for what I was calling "sampling") The other issue I was worried about was what's called "grinding". This is where you take the random seed for a process (or some other variable) and attempt to manipulate it to your advantage. But I proved to myself that grinding this sort of thing would actually take more work than doing it honestly, even if the sample size is as low as 2.