This talk will be a brief overview of three results on designing robust, large-scale networks.  First, we will describe a result that resolves a 23 year old open question: "Does there exist a polynomial time Monte Carlo algorithm for asynchronous Byzantine agreement?" Next, we will describe how an alert can spread more quickly than a worm over a large network. Finally, we will discuss how a network can self-heal even when an omniscient adversary repeatedly removes carefully chosen nodes.

The talk will cover results previously published in FOCS, SODA, PODC and OPODIS.  We will end with many open problems. .