Paul's talk, as you would expect, was about Lisp, but it was very cunning, craftful, entertaining and thought provoking. He talked about programming language evolution, and how some languages (like COBOL and Java) are evolutionary dead ends -- neanderthal languages that don't have any intellectual dividends to speak of. Paul used an interesting theme for his talk: what will a good programming language look like in 100 years. Specifically, he is not so much interested in programming in that language in particular, but choosing programming languages today and tomorrow that are on the evolutionary path to the language we will be using 100 years from now.
He asserted that there is an "evolutionary trunk" that new and interesting programming languages extend. Avoiding the dead-end languages necessarily means staying close to the trunk. From here, he also slowly described the types of computers we can expect if Moore's law continues unimpeded for a century, where memory is effectively boundless, and computers are 78 quintillion times faster than they are today. In some situations, efficiency won't matter too much, and in other situations, efficiency will be at least as important as it is today. At this point, a good chunk of the talk was a connect-the-dots path to why something Lisplike will continue to be relevant, and why the 100 year language would be recognizable as a Lisp.
All in all an interesting talk. Other things of note: the DHCP server(s) appear to be overloaded with lease requests; I'm just now getting a connection for the first time today. It's a nice spring day in DC, but the air conditioning isn't kicking in enough in the breakout rooms; with ~250 python hackers in 3 crowded rooms, that entire wing of the 3rd floor is uncomfortably dry and hot, with little airflow. The grand ballroom is quite pleasant at the moment, with doors open to the balcony, but is generally unused except for keynotes, lunch and "open space".
All in all, this is shaping up to be a good YAPC, even if it's a Python event, and the conference planning and organization is rough around the edges.