As for the future of physics engines, Gustafsson described networked multiplayer physics as the final frontier.
"Networked physics can be done in two ways — either you send the state of all objects to all clients all the time, or you just send the inputs and make sure all clients behave exactly the same. The first method takes a lot of bandwidth and will cause glitches when two players interact with the same object. The second one introduces an unacceptable lag, which can be hidden by predicting the world state and [making] corrections later on, but that is too slow with lots of physics, so none of them really work on a large scale."
There is light on the horizon though, he said. "When or if we move over to cloud streaming that will partially solve these issues because all clients run on some central server farm with really fast internal connections, but cloud streaming might introduce other issues. I guess we'll find out soon."
Multiplayer physics may be the holy grail of game physics engines, but Gustafsson himself is in no hurry to tackle it. His own project will be a single-player affair, for starters. But more than that, Gustafsson pointed to the soaring popularity of online multiplayer as one of the reasons mainline studios haven't made any Half-Life 2-tier physics advances recently.
"A lot of people, myself included, enjoy playing with physics, but in some ways physics in games even had to take a step back over the last decade in favor of online multiplayer. Nowadays people expect to play with their friends online and I think that makes a lot of developers steer away from physics-driven gameplay. Half-Life 2 is still one of the best implementations of physics in games and that released fifteen years ago, so I don't think it's only about technology."