We use audio in a web browser all the time, listening to songs on YouTube or reacting to a sound in a flash game. But upcoming (and to some extent already available) techniques promise more: Interactive, dynamic audio with signal processing and even access to the sample level. Especially the last feature allows for more than traditional audio programming environments, such as Max/MSP or SuperCollider, can offer: Basically you could implement any audio application (or at least any DSP algorithm). Or maybe not?