In this talk, we describe how web2py together with crowdsourced grading enable teaching web applications at scale. Web2py is unique in allowing the easy packaging and sharing of applications. A student can write a web app, package it into a single file, then share it; other students can with the click of a mouse unpack and load the application, and use it. This allows teaching web development in a peer-driven fashion: students submit solutions to homework assignments, then review and grade each other's work, in the process learning from studying the solutions that other students submitted for the same problem. We have successfully used this approach at UC Santa Cruz for three years now, in the process building a tool, CrowdGrader, for the peer review and grading of the applications. The talk will describe the peer-driven approach and the tool, and present results on tool accuracy and student satisfaction.