You could make a first evaluation on the client so that the user gets quick feedback whether his solution is right. To validate that the user didn't cheat, you can make a second evaluation (server sided). The user shouldn't be forced to wait for the server sided evaluation. Instead he should be able to solve the next task immediately.

But nevertheless: Great idea with a lot of potential! Maybe you can give more detailed compile errors, if the user wrote incorrect code.