Chapel is an emerging parallel language being developed by Cray Inc. with the goal of making large-scale parallel programming more productive. While Chapel was designed with petascale-era systems in mind, we also believe that it is better-positioned to target the increasingly heterogeneous architectures that are being considered for exascale computing. In this talk, I will provide an introduction to Chapel's central features, including abstractions for locality, task parallelism, and data parallelism. I will then describe current work that strives to make Chapel more amenable to targeting next-generation node architectures including hierarchical multicore processors, tiled processor architectures, and accelerators such as GPUs.