The unreasonable effectiveness of language models for source code

This blog post is deprecated in favour of an upcoming paper. Stay tuned!