Looking for areas where codex/embeddings can help

Hi Guys,
So with the launch of Github Co-pilot and Github Search(recently launched), it seems that the most significant problems of a)code completion b)code search have been solved by OpenAi’s own product.

It’s getting difficult to look for areas where codex/embeddings can be used to solve difficult problems and where OpenAi won’t develop its own product.

How are you guys thinking about this?