GPTs should cache code fragments and re-use them to make them faster

It would be amazing if GPTs could realise when they are repeatedly solving the same problem with the code interpreter in threads (generating equivalent code)- generalise, and store the generated code to make future conversations much quicker.