Automatically decompose large Git commits

Is there a way that the AI system gets two successive Git commits (or initial commit which is large = lots of new code was added), so it can take a diff between them and will create automatically a new "imaginary branch" with lots of small commits with proper commit messages, so we can go from first small commit to last small commit and visualize logically how something large was built by smaller chunks of code. Could you refer me, where to find more information about such logical decompositions, so I start like minimalist with something small and workable and to extend it to the final result, not with one big step, but rather with lots of smaller good documented steps. Such system should have good reasoning abilities and be interconnected with Git. I like reading code in Github from commit to commit, but this decomposition approach could be more enlightening and sensible. The motivation was to understand Django source code on Github where first commit is really not minimal. The system splitting it into smaller commits should recognize the core functionality that is still working and make a first commit from it and then add successively other not so important functionalities.

May 23, 2025 - 05:50

Is there a way that the AI system gets two successive Git commits (or initial commit which is large = lots of new code was added), so it can take a diff between them and will create automatically a new "imaginary branch" with lots of small commits with proper commit messages, so we can go from first small commit to last small commit and visualize logically how something large was built by smaller chunks of code.

Could you refer me, where to find more information about such logical decompositions, so I start like minimalist with something small and workable and to extend it to the final result, not with one big step, but rather with lots of smaller good documented steps. Such system should have good reasoning abilities and be interconnected with Git.

I like reading code in Github from commit to commit, but this decomposition approach could be more enlightening and sensible.

The motivation was to understand Django source code on Github where first commit is really not minimal. The system splitting it into smaller commits should recognize the core functionality that is still working and make a first commit from it and then add successively other not so important functionalities.