Description
LLM coding assistants do best when they have a summary of a codebase they can load to get context without expending as many tokens. I'm not totally sure what should be in such a file and if there needs to be different files for different LLM coding assistants. It's likely going to be an iterative process that should change over time as the assistants change.
Source code
Here is a sample file to get people thinking about the topic.
https://gist.github.com/schwehr/73b37abd8a38de7a4a41c2880a62c945
See also
I am inspired by https://github.com/Waveform-Analytics/PyAnimats where @michellejw shows working on a specification before making new code. See for example movement-models.md.
Description
LLM coding assistants do best when they have a summary of a codebase they can load to get context without expending as many tokens. I'm not totally sure what should be in such a file and if there needs to be different files for different LLM coding assistants. It's likely going to be an iterative process that should change over time as the assistants change.
Source code
Here is a sample file to get people thinking about the topic.
https://gist.github.com/schwehr/73b37abd8a38de7a4a41c2880a62c945
See also
I am inspired by https://github.com/Waveform-Analytics/PyAnimats where @michellejw shows working on a specification before making new code. See for example movement-models.md.