First of all, AI is still a very zelous intern. Codes great, but if not kept in the line will go all over the place and create something that even AI struggles with. So I learned how to make what looks like good code. I also learned how to make bad and very bad code. Most of the learning are coming from me building a [FinOps Toolkit](https://github.com/fcontrepois/finops-toolkit) following the principles of Context is everything. The toolkit is here to show what I feel should be FinOps tools development. When context is everything you want to understand the contexts (with an s): - There is the context of builing a strong foundation and approach - There is the context of the developer that wants to extend the toolkit more aligned with their context; usually answering a specific question - There is the context of users of the toolkit, the people using shell and pipe to connect all the tools together to get something uniquly theirs - There is the context of AI, and how to have AI to be of help for Four contexts that we need to keep in mind. When coding with AI, I now start by explaining what I am trying to achieve