The readme was written by claude. And also most of the code, I usually write very descriptive markdowns about what I want especially in term of architecture, concepts and how it should be tested. Then I ask AI to created another markdown to plan the development of a specific feature I start iterate with it, function by function. And check every step that are done. I this way the LLM have the context of project between in session.

I double check everything, read the tests and the code. It sometime fails at solving problems or doesn’t follow exaclty the instructions so in this cases I do it myself. But most of the time it is a 10x accelerator.

The biggest downside of this approach is that it doesn’t help me to learn as well as I would about the implementation and understanding of the matematical concepts around finance. But I expect to dig deeper as soon as I will be able to use Quant in “production”.