Everybody has that one book that changed a topic for them. For me it was the Dragon Book (Compilers: Principles, Techniques and Tools). I was able to use that book as a jumping of point to tons of other papers etc in a kind of iterative fashion - more importantly to build to scratch personal itches.

I am trying to do the same with ML/AI. I learn best by building (specifically building tooling) and I am struggling to find a path like this in ML/AI. I know the area is still relatively young so a book is not easy to hope for. Even a curated list of gradual papers would be awesome. Ive tried Karpathy's NN Zero to Hero, Andrew Young's bunch of ML courses and others. Not to mention tons of articles on tool X being amazing and another tons of articles on the same tool X being pointless (eg X = langchain) isnt helping establish a starting point either.

I also understand I am like the millionth person probably asking for this and yet (may be) like the millionth person I cant help feeling lost and not knowing how to start building fearlessly.