Analogy at the core of cognition - This might be one of the most influential things I have read recently. Once I have read it I can see it everywhere. Let me try to describe it succinctly.
The argument is that experienced people in any given field don't have a higher brain capacity or larger number of neurons because of which they can come up with brilliant insights or clever ideas. The reasoning ability of a chess novice and chess grandmaster isn't that different as to be granted the sole reason of grandmaster's abilities . What sets a grandmaster apart is that he reasons about larger "chunks" than a novice.
For me it clicked in a software project I'm working on. I don't reason or calculate about minute workings of a single function, although I do have to do that sometimes. I reason about entire files and sets of functions at a time.
I am building my own mini version of React. At first I was working hard on just attaching DOM nodes and displaying something on the webpage. Now I'm reasoning about entire sub-trees of components at a time. It isn't that the latter is more intensive task than former. I would argue it's easier. But it has the most impact. Because of that I'm now seeing the higher level details of React, and how it works. I now see why it has to re-render the whole sub-tree when state changes.
This chunking is exactly why we modularize code. Extract a function in a separate file, test it thoroughly and now you can assume it is a primitive of the environment you are working in. I wrote a function which compares two virtual DOM trees in React and labels each leaf as either new or update. Now having tested this function, I can treat this function as a primitive and think about entire trees at a point. While writing this function I was thinking about single leaves of a tree at a time.
Keep doing this, and eventually you'll start reasoning at much higher levels of abstractions.
Look where we have reached using "chunking". logic gates -> circuits -> ....many steps I don't know.... -> AGI.