A Model of Discovery: Context is King.
We don't know where new ideas come from.
In hindsight new ideas often seem obvious; the imaginative leaps that lead to the new ideas are often not that large if you are thinking about problems in the right way. The key insight – one that has been forced on me from working with AI agents – is that it is context that really matters; intelligence is now a commodity.
This model could explain why great discoveries are often made nearly simultaneously by different people; the "time is right" for an idea when the culture has loaded the correct context. (What is culture other than shared context?)
Context as the limiting factor of discovery could also explain the mechanism of serendipity: browsing physical book shelves or having interesting wide-ranging conversations are productive precisely because of the non-trivial collisions of context they create. This would also explain why our most productive thinkers are often polymaths.
This model also suggests a way to automate discovery with AI - search for non-trivial connections or common structures among rich, but randomly loaded contexts drawn from different fields/subjects/domains.