1) Paul is not ignorant of the existence of Smalltalk
2) Smalltalk is not a replacement for the fruits of modern language research. In fact, the academic work around Smalltalk is nothing more than unsupported opinion, and it's not even remotely useful as a source of research.
1) Posits new conjectures 2) Proves those conjectures 3) Those conjectures can be used to formulate new conjectures that conform to items 1-3?
I've read a great deal of smalltalk literature, and I can't recall any papers that would qualify. There's plenty of empirical exploration of ambiguous hypothesis, but nothing that actually provides anyone in the field anything on which they could actually build.
Back in the mid-90's, before Java was released to the world, I was using VisualWorks for an university project.
IBM was a big enterprise player with Visual Age for Smalltalk.
A small startup (Animorphic) was making Smalltalk faster (StrongTalk).
There were, of course, other companies searching their piece of the pie.
When Sun started pushing Java to the world, many in the Smalltalk world shifted direction.
Visual Age for Smalltalk architecture became the foundation of Eclipse.
Sun eventually bought Animorphic and Hotspot was born out of StrongTalk VM ashes.
Slowly, other Smalltalk players joined the party.
It's funny, because there's actually a language just like that: Smalltalk. The development environment is the same as the runtime environment, it has its own source control (Monticello), its own editor and a very regular language.