Brian,
I'd be happy to share any of the things I've been working on over the last few years, at the next Python gathering.
Most of my work has to do with text interpretation & object systems.
Phil Jones pointed out to me similarity between my work and "Phenotropic Programming," as a vision.
* ASCII tkinter
I made a system for creating tkinter GUIs from 2-D drawings in ASCII.
In principle, the same rectangle naming and recognition code can be put to other uses.
* Lions Sectioned Format (LSF)
A text file format for sectioned text.
Useful for anything that requires both human editable and computer editable text.
Much more human-friendly than XML.
NOT like YAML, in that the primary content is text, rather than nested hierarchies of data.
So for example: Journal entries, chapters in books, logs.
The focus is much more on sequence and text than hierarchy and bits.
Hence, "sectioned."
* Textual Drawn Databases
Draw a database in text, and then have it interpreted into named tuples and other such structures.
* nested text interpretation
You have a bunch of textual data of different formats, embedded inside each other.
Protocols for creating interpreters, mix & matching interpreters, keeping track of line numbers, indentation, change, etc.,.
If the user doesn't change anything, you don't want to mess up the human-chosen ordering scheme, white space, etc.,.
* cheap parsing
Parsing a file in multiple passes is computationally expensive, but often times, easier to think about, and orders of magnitude easier to think about.
Alternatively, parse "from the inside out" -- recognizing structures around a point, inferring larger structures. This way, you can parse data even when it's incorrect, filled with bugs.
Further, you don't need a "god's eye" view of the world, as is required in grammar construction, so you can build grammers part by part.
* Object2 systems & connectors
I've done a lot of work on high-powered object systems; Base object systems that can connect, default, serialize, alert, ...
* Object3 system: "A Membrane with Parts"
This idea, specifically, was inspired by ViewPoints Research, Ted Kaehler's short easy to read paper, "A Membrane with Parts: A new object model." The motto is, "Being a partner with other objects is more important than owning other objects."
* liberal argument interpretation
Demonstrations of functions that have very high capacities to interpret arguments liberally.
Being able to pass just about anything to anything, and having working stuff happen as a result.
* megalibraries
Modules with, say, 1,000 functions.
How do you find them? Tagging your code.
Python + Javascript for a Python webserver that will execute anything given to it, and return either JavaScript commands or HTML or JSON to be returned to the website.
Efforts to create a 2-D cross between EMACS & InkScape.
The word "visual" is overly overloaded, so I use the word "schematic" instead.
The idea is that iconography and shape need to be added to our editors, rather than to our languages.
If we add line and icon to our editors, then the languages that can make use of them will follow.
(Rather than the other way around.)
Based in the recognition that languages succeed in text because we have a solid concept of text editors.
I have implemented only the basic beginnings in Python/tkinter, and may move to JS/canvas.
I'm particularly interested in finding collaborators.
I am always filled with ideas, and explode outwards.
But when I have collaborators, things can take more solid form and become useful to others.
Ideas?