From the bullet points we made on the whiteboard in the presentation room:

1. Improve MEP usage. An MEP for MEPs?
2. Development/Release cycles
3. Online docs
4. Tidy up axes (currently it violates the Single Responsibility Principle)
5. Refactor GUI backends, remove pyplot from the GUI backends
6. Default styling (a la ggplot, etc), does changing the default style "break" backwards compatibility
7. Release more often?
8. Be more memory efficient (I'm looking at you, imshow). Downsample images for interactive exploration?
9. Store metadata in output image format for reproducibility purposes?
10. Store extra data in containers so other libs (speak to Andrew from plotly) can scrape matplotlib figure objects and tweak them
11. Interactive figure editor, have the GUI record state after tweaks and save a python script than can regenerate that figure post-tweakage
12. ax.plot interface sucks, add a ax.line_plot with a more sane call signature?
13. json backend? How does this play with display logic?