VPLs, understood as "nodes & edges", always (is that true?) provide some sort of infinite canvas to lay functions on, I'm curious if this combination is necessary - can the code still be text (or some other structural approach), while I can still "keep some functions here on the right of that other thing" - basically thinking spatially about my code
notable work in this direction is Dark (also featuring the "embodied data" idea):
another idea would be to allow working with "vertical slices": CSV data, transformation code and data visualization output visible at once, instead of dealing with parsing code, then somewhere else with visualization code, etc.
Combination of the background grid, and a quick way to build DOM elements, allows for adding pieces of documentation, notes, and inspiration images inline on the board, creating a canvas for programming: