I had this idea while driving to work...

Imagine if the dsp side of pd was multidemensional. If every tilde~ object's first two arguments described a matrix of objects, [osc~ x y] would be (x*y) sin oscillators, each with a unique ID (0-0, 0-1, 0-2, 1-0, 1-1, etc...). The entire matrix could be addressed with a control matrix, or a message/list of (x*y) components. Each matrixed object could be individually called or controlled by a specific message depending on it's location in the grid. Each matrix~ objects would have to have some sort of multiplexed or mixed i/o, as well as individual ins and outs.

A central object would be a matrix mixer, both 1d(x=1) and 2d. I feel that this kind of dataflow would help to easily bridge the gap between generative audio and video, as well as standard(16*n) sequencing; both of which work with matrices.

A quick hack could be accomplished with Pd's shaky-at-best dynamic object allocation, but the real answer is something I am incapable of creating at this time. Maybe the answer lies in another language?

would be mega cool - I imediately thought of GridFlow (but it doesn't support DSP yet...)

using GridFlow-alike syntax, a reverb with 8 delay lines could be something like this:

delaytimes
|
[#~delread (8) $0-delay]
|
[#~ +]\[#~inner]\input
|
[#~inner]\scattering
|\
| [#~inner]/output
|
[#~ *]\feedback
|
[#~delwrite~ (8) $0-delay (8#1000)]

a lot easier than implementing it in vanilla Pd (which I did, for hardcoded input/output/scattering matrices...)