[roenxi]

It is interesting, given that the corpuses are available, that keyboard design isn't framed as some sort of linear or nonlinear optimising problem.

Choose a set of tasks -> plug into an optimiser -> get a keyboard mapping tuned to the task you are doing. There'd be people who find that useful. The maths isn't at all hard, the data is there.

EDIT Maybe it would be a fun project to take something like top 100 C libraries from github and build up an optimised keyboard layout for C coding. It is an interesting idea.

[toyg]

The problem is generalizing the corpus enough to make it viable for more tasks than "writing C libraries", so that it can achieve mainstream support. The parameters to consider are also contentious - are we optimising for travel, finger stretching, speed, etc ...? It's not just about data.