My middle ground for this is walking. It gives me enough input to suppress my habit of entertaining myself using a smartphone, and it also makes staring at a screen quite impractical. But it gives me enough room to just process my thoughts.

Good choice. As Nietzsche said, never trust a thought you had while sitting down.

There is a commonly-used way to avoid having an open-ended think while walking, too: podcasts.

And music, too. It’s underrated what a cultural cognitive shift into multitasking and experiencing the world through an artificial auditory filter that the Walkman, and then the iPod, brought to society even before smartphones.

Good for podcasts though :)

Speech recognition developer here - I hear many people complaining about the accuracy of speech recognition hindering multiple use-cases. I work for a lesser known speech recognition company Speechmatics [1] and accuracy is our top priority. We do have a real-time speech recognition API available, adding it as an alternative speech recognition provider for some of the tools mentioned here might improve the experience for end-users. I've got in touch with our team to see if there is any way we can get involved.

[1] https://www.speechmatics.com/