Consider the humble spreadsheet. :) My first thought was "the y-axis sure is nice for showing dependency order, so you need to compensate for losing that". In limited experience with IIRC Jupyter, I remember having to think a lot about cell dependencies and cache invalidation, which would have been prohibitive if the cells had been laid out in every direction. If all that stuff is automatic, then you can get away with making data flow less brutally visible.