I like the concept of @jashkenas’s geographical coordinates inputs, https://observablehq.com/@jashkenas/inputs#worldMapCoordinatesDemo but it’s pretty hard to get a precise position into it, which makes it not so useful for many purposes.
Minor feedback: the labels on the map should say e.g. Longitude: 119°30′ W instead of Longitude: -119.50. Also, click and drag should ideally cause the red dot to follow the mouse and coordinates as written below to change, instead of only updating on mouseup. (Perhaps the input value – and therefore downstream cells – should change only at mouseup, or perhaps that should be configurable, I’m not sure.)
More substantial idea: I don’t have spare attention to tackle this right now, but here’s an idea if someone else wants to take a shot: what if the input showed ~4 different maps from left to right (or on small screens maybe in 2 rows) at different zoom levels, starting from the whole world but then getting scaled 1000 times closer by the last map. Then someone trying to pinpoint a specific place could do a reasonable job at it.
I’m not sure what the best method would be for getting appropriate outline data for closer zoom levels.