Painters are taught to begin in generalities. Block in the big shapes, establish the masses of light and shadow, find the largest value relationships before any detail earns its place. The wisdom is old and sound: details seduce, and a painter who renders a perfect eyelash before locating the head in space will make the work more difficult. The broad strokes are the structure; the details are the reward you grant yourself once the structure can hold them. Work light to specific, large to small, and the painting stays coherent at every stage. Squint, and the thing should still read. 

...the eye wants structure first because it sees the whole and needs the parts kept in their place

In my work, sound design tends to run the opposite direction, and the reason is rooted in how the two media meet the senses. A painting is apprehended all at once, spatially, the eye takes in the whole field and then roves. Sound is apprehended in time, serially, and the ear builds its sense of a world from particulars accreting moment to moment. A "broad stroke" in sound …  a wash of generic ambience, a bed of undifferentiated room tone … communicates almost nothing on its own. It is the audio equivalent of a gray underpainting that never gets developed. What persuades the ear that a place is real or worth listening to is the specific: the particular tick of one clock, the individual creak of one floorboard, the exact grain of one voice in one room. Detail is not the reward in sound; detail is the foundation. 

So, I often work from the inside out. I think of the opening of Contact … the long pull-back from Earth out through the solar system and into the galaxy, carried almost entirely by sound. That sequence isn't built from a wash of "space" or a generic bed of broadcast noise. It's built from particulars: specific songs, specific news bulletins, specific human voices, each one identifiable, each one anchored to a moment. The camera travels outward and the signals grow older, because the farther out you go the longer ago they left Earth, until the individuated voices thin and recede into the silence of deep space. The whole emotional arc … an entire receding history of human noise … is assembled from small, true, specific events. And the one genuinely broad stroke, the silence at the far end, lands only because of the dense particularity it emerged from. Lead with the silence and it's nothing. Earn it with the details and it's the loudest thing in the film. 

That's the principle in miniature: I find the specific events first, the ones that carry character and point of view, and let the broad strokes arrive as connective tissue, as needed, to bind the particulars into a continuous space. Get the one right detail and the imagination supplies the rest of the room. Lead with the wash and the room stays generic forever. 

Lots of sound designers/editors prefer to begin building a track with ambiences, but I think that even in the case of “ambiences,” it is usually specific sounds within the ambiences that give the ambiences their character. 

This isn't a contradiction of the painter's wisdom so much as its mirror. The eye wants structure first because it sees the whole and needs the parts kept in their place. The ear wants specificity first because it hears in sequence and builds the whole from parts it trusts. The danger, in both arts, is leading with the wrong thing for the medium … and in sound, leading broad is how you lose the point of view, because point of view lives in which particulars are heard and which are skipped. Knowing which way your medium runs is most of knowing how to begin.