bufferless - Don't wait for an image, just stream pixels as fast as you find them

clockless - No world "ticks", the world is an append-only log of "percepts"¹ which can be projected onto any time.

stochastic - Don't wait for certainty about a pixel, just push out the most probable ones first

signed distance field - Afformentioned "percepts" don't have well defined boundaries like a polygon, instead "fields" centered on a point describe how light moves around them. Any two fields can be trivially summed, so you can ignore most of a scene when searching for a specific pixel near a small number of local fields.

Together they allow you to supply the eyes with nearly zero-latency data with arbitrarily low computing power.

¹ As an aside, there is evidence humans don't see a "now" tick either, we perceive "fields out of time" directly, log them, and interpolate their relationship to "now" thereafter, such that we feel that we are "seeing" something which our eyes have already stopped reporting about. Thus SDFs and clockless rendering are a natural fit and map well to human perception.