A phrase often used to characterize the operation of modern A/D and D/A converters. In the old days of digital audio, a multi-bit word was derived from each sample taken. Later, as technology progressed, designers began to employ what’s known as Delta Sigma Modulation to create a stream of pulses at a very fast rate. Each pulse provides only one bit of data (a zero or a one), but at a high enough speed, these pulses accurately (enough) capture an audio waveform for the purposes of digital recording (there are other applications, too).
This idea is based in part on the fact that quantization distortion/noise is at its worst at the Nyquist frequency. At each octave below that, the noise is reduced by 3dB. So, if your sampling rate is high enough, then you can actually use only one bit to represent the audio with enough dynamic range to satisfy the human auditory system. This alone would require a very high sampling rate. So, in practice, there are other technologies employed, such as noise shaping, to fully accomplish the goal.
In layman’s terms, you can think of it in the following way: Each pulse merely tells you whether the waveform is headed up or down. Digital ones indicate the voltage is increasing, and digital zeros mean it is decreasing. Alternating ones and zeros mean it is staying the same. Put enough of those together, and a picture of the audio waveform emerges.