The majority of my pieces are created using a process not very different from how I worked in the 70's. I usually begin with a concept for a modular synth patch. These are usually pretty basic notions of ways to connect (and sometimes cross-connect) modules in the signal and control paths, such as "use the interference pattern of two low-frequency oscillators to create a rhythm," "use pitch sweeps driven by an envelope generator on the inputs to a ring modulator to create dynamic timbre" and the like. I then simply play around with such patches -- changing the underlying wave forms, inverting the phase of some of the low-frequency oscillator signals, changing the base frequency relationships when using FM, and so on -- until I find something that appeals to my very idiosyncratic tastes sufficiently to be worth recording. I then repeat the same process to create multiple layers of sound for a given composition. Where I used a multitrack analog recorder back in the day, I now use Audacity to record each sonic layer. Finally, I align and trim the tracks, add fades and similar basic editing in the DAW to produce a final mix as a FLAC file. That is where the greater flexibility of a DAW shines, but I do miss some of the tricks you could play with tape. One of these days I'll have to find or write a time-reversal plugin for Audacity that would work like turning the cassette over between recording two tracks on my old PortaStudio.
Most of the time, when I "play" a piece during the recording process it is by adjusting potentiometers to alter the patch in real time. Often I don't do even that during the recording but, instead, allow the "logic" of a given patch to play itself out over time. When I use a keyboard at all when working this way, it is usually to provide trigger and gate voltages rather than to play "notes" in the traditional sense. One characteristic of modular analog synthesis is that small changes to a given patch can produce dramatic differences in the resulting sound. Multiple tracks on a given album may be the result of the process I just described, using relatively minor variations of a single patch.
An aspect of this approach is that there usually is neither a musical score in the conventional sense nor any way to recover the exact combinations of settings and initial phase relationships of the various modules contributing to a patch. Once a track is recorded, that recording is the only representation of that exact "composition" and there would be no way for me or anyone else to perform it again and get the same result. This is one of the features of "West Coast" analog electronic music that I find conceptually attractive. There is a complete break not only from traditional tunings, structures and forms but the totality of a musical composition created in this way is represented entirely by its acoustic signature and nothing else.