I went through a similar transition in hardware, from just using modulars as a ‘sound engine’ to doing almost everything with them, composition, effects routing and so on. I still use the DAW for basic recording and effects (because I don’t want right now to mess with tape and outboard) but this is only a minimal use and doesn’t get much in the way.
It was one of the best decisions I’ve made. I no longer have to think anything out beforehand, everything can be developed more or less spontaneously, not just the sound design. Things like feedback sequencing, multiple interacting sequencers, also sequenced gating/filtering before effects send, etc. have now become standard and it’s enabled me to push things much further.
I was curious and spent some time with VCV and many aspects of the experience are similar. I didn’t mind as much having to control everything with mouse and keyboard, things are a lot slower this way than the direct hands-on approach of hardware, but the way of working is more or less the same.
What keeps me from VCV are all the usual, built-in limitations on sound quality inherent to basically all DSP-based approaches since the 1990’s. In particular, aliasing artifacts and/or the usual muffled sound when you try to push oscillator sync+FM. Running at the highest available sampling rate (for me: 176 kHz) didn’t have much of an effect.
Unfortunately I don’t think there’s a solution here, their entire way of doing things needs to be rethought. So for me unfortunately, the entire format is DOA. It’s a shame because other than these artifacts the quality of the built-in stuff is quite good, reminiscent of the Roland 100m modules in many aspects.
The idea that complex computations must be run at the audio sampling rate rather than much simpler computations at much higher rates is the culprit. This may have been a necessary limitation in the 1980’s but shouldn’t be one with the processing power available today.
Much research has been poured into things like band-limited oscillator design and delay-free loop approximations, which are simply band-aids over problems that should not have been created in the first place. Band-limited oscillators do not stay that way when modulated and this aspect was very obvious to me with VCV. Sure, one could then just not apply various modulations, but being forced to work in a much more limited way is kind of antithetical to the modular idea in the first place.
In theory an individual module could still do computations internally at these much higher rates, but the format then would force implicit rate conversions at all inputs and outputs of each module. This adds latency which will mess up feedback patches, and is not a truly modular solution.
Of course there are then huge technical challenges in how to optimally run DSP at much higher rates (i.e. 10 MHz), in particular resonant filters become very difficult due to numerical quantization effects. Third-party developers would need to be given lots of “blocks” to get around these issues, and it’s not clear the appropriate research has even been done. Aside from the filtering problem, there’s the fact that higher rates would otherwise favor a return to integer or even low-bit computations, due to the ability to shape quantization noise. The point is that raw DSP would become very “weird”, and probably unfriendly to developers who have not grown up in this alternative path – so that another API layer would almost have to be developed and standardized to attract developers at all. A great candidate for the API would be simply a collection of analog circuit elements plus a few other ‘ideal’ modules. The point is while this totally different approach to DSP will yield “true modularity” on the part of users, it may yield just the opposite for developers, favoring only ‘closed’ approaches.
Anyway, these considerations are probably irrelevant to most. VCV seems to keep growing in popularity, and its accessibility ($0 vs. > $10K in most equivalent hardware setups, considering the usual levels of patch complexity) is obviously unmatched. But I see VCV as continuous with all-hardware approaches (i.e. doing everything with modular rather than using keyboards+MIDI) rather than as anything related specifically to DAWs.