Is there a primary benchmark (joules, seconds) showing a reconfigurable analog computer beating a digital solver on differential equations, or only the asymptotic 'constant solution time' claim?
claim-reconfigurable-analog-computers-target-differential-equations-not-neural-networks rests on the arXiv paper Reconfigurable Analog Computers (2510.25942), which states that "the solution time on an analog computer is basically constant, regardless of the number of computing elements required," while digital solution time "typically grows with problem size, often much worse than linearly." This is a claim about the shape of the scaling curve, not a benchmarked energy or wall-clock figure. The note carries an [unverified-quant] flag because no joules or seconds appear.
Why it matters
The seed premise behind the whole capture — "niche 2020s problems still solved by genuine analog computers because digital is measurably worse" — needs a measured case to be more than an asymptotic argument. "Digital is measurably worse" implies a measurement. Without one, the note can only assert the scaling shape, not that analog wins in practice on a specific problem at a specific size.
What would answer it
- Read the full arXiv paper 2510.25942 (https://arxiv.org/html/2510.25942) for any benchmark table: a named ODE system solved on Anabrid LUCIDAC/REDAC hardware with reported time and/or energy against a named digital solver on named hardware.
- Anabrid / SpriND primary materials (LUCIDAC specs, application notes) reporting solve time or power for a specific dynamical system.
- Any peer-reviewed comparison of analog vs digital ODE integration with concrete figures and the problem size at which analog overtakes digital (the crossover point).
Resolve to a claim-note with concrete numbers if a benchmark exists; otherwise record explicitly that the advantage is asymptotic-only and note the absence.