He says if you know what to measure ( a more comprehensive set than usually done) you can judge an amp only by measurements - if it is extremely good in all aspects it will sound good. But he says most measurements sets of amps don't measure enough aspects of the performance to find the weaknesses of the amp, and therefore aren't reliably predictive of it's sound.
That is a good summary.
This interview was good especially for this site because it is with someone who creates designs that measure well (at least under the usual metrics) and believes in measurements but is realistic about the limitations of measurements. Typically, the latter happens from those who are only looking to discount measurements and so not taken seriously (at least by this group).
Examples include doing IMD measurements with a low frequency along with a high frequency rather than two bunched frequencies as the former might expose some weakness in that particular design of the amp. In Class D, since the design tries to keep the audible range do well and "bugger the rest by design" (his words), it might indicate a high level of distortion at 20khz signal but may not be as important since the higher harmonics are out of audible bandwidth. So, it would be better to test at increasingly lower frequencies to see if any of the harmonics manifest themselves in the audible range.
Those are good points and food for thought as to whether a standard 1khz test for distortion numbers (along with multi-tone) is enough for all amp designs or even sufficient to expose potential issues within a specific amp design.
The main takeaway from that interview is that amps that measure "clean" in the limited set of measurements as done here aren't guaranteed to "sound neutral" unless they include measurements that test for the potential weaknesses in that design. Only when they measure well in all of those ways (and he does realize the practical limitations of doing so exhaustively) can one be sure about the amp being "neutral" (or transparent as used synonymously). Another way to look at it is that you can obtain confidence about "transparency" only asymptotically in number of measurements and have to make a cut-off at some point as enough.
The flip side of that thesis is that if you haven't done that exhaustive testing OR an amp doesn't fully ace in the measurements you have done then you have no way to predict how the amp will actually sound.
There are some strong views amongst people I have read here that the measures done here are all one needs to judge an amplifier by. Bruno's views are counter to that.
There is a preference for "standardized tests" here over idiosyncratic ones. For example, linearity for amps is not normally done because there is no standardization of what levels to measure at but that also means it may not expose a potential problem with an amp where that measurement has not been done. Then there is the practical aspect of it. There are only so many tests that can be done within anyone's time and resources. But that does not prove
completeness or
sufficiency of measurements.
All are good reasons for the limited set of tests used here but I think it is also OK to believe that these tests are not necessarily
complete for a given amp design and the final word. Bruno has a persuasive argument for this.