I'm outside my edit window but I'm surprised at the negative reaction to this.

This experimental process on small NNs us exactly how science is done. You develop and test theories of what will work using small networks because you can do them quickly.

Then you test on large, production size networks based on the lessons from the small ones.

Probably because one word replies come across as low effort. IMO your second reply is great.

It had the right amount of effort that the parent question deserved.