I had an interesting experience this afternoon and opportunity to compare Google to Grok. I asked both how welding shade correlated to O.D. (optical density). Both started out saying the higher the OD the higher the shade number (duh), but nothing quantitative. Grok did give some quantitative examples, and of course, qualified its assessment with an accurate comment about wavelength.
I then asked Grok what OD corresponded to shades 11 & 12. It answered 21.x and 23.3 for shades 11 & 12, respectively. I responded that those results seemed to be an order of magnitude off. What surprised me is Grok responded with the correct calculations, apologized, and pointed out its own error. I am beginning to like (maybe not trust) Grok a lot more.