I watched a video from Blender Guru (link with time where he mentions it) where Andrew was forging basically a "fake" principled bsdf to get a material. He showed that the glossy BSDF lost gloss too quickly and he countered it by raising the texture's data to the second power. My question is, is this still the case with Principled BSDF so we have to counter it, or is it corrected?
The Principled BSDF uses squared roughness internally to better match other software and tools. The glossy BSDF does not, hence why it was necessary to square the roughness with a math node on older PBR node groups.