I wonder what fonts were used. Some of Microsoft’s fonts (e.g. Cambria, Georgia, Sitka, Tahoma, Trebuchet and Verdana) were designed specifically for use on-screen and when printing at very small point sizes. Other fonts, like Arial, Palatino and Times New Roman, are designed for printing and can suffer when used on-screen, especially at small point sizes.
See also Font List Windows 10 - Typography | Microsoft Learn
Apple’s set of fonts similarly has some optimized for screen usage and some optimized for print usage. Any test that attempts to compare font-rendering systems (e.g. Apple vs. Microsoft) must use the same fonts (preferably the exact same TTF files) or it will be impossible to know if the differences are due to the OS or the font.
And interesting article, but I can’t help but notice that he isn’t presenting all of the options in his comparison.
He presents a 5K screen at its native resolution (which I assume to mean configured with HiDPI disabled, because otherwise every resolution is scaled from a larger internal image), 4K at its native resolution, and a 5K image scaled down to a 4K screen (by the GPU? by the display?).
But are those the only choices? I haven’t worked with large screens on macOS before, but on Windows, you can do more than just pick a resolution. You can also pick a scaling factor. If you have a 4K display (which the author thinks is rendering text too large), and choose a scaling factor less than 100%, it will reduce the size of everything, but it won’t do it by making a larger desktop that is scaled down to 4K. It will apply the scaling to the rendering pipeline, causing the fonts to render at smaller point sizes. This should be less blurry than downscaling a 5K image. Vector graphics should be similarly cleanly scaled.
Of course, there is no substitute for having more pixels, but there are different ways to scale images and some will definitely look better than others.
I’d love to see Mr. Wellborn’s actual system configuration (on the Mac and Windows), fonts and sizes chosen for the test. What we’ve got here isn’t enough for us to judge if he’s making a legitimate point or if there may be a solution to his problem that he hasn’t tried yet.