You're not wrong, but largely as a result of dubious architectural decisions made in the name of backwards compatibility and minimal hardware requirements, Microsoft sold 40 million copies of Windows 95 in its first year, compared to 300,000 copies of Windows NT 3.1.
Consider:
Windows 95 ran the vast majority of MS-DOS and Windows 3.1 applications with minimal performance loss, supported MS-DOS and Windows 3.x drivers for hardware that lacked 32-bit driver support, and ran acceptably on a 386 with as little as 4 MB RAM.
The properly architected Windows NT 3.1, released two years before Windows 95, had limited MS-DOS and Windows 3.1 application support, required NT-specific drivers for all hardware, and required 12 MB RAM to boot, 16 MB to do anything useful, and you really wanted a 486 for decent performance.
Now try a 3rd comment that actually connects to the design deficiency described in the article instead of a generic grievance about rearchitectrue that included a gazillion of changes
Consider:
Windows 95 ran the vast majority of MS-DOS and Windows 3.1 applications with minimal performance loss, supported MS-DOS and Windows 3.x drivers for hardware that lacked 32-bit driver support, and ran acceptably on a 386 with as little as 4 MB RAM.
The properly architected Windows NT 3.1, released two years before Windows 95, had limited MS-DOS and Windows 3.1 application support, required NT-specific drivers for all hardware, and required 12 MB RAM to boot, 16 MB to do anything useful, and you really wanted a 486 for decent performance.