Not so. Even today, Mac and iOS apps run on the main thread and require extra effort to utilize multiple cores efficiently. Apple has made this easier over the years, but it still takes effort.

Apple assumes that when speed is important, developers will use code that has been carefully optimized for its platform; not merely balancing data transfer time against data processing time, but also utilizing the GPU and neural coprocessor. This is true for well-funded companies. It's also true for Python developers when they can leverage highly-optimized numerical libraries. It's not true for amateur code and problems that don't easily leverage optimized libraries.

For example, last week my son wanted to simplify a large 3D model he found on the internet. He found a tool on a web page for doing that. (It ran entirely in the local browser, so didn't require uploading.) The code was written in single-threaded C++ and compiled to WebAssembly. It was fast up to a certain point, but bogged down on his huge file.

If you limit yourself to highly optimized commercial software, single-threaded performance will rarely be an issue. But never is an overstatement.