The standard thread implementation on linux is pthreads and processes can easily be created by issuing a fork function call. While I was studying processes recently and reading the man pages for clone I found two important pieces of information:

• threads are commonly implemented using clone.
• the fork function calls the clone syscall.

The conclusion from these is that on linux threads == processes. So threads are in reality processes with a shared address space and processes are really processes with COW¹ pages. You can find confirmation in many answers on stackoverflow such as this one.

To me this is amazing, and kudos to the kernel devs for deciding that the two can be implemented in the same way and for optimizing this one path to where we are now.