I think that's oversimplified. TOS was single tasking but two things a) unlike the Amiga a proper 68000 syscall TRAP mechanism was used meaning the OS etc ran using proper supervisor / user separation and b) the application toolkit (AES) built overtop supported message passing and multiple application semantics and c) the graphics subsystem (VDI) was also in theory abstracted away from the physical. In fact the OS components themselves had originally been developed (by Digital Research) for both x86 and 68000 and the original work was done on the Lisa before it ever got put on a Atari hardware. And DR's GEM had a life of its own on PC hardware (see Ventura Publisher, etc.), though handicapped by the Apple lawsuit and competition from Microsoft.
These things meant that later the Atari community and Atari themselves were able to extend the OS in a proper multitasking almost Unix-like direction (MiNT and MultiGEM) and bring it to new hardware, and new display formats and architectures, etc. Provided the applications being run were cleanly written (well, that's a big caveat...). For example -- we can now run TOS/GEM on an Amiga, that's pretty neat (though totally pointless).
In true Tramiel fashion they shipped the cheapest simplest thing they could. But it was something they were able to iterate on -- unfortunately they just did this too slowly. As others have pointed out, the Amiga had amazing hardware from the go, but its architecture became somewhat tied to that original hardware. It was more like a video games console than a workstation. They had a few years headstart on everyone else and then having (then dated) specialized hardware became a liability, not an advantage.