Assembly language is not difficult, but it's finicky. You have to be very closely aware of the hardware you're running on, especially the CPU and its registers. You need to learn its instruction set, and its idiosyncrasies - what operations set what flags? Does just loading a register set flags, or do you have to do an operation on it first? Does INC change the Carry flag? Some study required. You have to initialize all variables yourself, or be aware which ones need it and which don't. There is no type casting in assembler - all variables are the same, just labels associated with addresses in RAM. You have to be disciplined with variables too, since they are all global in assembler. If you need math functions, other than the simple add subtract multiply and (sometimes) divide that come with the CPU, you have to write them yourself. And of course there are no convenient library functions you can include for anything else, either. If you can't find the source for a driver for (say) an LCD that someone else wrote, guess who's going to write it! (Be generous and share.)

It's possible to write good, clear, structured code in assembler that is usually a little smaller and a little faster than a C program, although it's possible to write absolutely terrible code as well, since assembler imposes no discipline on you. Skill comes with experience. I grew up with assembler and came late to C, so I find C irritating and restrictive - goddam case sensitivity, aggravating function prototypes, and AAAARGHH!!! macros that look exactly like f'ing functions and you can't find them when you're trying to edit someone's code. Over the years I've built up a huge collection of code I can copy and paste to do nearly everything, and I can usually turn out a working assembler program faster than the next guy can do it in C. I can certainly debug mine faster than he can, since the debugger presents in assembler, and I have the commented source code and can follow it line by line.