Are you talking about what I know as Machine code? Binary is just a numbering system with two states, 0 and 1. It's a bit boring as it stands, so processor commands were usually input in hexadecimal; a numbering system going from 1 to F, er.. counting in 16s. The symbols used are 0123456789abcdef. Processor commands would be generally input as quads on a 32 bit system, i.e. 29C3. You'd have codes which would allow registers within the CPU to be loaded and manipulated with hex data. This all at a very basic level. One code may load a register with a hex number, bits within this register would be set or not set which would affect other stuff happening, etc. you'd need a code just to load the register, another to shift the data left or right by 1 bit, etc,etc. So doing something we'd think was quite simple, like adding 1 + 1, when coded, would be quite complex.
This is why high level programming languages were developed, it made life simpler for humans. It does generate a load of redundant code though; unused libraries, etc. Machine code, when it worked, was very lean, fast, and took little memoryspace. The Sinclair ZX81 only had 1K of memory!