diff --git a/MIPS.txt b/MIPS.txt
new file mode 100644
index 0000000..5b0b3d7
--- /dev/null
+++ b/MIPS.txt
@@ -0,0 +1,60 @@
+MIPS 32 bits
+
+== Bits, Bytes, Hex ==
+-- 8 bits = 1 byte = 2 hex
+-- 32 bits = 4 bytes = 8 hex
+-- e.g. 00000000 00000000 00000000 00000000 -> 0x00000000
+
+== CPU ==
+inputs: 
+-- manual MIPS assembly code -> instructions in binaries
+-- C code -> compiled c program under mips -> disassemble binaries -> assembly code -> instruction in binaries
+-- these binary instructions goes into instruction memory
+outputs:
+-- output of the instructions
+errors:
+-- ?????? how do we detect errors ??????
+
+== Submodules ==
+-- ALU
+-- Register File
+-- Data Memory
+-- Instruction Register
+-- PC
+-- Control Unit
+
+== Testbench ==
+-- ????? not so sure yet ?????
+
+== Endianess ==
+-- big endian: bytes are numbered starting with byte 0 at MSB
+-- use -EB flag to ensure big endian
+
+== Instruction Access ==
+-- PC (program counter): 32 bit register
+-- PC is initialised to 0xBFC00000
+-- PC changed as instructions are executed
+-- IR = Mem[PC] -> instruction is fetched from data memory using data at the address given by program counter
+
+Address bus: CPU -> Memory
+Data bus: CPU <=> Memory
+
+== Register File ==
+-- 32 general-purpose registers
+
+== Program Counter ==
+-- PC is just a 32 bit register in which the value (address of instruction) get updated by other blocks
+-- Controlled by PCSrc (for branching or regular increment by 4 bytes)
+
+== Questions ==
+Pseudo-instructions -> how to deal with them -> convert to actual instructions?
+Do we implement big-endian mips?
+What verilator could be useful for
+How would a testbench in c++ be helpful?
+Don't understand that part where we need to implement cache
+
+== Todo ==
+Testbench in c++
+Cache
+CPU stall cycle
+