diff --git a/assembler/grammar/assembly.pest b/assembler/grammar/assembly.pest
index 530aa40..82cea9e 100644
--- a/assembler/grammar/assembly.pest
+++ b/assembler/grammar/assembly.pest
@@ -3,7 +3,7 @@ comment = { ";" ~ (!"\n" ~ ANY)* ~ "\n" }
 label = { (!":" ~ !"\n" ~ ANY)+ ~ ":" ~ "\n" }
 instruction = { mnemonic ~ (operand ~ ", "?)+? ~ "\n"? }
 mnemonic = { 
-    "lw" | "sw" | "jalv" | "jal" | "beqi" | "beq" | "bnei" | "bne" | "imm32" | "stop" |
+    "lw" | "sw" | "loadu8" | "loads8" | "storeu8" | "jalv" | "jal" | "beqi" | "beq" | "bnei" | "bne" | "imm32" | "stop" |
     "advread" | "advwrite" | 
     "addi" | "add" | "subi" | "sub" | "muli" | "mul" | "mulhsi"| "mulhui"| "mulhs"| "mulhu"  | "divi" | "div" | "sdiv" | "sdivi" |
     "ilte" | "ltei" | "lte" | "ilt" | "lti" | "lt" | "shli" | "shl" | "shri" | "shr" | "srai" | "sra" |
diff --git a/assembler/src/lib.rs b/assembler/src/lib.rs
index f24e48f..b3ed64b 100644
--- a/assembler/src/lib.rs
+++ b/assembler/src/lib.rs
@@ -58,7 +58,10 @@ pub fn assemble(input: &str) -> Result<Vec<u8>, String> {
                 let opcode = match mnemonic {
                     // Core CPU
                     "lw" => LOAD32,
+                    "loadu8" => LOADU8,
+                    "loads8" => LOADS8,
                     "sw" => STORE32,
+                    "storeu8" => STOREU8,
                     "jal" => JAL,
                     "jalv" => JALV,
                     "beq" | "beqi" => BEQ,
@@ -101,12 +104,12 @@ pub fn assemble(input: &str) -> Result<Vec<u8>, String> {
 
                 // Insert zero operands if necessary
                 match mnemonic {
-                    "lw" => {
+                    "lw" | "loadu8" | "loads8" => {
                         // (a, 0, c, 0, 0)
                         operands.insert(1, 0);
                         operands.extend(vec![0; 2]);
                     }
-                    "sw" => {
+                    "sw" | "storeu8" => {
                         // (0, b, c, 0, 0)
                         operands.insert(0, 0);
                         operands.extend(vec![0; 2]);
diff --git a/basic/src/bin/valida.rs b/basic/src/bin/valida.rs
index 71aa70e..c59b45e 100644
--- a/basic/src/bin/valida.rs
+++ b/basic/src/bin/valida.rs
@@ -166,7 +166,7 @@ fn last_frame(_: ArgMatches, context: &mut Context) -> Result<Option<String>> {
     frame += &format!("Current FP: 0x{:x}\n", fp).as_str();
 
     // print last frame
-    for i in (-5..(last_size / 4) + 1).rev() {
+    for i in (-10..(last_size / 4) + 1).rev() {
         let offset = (i * 4) as i32;
         let read_addr = (fp + offset) as u32;
         let string_val = context.machine_.mem().examine(read_addr);
diff --git a/basic/src/lib.rs b/basic/src/lib.rs
index 55fea07..67fe80f 100644
--- a/basic/src/lib.rs
+++ b/basic/src/lib.rs
@@ -37,8 +37,8 @@ use valida_bus::{
 };
 use valida_cpu::{
     BeqInstruction, BneInstruction, Imm32Instruction, JalInstruction, JalvInstruction,
-    Load32Instruction, LoadFpInstruction, ReadAdviceInstruction, StopInstruction,
-    Store32Instruction,
+    Load32Instruction, LoadFpInstruction, LoadS8Instruction, LoadU8Instruction,
+    ReadAdviceInstruction, StopInstruction, Store32Instruction, StoreU8Instruction,
 };
 use valida_cpu::{CpuChip, MachineWithCpuChip};
 use valida_machine::__internal::p3_challenger::{CanObserve, FieldChallenger};
@@ -63,7 +63,11 @@ use valida_machine::StarkConfig;
 pub struct BasicMachine<F: PrimeField32 + TwoAdicField> {
     // Core instructions
     load32: Load32Instruction,
+    loadu8: LoadU8Instruction,
+    loads8: LoadS8Instruction,
     store32: Store32Instruction,
+    storeu8: StoreU8Instruction,
+
     jal: JalInstruction,
     jalv: JalvInstruction,
     beq: BeqInstruction,
@@ -1071,9 +1075,18 @@ impl<F: PrimeField32 + TwoAdicField> Machine<F> for BasicMachine<F> {
             <Load32Instruction as Instruction<Self, F>>::OPCODE => {
                 Load32Instruction::execute_with_advice::<Adv>(self, ops, advice)
             }
+            <LoadU8Instruction as Instruction<Self, F>>::OPCODE => {
+                LoadU8Instruction::execute_with_advice::<Adv>(self, ops, advice)
+            }
+            <LoadS8Instruction as Instruction<Self, F>>::OPCODE => {
+                LoadS8Instruction::execute_with_advice::<Adv>(self, ops, advice)
+            }
             <Store32Instruction as Instruction<Self, F>>::OPCODE => {
                 Store32Instruction::execute_with_advice::<Adv>(self, ops, advice)
             }
+            <StoreU8Instruction as Instruction<Self, F>>::OPCODE => {
+                StoreU8Instruction::execute_with_advice::<Adv>(self, ops, advice)
+            }
             <JalInstruction as Instruction<Self, F>>::OPCODE => {
                 JalInstruction::execute_with_advice::<Adv>(self, ops, advice)
             }
diff --git a/cpu/src/columns.rs b/cpu/src/columns.rs
index b19bb1f..589cb99 100644
--- a/cpu/src/columns.rs
+++ b/cpu/src/columns.rs
@@ -48,7 +48,10 @@ pub struct OpcodeFlagCols<T> {
     pub is_bus_op_with_mem: T,
     pub is_imm_op: T,
     pub is_load: T,
+    pub is_load_u8: T,
+    pub is_load_s8: T,
     pub is_store: T,
+    pub is_store_u8: T,
     pub is_beq: T,
     pub is_bne: T,
     pub is_jal: T,
diff --git a/cpu/src/lib.rs b/cpu/src/lib.rs
index 2273f2c..ded7772 100644
--- a/cpu/src/lib.rs
+++ b/cpu/src/lib.rs
@@ -11,11 +11,13 @@ use core::marker::Sync;
 use core::mem::transmute;
 use valida_bus::{MachineWithGeneralBus, MachineWithMemBus, MachineWithProgramBus};
 use valida_machine::{
-    instructions, AdviceProvider, Chip, Instruction, InstructionWord, Interaction, Operands, Word,
+    addr_of_word, index_of_byte, instructions, AdviceProvider, Chip, Instruction, InstructionWord,
+    Interaction, Operands, Word,
 };
 use valida_memory::{MachineWithMemoryChip, Operation as MemoryOperation};
 use valida_opcodes::{
-    BEQ, BNE, BYTES_PER_INSTR, IMM32, JAL, JALV, LOAD32, LOADFP, READ_ADVICE, STOP, STORE32,
+    BEQ, BNE, BYTES_PER_INSTR, IMM32, JAL, JALV, LOAD32, LOADFP, LOADS8, LOADU8, READ_ADVICE, STOP,
+    STORE32, STOREU8,
 };
 
 use p3_air::VirtualPairCol;
@@ -31,7 +33,10 @@ pub mod stark;
 #[derive(Clone)]
 pub enum Operation {
     Store32,
+    StoreU8,
     Load32,
+    LoadU8,
+    LoadS8,
     Jal,
     Jalv,
     Beq(Option<Word<u8>> /*imm*/),
@@ -175,6 +180,15 @@ impl CpuChip {
             Operation::Load32 => {
                 cols.opcode_flags.is_load = SC::Val::one();
             }
+            Operation::StoreU8 => {
+                cols.opcode_flags.is_store_u8 = SC::Val::one();
+            }
+            Operation::LoadU8 => {
+                cols.opcode_flags.is_load_u8 = SC::Val::one();
+            }
+            Operation::LoadS8 => {
+                cols.opcode_flags.is_load_s8 = SC::Val::one();
+            }
             Operation::Jal => {
                 cols.opcode_flags.is_jal = SC::Val::one();
             }
@@ -349,7 +363,10 @@ pub trait MachineWithCpuChip<F: Field>: MachineWithMemoryChip<F> {
 
 instructions!(
     Load32Instruction,
+    LoadU8Instruction,
+    LoadS8Instruction,
     Store32Instruction,
+    StoreU8Instruction,
     JalInstruction,
     JalvInstruction,
     BeqInstruction,
@@ -441,6 +458,119 @@ where
     }
 }
 
+impl<M, F> Instruction<M, F> for LoadU8Instruction
+where
+    M: MachineWithCpuChip<F>,
+    F: Field,
+{
+    const OPCODE: u32 = LOADU8;
+
+    fn execute(state: &mut M, ops: Operands<i32>) {
+        let opcode = <Self as Instruction<M, F>>::OPCODE;
+        let clk = state.cpu().clock;
+        let pc = state.cpu().pc;
+        let fp = state.cpu().fp;
+
+        let read_addr_loc = (fp as i32 + ops.c()) as u32;
+
+        let read_addr = state
+            .mem_mut()
+            .read(clk, read_addr_loc, true, pc, opcode, 0, "");
+        let read_addr_index = addr_of_word(read_addr.into());
+
+        // The word from the read address.
+        let cell = state.mem_mut().read(
+            clk,
+            read_addr_index,
+            true,
+            pc,
+            opcode,
+            1,
+            &format!(
+                "fp = {}, c = {}, [fp+c] = {:?}",
+                fp as i32,
+                ops.c() as u32,
+                read_addr_index
+            ),
+        );
+
+        // The array index of the word for the byte to read from
+        let index_of_read = index_of_byte(read_addr.into());
+        // The byte from the read cell.
+        let cell_byte = cell[index_of_read];
+
+        let write_addr = (state.cpu().fp as i32 + ops.a()) as u32;
+        // The address, converted to a multiple of 4.
+        let write_addr_index = addr_of_word(write_addr);
+
+        // The Word to write, with one byte overwritten to the read byte
+        let cell_to_write = Word::zero_extend_byte(cell_byte);
+
+        state
+            .mem_mut()
+            .write(clk, write_addr_index, cell_to_write, true);
+        state.cpu_mut().pc += 1;
+        state.cpu_mut().push_op(Operation::LoadU8, opcode, ops);
+    }
+}
+
+impl<M, F> Instruction<M, F> for LoadS8Instruction
+where
+    M: MachineWithCpuChip<F>,
+    F: Field,
+{
+    const OPCODE: u32 = LOADS8;
+
+    fn execute(state: &mut M, ops: Operands<i32>) {
+        let opcode = <Self as Instruction<M, F>>::OPCODE;
+        let clk = state.cpu().clock;
+        let pc = state.cpu().pc;
+        let fp = state.cpu().fp;
+
+        let read_addr_loc = (fp as i32 + ops.c()) as u32;
+
+        let read_addr = state
+            .mem_mut()
+            .read(clk, read_addr_loc, true, pc, opcode, 0, "");
+
+        let read_addr_index = addr_of_word(read_addr.into());
+
+        // The word from the read address.
+        let cell = state.mem_mut().read(
+            clk,
+            read_addr_index,
+            true,
+            pc,
+            opcode,
+            1,
+            &format!(
+                "fp = {}, c = {}, [fp+c] = {:?}",
+                fp as i32,
+                ops.c() as u32,
+                read_addr_index
+            ),
+        );
+
+        // The array index of the word for the byte to read from
+        let index_of_read = index_of_byte(read_addr.into());
+        // The byte from the read cell.
+        let cell_byte = cell[index_of_read];
+
+        let write_addr = (state.cpu().fp as i32 + ops.a()) as u32;
+        // The address, converted to a multiple of 4.
+        let write_addr_index = addr_of_word(write_addr);
+
+        // The Word to write, with one byte overwritten to the read byte
+        let cell_to_write = Word::sign_extend_byte(cell_byte);
+
+        state
+            .mem_mut()
+            .write(clk, write_addr_index, cell_to_write, true);
+        state.cpu_mut().pc += 1;
+        state.cpu_mut().push_op(Operation::LoadS8, opcode, ops);
+    }
+}
+
 impl<M, F> Instruction<M, F> for Store32Instruction
 where
     M: MachineWithCpuChip<F>,
@@ -456,7 +586,7 @@ where
         let pc = state.cpu().pc;
         let write_addr = state
             .mem_mut()
-            .read(clk, write_addr_loc.into(), true, pc, opcode, 0, "");
+            .read(clk, write_addr_loc, true, pc, opcode, 0, "");
         let cell = state
             .mem_mut()
             .read(clk, read_addr, true, pc, opcode, 1, "");
@@ -466,6 +596,59 @@ where
     }
 }
 
+impl<M, F> Instruction<M, F> for StoreU8Instruction
+where
+    M: MachineWithCpuChip<F>,
+    F: Field,
+{
+    const OPCODE: u32 = STOREU8;
+
+    fn execute(state: &mut M, ops: Operands<i32>) {
+        let opcode = <Self as Instruction<M, F>>::OPCODE;
+        let clk = state.cpu().clock;
+        let read_addr = (state.cpu().fp as i32 + ops.c()) as u32;
+
+        // Make sure we get to the correct and non empty map for the byte.
+        let read_addr_index = addr_of_word(read_addr);
+        let write_addr_loc = (state.cpu().fp as i32 + ops.b()) as u32;
+        let pc = state.cpu().pc;
+        let write_addr = state
+            .mem_mut()
+            .read(clk, write_addr_loc.into(), true, pc, opcode, 0, "");
+
+        // Read the cell from the read address.
+        let cell = state
+            .mem_mut()
+            .read(clk, read_addr_index, true, pc, opcode, 1, "");
+
+        // The array index of the word for the byte to read from
+        let index_of_read = index_of_byte(read_addr);
+
+        // The word from the read address.
+        let cell_read = cell.0;
+        // The byte from the read cell.
+        let cell_byte = cell_read[index_of_read];
+
+        // The array index of the word for the byte to write to
+        let index_of_write = index_of_byte(write_addr.into());
+
+        // The key to the memory map, converted to a multiple of 4.
+        let write_addr_index = addr_of_word(write_addr.into());
+
+        // The original content of the cell to write to. If the cell is empty, initiate it with a default value.
+        let cell_write = state.mem_mut().read_or_init(clk, write_addr_index, true);
+
+        // The Word to write, with one byte overwritten to the read byte
+        let cell_to_write = cell_write.update_byte(cell_byte, index_of_write);
+
+        state
+            .mem_mut()
+            .write(clk, write_addr_index, cell_to_write, true);
+        state.cpu_mut().pc += 1;
+        state.cpu_mut().push_op(Operation::StoreU8, opcode, ops);
+    }
+}
+
 impl<M, F> Instruction<M, F> for JalInstruction
 where
     M: MachineWithCpuChip<F>,
diff --git a/machine/src/core.rs b/machine/src/core.rs
index e37aabb..4ff8d2f 100644
--- a/machine/src/core.rs
+++ b/machine/src/core.rs
@@ -7,6 +7,18 @@ use p3_field::{Field, PrimeField};
 #[derive(Copy, Clone, Debug, Default)]
 pub struct Word<F>(pub [F; MEMORY_CELL_BYTES]);
 
+// Functions for byte manipulations
+/// Get the index of a byte in a memory cell.
+pub fn index_of_byte(addr: u32) -> usize {
+    (addr & 3) as usize
+}
+
+/// Get the address of the memory cells which is not empty (a multiple of 4).
+pub fn addr_of_word(addr: u32) -> u32 {
+    (addr & !3) as u32
+}
+//----------------------------------
+
 impl Word<u8> {
     pub fn from_u8(byte: u8) -> Self {
         let mut result = [0; MEMORY_CELL_BYTES];
@@ -15,6 +27,31 @@ impl Word<u8> {
     }
 }
 
+impl Word<u8> {
+    pub fn sign_extend_byte(byte: u8) -> Self {
+        let sign = byte as i8 >> 7;
+        let mut result: [u8; MEMORY_CELL_BYTES] = [sign as u8; MEMORY_CELL_BYTES];
+        result[0] = byte;
+        Self(result)
+    }
+}
+
+impl Word<u8> {
+    pub fn zero_extend_byte(byte: u8) -> Self {
+        let mut result: [u8; MEMORY_CELL_BYTES] = [0; MEMORY_CELL_BYTES];
+        result[0] = byte;
+        Self(result)
+    }
+}
+
+impl Word<u8> {
+    pub fn update_byte(self, byte: u8, loc: usize) -> Self {
+        let mut result: [u8; MEMORY_CELL_BYTES] = self.0;
+        result[loc] = byte;
+        Self(result)
+    }
+}
+
 impl<F: Copy> Word<F> {
     pub fn transform<T, G>(self, mut f: G) -> Word<T>
     where
diff --git a/machine/src/program.rs b/machine/src/program.rs
index f61bcba..3032f8c 100644
--- a/machine/src/program.rs
+++ b/machine/src/program.rs
@@ -102,9 +102,18 @@ impl InstructionWord<i32> {
             valida_opcodes::LOAD32 => {
                 format!("{}(fp), {}(fp)", self.operands.0[0], self.operands.0[2])
             }
+            valida_opcodes::LOADU8 => {
+                format!("{}(fp), {}(fp)", self.operands.0[0], self.operands.0[2])
+            }
+            valida_opcodes::LOADS8 => {
+                format!("{}(fp), {}(fp)", self.operands.0[0], self.operands.0[2])
+            }
             valida_opcodes::STORE32 => {
                 format!("{}(fp), {}(fp)", self.operands.0[1], self.operands.0[2])
             }
+            valida_opcodes::STOREU8 => {
+                format!("{}(fp), {}(fp)", self.operands.0[1], self.operands.0[2])
+            }
             _ => {
                 format!(
                     "{}(fp), {}, {}",
diff --git a/memory/src/lib.rs b/memory/src/lib.rs
index 0c9725a..1bb80ea 100644
--- a/memory/src/lib.rs
+++ b/memory/src/lib.rs
@@ -2,6 +2,8 @@
 
 extern crate alloc;
 
+use valida_machine::MEMORY_CELL_BYTES;
+
 use crate::alloc::string::ToString;
 use crate::columns::{MemoryCols, MEM_COL_MAP, NUM_MEM_COLS};
 use alloc::collections::BTreeMap;
@@ -79,6 +81,7 @@ impl MemoryChip {
         }
     }
 
+    /// Read from a cell. If the cell is empty, panic.
     pub fn read(
         &mut self,
         clk: u32,
@@ -100,6 +103,22 @@ impl MemoryChip {
         value
     }
 
+    /// Read from a cell. If the cell is empty, initialize it with the default values.
+    pub fn read_or_init(&mut self, clk: u32, address: u32, log: bool) -> Word<u8> {
+        let value = self
+            .cells
+            .get(&address.into())
+            .copied()
+            .unwrap_or_else(|| Word([0; MEMORY_CELL_BYTES]));
+        if log {
+            self.operations
+                .entry(clk)
+                .or_insert_with(Vec::new)
+                .push(Operation::Read(address.into(), value));
+        }
+        value
+    }
+
     pub fn write(&mut self, clk: u32, address: u32, value: Word<u8>, log: bool) {
         if log {
             self.operations
diff --git a/opcodes/src/lib.rs b/opcodes/src/lib.rs
index 161be3f..5ffd910 100644
--- a/opcodes/src/lib.rs
+++ b/opcodes/src/lib.rs
@@ -15,6 +15,10 @@ pub enum Opcode {
     STOP = 8,
     READ_ADVICE = 9,
     LOADFP = 10,
+    LOADU8 = 11,
+    LOADS8 = 12,
+    STOREU8 = 13,
+
     ADD32 = 100,
     SUB32 = 101,
     MUL32 = 102,
@@ -56,6 +60,9 @@ declare_opcode!(BNE);
 declare_opcode!(IMM32);
 declare_opcode!(STOP);
 declare_opcode!(LOADFP);
+declare_opcode!(LOADU8);
+declare_opcode!(LOADS8);
+declare_opcode!(STOREU8);
 
 /// NONDETERMINISTIC
 declare_opcode!(READ_ADVICE);