diff --git a/ex.py b/ex.py
index d0f4abe..41a69b6 100644
--- a/ex.py
+++ b/ex.py
@@ -1,3 +1,4 @@
+from gbso.program.mutate import create_random_program
 from time import time
 
 from gbso.cpu.state import CPUState
@@ -20,11 +21,12 @@ prgm = Program(
 outputs = [R8.A]
 test_cases = [TestCase()]
 max_size = 4
-num_iters = 100_000_000
+num_iters = 1_000_000
 
 initial_cost = cost(prgm, test_cases, outputs, prgm)
 initial_cycles = prgm.perf()
 
+print("Program to optimize:")
 prgm.display()
 
 print(f"Cost: {initial_cost}")
@@ -34,6 +36,7 @@ start_time = time()
 
 optimized_prgm = optimize(
     prgm,
+    init_prgm=create_random_program(max_size),
     max_size=max_size,
     test_cases=test_cases,
     outputs=outputs,
@@ -42,6 +45,7 @@ optimized_prgm = optimize(
 
 end_time = time()
 
+print("Optimized result:")
 optimized_prgm.display()
 
 optimized_cost = cost(prgm, test_cases, outputs, optimized_prgm)
diff --git a/gbso/optimize.py b/gbso/optimize.py
index b20f167..881e472 100644
--- a/gbso/optimize.py
+++ b/gbso/optimize.py
@@ -1,6 +1,6 @@
-from math import exp, log
+from math import log
 from random import random
-from typing import List, Tuple
+from typing import List, Optional, Tuple
 
 from gbso.program.test_case import Output, TestCase, eq_on_testcase
 from gbso.program.mutate import mutate_program
@@ -31,11 +31,12 @@ def cost(orig_prgm, test_cases, outputs, prgm) -> Tuple[int, bool]:
 
 
 def optimize(
-    prgm: Program,
+    target_prgm: Program,
     max_size: int,
     test_cases: List[TestCase],
     outputs: List[Output],
     beta: int = 0.5,  # How far away in cost you are allowed to search
+    init_prgm: Optional[Program] = None,
     num_iters: int = DEFAULT_NUM_ITERS,
     prob_opcode: float = DEFAULT_PROB_OPCODE,
     prob_operand: float = DEFAULT_PROB_OPERAND,
@@ -43,31 +44,32 @@ def optimize(
     prob_insn: float = DEFAULT_PROB_INSN,
     prob_insn_unused: float = DEFAULT_PROB_INSN_UNUSED,
 ) -> Program:
-    padded_prgm = prgm.pad(max_size)
+    padded_prgm = (init_prgm or target_prgm).pad(max_size)
 
     last_prgm = padded_prgm
-    last_cost, _last_eq = cost(padded_prgm, test_cases, outputs, last_prgm)
+    last_cost, _last_eq = cost(target_prgm, test_cases, outputs, last_prgm)
 
-    best_prgm = padded_prgm
-    best_cost = last_cost
+    best_prgm = target_prgm.pad(max_size)
+    best_cost = 0
+
+    num_candidates = 0
 
     for _ in range(num_iters):
         candidate_prgm = mutate_program(
             last_prgm, prob_opcode, prob_operand, prob_swap, prob_insn, prob_insn_unused
         )
         candidate_cost, candidate_eq = cost(
-            padded_prgm, test_cases, outputs, candidate_prgm
+            target_prgm, test_cases, outputs, candidate_prgm
         )
 
         if candidate_cost < best_cost and candidate_eq:
             best_prgm = candidate_prgm
             best_cost = candidate_cost
+            num_candidates += 1
 
         if candidate_cost < last_cost - log(random()) / beta:
             last_prgm = candidate_prgm
             last_cost = candidate_cost
 
-    print("last")
-    last_prgm.display()
-
+    print(f"Optimization complete. Total candidates: {num_candidates}")
     return best_prgm
diff --git a/gbso/program/mutate.py b/gbso/program/mutate.py
index 8c1cd3f..025bdd3 100644
--- a/gbso/program/mutate.py
+++ b/gbso/program/mutate.py
@@ -146,3 +146,13 @@ def create_random_operand(op: Operand) -> Union[int, R8, R16]:
         value = randint(-128, 127)
 
     return value
+
+
+def create_random_program(size: int) -> Program:
+    insns = []
+
+    for _ in range(size):
+        cls = choice(ALL_INSN_CLASSES + [UNUSED])
+        insns.append(create_random_insn(cls))
+
+    return Program(insns=insns)