diff --git a/2015/day19.py b/2015/day19.py
new file mode 100644
index 0000000..f3dc0fe
--- /dev/null
+++ b/2015/day19.py
@@ -0,0 +1,56 @@
+import re
+with open('day19-input', 'r') as file:
+	replacements, _, medicine_molecule = file.read().strip().partition('\n\n')
+replacements = replacements.strip().split('\n')
+replacements = [(old, new) for old, _, new in (rep.partition(' => ') for rep in replacements)]
+medicine_molecule = medicine_molecule.strip()
+
+
+unique_molecules = set()
+for old, new in replacements:
+	r = re.compile(old)
+	for match in r.finditer(medicine_molecule):
+		unique_molecules.add(medicine_molecule[:match.start()] + new + medicine_molecule[match.end():])
+
+print(f'Part 1: {len(unique_molecules)}')
+
+
+starting_molecule = 'e'
+replacements_ordered = sorted(replacements, key=lambda x: (len(x[1]), x[1]))  # Shortest output to longest output
+regexes_ordered = [(old, re.compile(new)) for old, new in replacements_ordered]
+biggest_reduction = max((len(new) - len(old) for old, new in replacements_ordered))
+print(f'Biggest reducing step is {biggest_reduction}')
+
+# stack = [(0, medicine_molecule)]  # We will work backwards from medicine to 'e' as it's easier to not make garbage
+stack = {medicine_molecule: 0}  # We will work backwards from medicine to 'e' as it's easier to not make garbage
+best = 500  # Set a hard limit to prevent any potential cycles going forever
+best_molecules = {}
+
+while stack:
+	# Retrieve state
+	# replacements, molecule = stack.pop()
+	molecule, replacements = stack.popitem()
+	# Don't process if this is already worse
+	if replacements >= best:
+		continue
+	# If the best case can't reduce enough, abort
+	target_reduction = len(molecule) - 1
+	min_replacements = -(-target_reduction//biggest_reduction)  # ceildiv
+	if replacements + min_replacements > best:
+		continue
+	# If we have worked back to the starting molecule, record this result
+	if molecule == starting_molecule:
+		best = replacements
+		print(f'New best found: {best}')
+		continue
+
+	r1 = replacements + 1
+	for old, r in regexes_ordered:
+		for match in r.finditer(molecule):
+			s = molecule[:match.start()] + old + molecule[match.end():]
+			if s not in best_molecules or best_molecules[s] < r1:
+				best_molecules[s] = r1
+				# stack.append((r1, s))
+				stack[s] = r1
+
+print(f'Part 2: fewest replacements = {best}')