diff --git a/2023/src/bin/day08.rs b/2023/src/bin/day08.rs
index 884d481..8a0db2c 100644
--- a/2023/src/bin/day08.rs
+++ b/2023/src/bin/day08.rs
@@ -7,7 +7,7 @@ use aoc2023::{boilerplate, common::*};
 const DAY: usize = 08;
 type Parsed<'a> = (Vec<Direction>, HashMap<&'a str, (&'a str, &'a str)>);
 
-#[derive(Debug, PartialEq, Copy, Clone)]
+#[derive(Debug, PartialEq, Eq, Copy, Clone, Hash)]
 enum Direction {
     Left,
     Right,
@@ -27,30 +27,48 @@ fn parse_input(raw: &str) -> Parsed {
     (directions, map)
 }
 
-fn part1((directions, map): &Parsed) -> usize {
+fn part1(parsed: &Parsed) -> usize {
+    steps_until(parsed, "AAA", "ZZZ")
+}
+
+fn steps_until((directions, map): &Parsed, start: &str, target: &str) -> usize {
     directions
         .iter()
         .cycle()
-        .scan("AAA", |pos, dir| {
+        .scan(start, |pos, dir| {
             let next = match dir {
                 Direction::Left => map.get(pos)?.0,
                 Direction::Right => map.get(pos)?.1,
             };
-            if next == "ZZZ" {
-                None
-            } else {
+            (!next.ends_with(target)).then(|| {
                 *pos = next;
                 Some(next)
-            }
+            })
         })
         .count()
         + 1
 }
 
+// I’m honestly not sure why this works. It seems each path only has a single ghost node, and that
+// node occurs right before looping, so we can just compute the least common multiple of their step counts.
+// I assume this holds true for other inputs (it can’t have been random),
+// but I don’t see it anywhere in the task and only found out by experimentation.
 fn part2(parsed: &Parsed) -> usize {
-    unimplemented!()
+    parsed.1.keys().filter(|start| start.ends_with("A")).map(|start| steps_until(parsed, start, "Z")).fold(1, |acc, n| lcm(acc, n))
 }
 
+#[cfg(test)]
+const TEST_INPUT_P2: &str = "LR
+
+11A = (11B, XXX)
+11B = (XXX, 11Z)
+11Z = (11B, XXX)
+22A = (22B, XXX)
+22B = (22C, 22C)
+22C = (22Z, 22Z)
+22Z = (22B, 22B)
+XXX = (XXX, XXX)";
+
 boilerplate! {
     TEST_INPUT == "LLR
 
@@ -59,9 +77,9 @@ BBB = (AAA, ZZZ)
 ZZZ = (ZZZ, ZZZ)",
     tests: {
         part1: { TEST_INPUT => 6 },
-        part2: { TEST_INPUT => 0 },
+        part2: { TEST_INPUT_P2 => 6 },
     },
     bench1 == 12083,
-    bench2 == 0,
+    bench2 == 13385272668829,
     bench_parse: |(v, m): &Parsed| { assert_eq!(m["AAA"], ("MJJ", "QBJ")); (v.len(), v[0]) } => (281, Direction::Left),
 }
diff --git a/2023/src/common.rs b/2023/src/common.rs
index b0a69de..6eb97c8 100644
--- a/2023/src/common.rs
+++ b/2023/src/common.rs
@@ -68,3 +68,17 @@ pub fn parse_num<I: ParseableNumber<I>>(s: &str) -> I {
     let start = unsafe { digits.next().unwrap_unchecked() };
     digits.fold(start, |acc, n| acc * I::from(10) + n)
 }
+
+fn gcd(mut x: usize, mut y: usize) -> usize {
+    let mut remainder;
+    while y != 0 {
+        remainder = x % y;
+        x = y;
+        y = remainder;
+    }
+    x
+}
+
+pub fn lcm(x: usize, y: usize) -> usize {
+    x * y / gcd(x, y)
+}