Kotlin

Part 1 is easily solved by simulating the beams and modifying the grid in-place.

This does not work for part 2, however. Here I opted for a BFS algorithm, moving down row-by-row.
Similar to other solutions, I store the amount of incoming paths to a splitter, so that I can reference it later. Practically, the two beams from each splitter are representatives of all incoming beams. This reduces the search complexity by a lot!
The final count of timelines is the sum of the array that collects the incoming beam counts for each bottom-row of the diagram.

Code on GitHub

class Day07 : AOCSolution {
    override val year = 2025
    override val day = 7

    override fun part1(inputFile: String): String {
        val diagram = readResourceLines(inputFile)
            .mapArray { line -> line.mapArray { char -> Cell.byChar(char) } }
            .toGrid()

        var count = 0
        for (y in 1 until diagram.height) {
            for (x in 0 until diagram.width) {
                // Search for beam sources in the preceding row
                if (diagram[x, y - 1] in Cell.beamSourceTypes) {
                    // Simulate the beam moving down
                    when (diagram[x, y]) {
                        Cell.Empty -> diagram[x, y] = Cell.Beam
                        Cell.Splitter -> {
                            // Split the beam and count this splitter
                            diagram[x - 1, y] = Cell.Beam
                            diagram[x + 1, y] = Cell.Beam
                            count++
                        }

                        else -> continue
                    }
                }
            }
        }

        return count.toString()
    }

    override fun part2(inputFile: String): String {
        val diagram = readResourceLines(inputFile)
            .mapArray { line -> line.mapArray { char -> Cell.byChar(char) } }
            .toGrid()
        val height = diagram.height.toLong()

        val startPosition = diagram.positionOfFirst { it == Cell.Start }

        // Working stack of beam origin and split origin positions
        val stack = ArrayDeque<Pair<Position, Position>>()
        stack.add(startPosition to startPosition)

        // Splitter positions mapped to the count of timelines to them
        // Start with the start position and 1 timeline.
        val splitters = mutableMapOf<Position, Long>(startPosition to 1)

        // Keep track of all splitters for which new beams have been spawned already
        // Could be used to solve part 1, as well
        val spawnedSplitters = mutableSetOf<Position>()

        // Count the timelines per diagram exit, which is the bottom-most row
        val diagramExits = LongArray(diagram.width)

        while (stack.isNotEmpty()) {
            // Breadth first search for memorizing the amount of paths to a splitter
            val (beamOrigin, splitOrigin) = stack.poll()
            val originPathCount = splitters.getValue(splitOrigin)

            val nextPosition = beamOrigin + Direction.DOWN

            if (nextPosition.y < height) {
                if (diagram[nextPosition] == Cell.Splitter) {
                    if (nextPosition !in spawnedSplitters) {
                        // Only spawn new beams, if they weren't spawned already
                        stack.add((nextPosition + Direction.LEFT) to nextPosition)
                        stack.add((nextPosition + Direction.RIGHT) to nextPosition)
                        spawnedSplitters.add(nextPosition)
                        // Initialize the count
                        splitters[nextPosition] = originPathCount
                    } else {
                        splitters.computeIfPresent(nextPosition) { _, v -> v + originPathCount }
                    }
                } else {
                    // Just move down
                    stack.add(nextPosition to splitOrigin)
                }
            } else {
                diagramExits[nextPosition.x.toInt()] += originPathCount
            }
        }

        // Sum the count of timelines leading to the bottom row, i.e. leaving the diagram for each position
        return diagramExits.sum().toString()
    }

    private companion object {
        enum class Cell(val char: Char) {
            Start('S'),
            Empty('.'),
            Splitter('^'),
            Beam('|');

            override fun toString(): String {
                return char.toString()
            }

            companion object {
                fun byChar(char: Char) = entries.first { it.char == char }

                val beamSourceTypes = arrayOf(Start, Beam)
            }
        }
    }
}

QIR (Quantum Intermediate Representation)

Nah, just kidding - I used Rust. The only tricky part seemed to be finding time on a Sunday to get it coded!

Part 1 I swept down with a bool array for beams and part 2 I swept up with a score array and summed when it split (joined).

struct Teleporter(String);

impl Teleporter {
    fn new(s: String) -> Self {
        Self(s)
    }

    fn start_pos(line: &str) -> Result<usize> {
        line.find('S').ok_or_eyre("Start not found")
    }

    fn splits(&self) -> Result<usize> {
        let mut input = self.0.lines();
        let start_line = input.next().ok_or_eyre("No start line")?;
        let start = Self::start_pos(start_line)?;

        let mut beams = vec![false; start_line.len()];
        beams[start] = true;

        let mut splits = 0;
        for line in input {
            for (i, ch) in line.bytes().enumerate() {
                if beams[i] && ch == b'^' {
                    splits += 1;
                    beams[i] = false;
                    beams[i - 1] = true;
                    beams[i + 1] = true;
                }
            }
        }
        Ok(splits)
    }

    fn timelines(&self) -> Result<usize> {
        let mut input = self.0.lines();
        let start_line = input.next().ok_or_eyre("No start line")?;
        let start = Self::start_pos(start_line)?;

        let mut num_paths = vec![1; start_line.len()];
        for line in input.rev() {
            for (i, c) in line.bytes().enumerate() {
                if c == b'^' || c == b'S' {
                    num_paths[i] = num_paths[i - 1] + num_paths[i + 1];
                }
            }
        }
        Ok(num_paths[start])
    }
}

Rust

Dynamic programming? Nah, just keep track of the number of overlapping beams and part 2 becomes no different to part 1.

View on github

use std::collections::VecDeque;

fn parse_input(input: &str) -> (Vec<Vec<bool>>, (usize, usize)) {
    let splits = input
        .lines()
        .map(|l| l.chars().map(|c| c == '^').collect())
        .collect();
    // Assume start is on first row
    let start = (input.chars().position(|c| c == 'S').unwrap(), 0);
    (splits, start)
}

fn solve(input: String) {
    let (splits, start) = parse_input(&input);
    let mut nsplits = 0u32;
    let mut timelines = 1u64;
    let mut frontier = VecDeque::from([(start, 1)]);
    while let Some((pos, multiplicity)) = frontier.pop_front() {
        let (x, y) = (pos.0, pos.1 + 1);
        if y == splits.len() {
            // Falls out of bottom
            continue;
        }
        if splits[y][x] {
            nsplits += 1;
            timelines += multiplicity;
            if let Some((b, m2)) = frontier.back_mut()
                && *b == (x - 1, y)
            {
                *m2 += multiplicity;
            } else {
                frontier.push_back(((x - 1, y), multiplicity));
            }
            frontier.push_back(((x + 1, y), multiplicity));
        } else if let Some((b, m2)) = frontier.back_mut()
            && *b == (x, y)
        {
            *m2 += multiplicity;
        } else {
            frontier.push_back(((x, y), multiplicity));
        }
    }
    println!("Part 1: {nsplits}");
    println!("Part 2: {timelines}");
}

fn main() -> std::io::Result<()> {
    let (input, _) = util::get_input("day7.txt")?;
    solve(input);
    Ok(())
}

Uiua

A bit late getting to this, but happy with this solution, despite it being nothing more than just building/summing all paths. As usual, you can drop your own file onto that solution.

D ← ".......S.......\n...............\n.......^.......\n...............\n......^.^......\n...............\n.....^.^.^.....\n...............\n....^.^...^....\n...............\n...^.^...^.^...\n...............\n..^...^.....^..\n...............\n.^.^.^.^.^...^.\n..............."
# D ← &fras"AOC2025day07.txt" # <- Uncomment and drop file here.

Parse ← ⊃↘↙1≠@.⊜∘⊸≠@\n
Flow  ← ⊃(+⊃↻₁↻₋₁×|׬)⊙⊸⊣
P₁    ← /+>0♭⬚0×⟜∧(˜⊂↥Flow)
P₂    ← /+⊣∧(˜⊂+Flow)
⊃P₁ P₂ Parse D

Haskell

That was a fun little problem.

import Data.Map qualified as Map  
import Data.Set qualified as Set  
import Data.Tuple (swap)  

readInput s =  
  Map.fromDistinctAscList  
    [((i, j), c) | (i, l) <- zip [0 ..] $ lines s, (j, c) <- zip [0 ..] l]  

beamPaths input = scanl step (Map.singleton startX 1) [startY .. endY]  
  where  
    Just (startY, startX) = lookup 'S' $ map swap $ Map.assocs input  
    Just ((endY, _), _) = Map.lookupMax input  
    step beams y =  
      Map.unionsWith (+) $  
        [ if input Map.!? (y + 1, j) == Just '^'  
            then Map.fromList [(j - 1, n), (j + 1, n)]  
            else Map.singleton j n  
          | (j, n) <- Map.assocs beams  
        ]  

part1 = sum . map Set.size . (zipWith (Set.\\) <*> tail) . map Map.keysSet . beamPaths  

part2 = sum . last . beamPaths  

main = do  
  input <- readInput <$> readFile "input07"  
  print $ part1 input  
  print $ part2 input  

Uiua

Heavily copied ahem, inspired by @[email protected]’s solution :)

Parse     ← °⊂ ▽⊸≡/↥≠@.⊜∘⊸≠@\n
Flow      ← +⊃(/+≡↻1_¯1¤×|×⊙¬)
Propagate ← ˜∧(⊸˜Flow)
Part₁     ← /+♭↧ ⊸Propagate
Part₂     ← /+⊣Propagate ⊙(˜⊂0)

$ .......S.......
$ ...............
$ .......^.......
$ ...............
$ ......^.^......
$ ...............
$ .....^.^.^.....
$ ...............
$ ....^.^...^....
$ ...............
$ ...^.^...^.^...
$ ...............
$ ..^...^.....^..
$ ...............
$ .^.^.^.^.^...^.
$ ...............
&fras "7.txt"

⊃(Part₁|Part₂) Parse

Janet

I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming I hate dynamic programming

• part1
• part2

Kotlin

Tried recursive for part1, didn’t work. LUCKILY for once I was smart and committed anyway, as it was the right solution for part2! I was losing my mind for a bit though as I had originally written my method with Integers…

class Day07 : Puzzle {

    val grid = mutableListOf<MutableList<Char>>()
    val partTwoCache = mutableMapOf<Pair<Int, Int>, Long>()

    override fun readFile() {
        val input = readInputFromFile(2025, 7, false)
        for (line in input.lines().filter { it.isNotBlank() }) {
            grid.add(line.toCharArray().toMutableList())
        }
    }

    override fun solvePartOne(): String {
        grid[1][grid[0].indexOf('S')] = '|'

        var splits = 0
        for (r in 1..<grid.size - 1) {
            for (c in 0..<grid[r].size) {
                if (grid[r][c] == '|') {
                    if (grid[r+1][c] == '.') {
                        grid[r+1][c] = '|'
                    } else if (grid[r+1][c] == '^') {
                        grid[r+1][c-1] = '|'
                        grid[r+1][c+1] = '|'
                        splits++
                    }
                }
            }
        }
        return splits.toString()
    }

    override fun solvePartTwo(): String {
        val start = grid[0].indexOf('S')
        return (1 + processBeamPartTwo(1, start)).toString() // don't forget to count the original timeline
    }

    private fun processBeamPartTwo(row: Int, column: Int): Long {
        if (partTwoCache.contains(Pair(row, column))) {
            return partTwoCache[Pair(row, column)]!!
        }

        if (row == grid.size) return 0L
        if (column == grid[row].size || column < 0) return 0L

        val out = if (grid[row][column] == '^') { // splitter
            1L + processBeamPartTwo(row, column - 1) + processBeamPartTwo(row, column + 1)
        } else {
            processBeamPartTwo(row + 1, column)
        }
        partTwoCache[Pair(row, column)] = out
        return out
    }
}

full code on Codeberg

Python

Not too difficult, especially since there are no edge cases (particles leaving the grid, adjacent splitters).

My initial solution mutated the input for the simulation which I cleaned up after by creating an array that would record the number of particle paths at every column location + some other optimizations. I chose to implement both parts in the same function because they share the majority of the logic.

def solve(data: str):
    grid = data.splitlines()
    m, n = len(grid), len(grid[0])

    # find the first particle
    particle_paths = [0] * n  # count of particle paths that will reach this column
    for j in range(n):
        if grid[0][j] == 'S':
            particle_paths[j] = 1
            break
    
    # count the number of splits for part 1
    splits = 0

    # simulate the particle moving down the grid
    #   optimization 1: we can start from the 3rd row (index 2) because that's where the first splitter is
    #   optimization 2: we can skip alternating rows because every other row is empty
    for i in range(2, m, 2):
        # particle paths per column after this row is processed
        next_particle_paths = [0] * n

        for j in range(n):
            if particle_paths[j] == 0:
                # skip if there are no particle paths coming from above in this column
                continue
            
            if grid[i][j] == '.':
                # no splitter here, the number of paths in this column remains the same
                # make sure to use += to account for neighboring splitters dumping additional paths into this column
                next_particle_paths[j] += particle_paths[j]
            else:
                # splitter activated here, any particle arriving here can end up in the left or right column
                # this can be simulated by adding the number of paths to the columns on either side
                splits += 1
                next_particle_paths[j-1] += particle_paths[j]
                next_particle_paths[j+1] += particle_paths[j]
        
        # update vars for next iteration
        particle_paths = next_particle_paths

    # return both 
    #   the number of splits AND 
    #   the count of timelines a particle would create
    return splits, sum(particle_paths)

sample = """.......S.......
...............
.......^.......
...............
......^.^......
...............
.....^.^.^.....
...............
....^.^...^....
...............
...^.^...^.^...
...............
..^...^.....^..
...............
.^.^.^.^.^...^.
..............."""
assert solve(sample) == (21, 40)

Nim

Another simple one.

Part 1: count each time a beam crosses a splitter.
Part 2: keep count of how many particles are in each column in all universes
(e.g. with a simple 1d array), then sum.

Runtime: 116 μs 95 µs 86 µs

type
  AOCSolution[T,U] = tuple[part1: T, part2: U]

proc solve(input: string): AOCSolution[int, int] =
  var beams = newSeq[int](input.find '\n')
  beams[input.find 'S'] = 1

  for line in input.splitLines():
    var newBeams = newSeq[int](beams.len)
    for pos, cnt in beams:
      if cnt == 0: continue
      if line[pos] == '^':
        newBeams[pos-1] += cnt
        newBeams[pos+1] += cnt
        inc result.part1
      else:
        newbeams[pos] += cnt
    beams = newBeams
  result.part2 = beams.sum()

Update: found even smaller and faster version that only needs a single array.
Update #2: small optimization

type
  AOCSolution[T,U] = tuple[part1: T, part2: U]

proc solve(input: string): AOCSolution[int, int] =
  var beams = newSeq[int](input.find '\n')
  beams[input.find 'S'] = 1

  for line in input.splitLines():
    for pos, c in line:
      if c == '^' and beams[pos] > 0:
        inc result.part1
        beams[pos-1] += beams[pos]
        beams[pos+1] += beams[pos]
        beams[pos] = 0
  result.part2 = beams.sum()

Full solution at Codeberg: solution.nim

Rust

Part 1 was quite straightforward: just keep track of a frontier of every beam, and keep following it, adding splitter locations to a set. Easy.

I had to give some thought to Part 2 and ended up doing dynamic programming: for every position, keep track of how many timelines go through it, solving recursively where necesary.

use std::collections::{HashMap, HashSet, VecDeque};

use crate::{grid::{Coordinate, Direction, Grid}, solver::DaySolver};

pub struct Day07Solver;

impl DaySolver for Day07Solver {
    fn part1(&self, input: String) -> String {
        let grid = Grid::from_grid_string(
            &input,
            |c| match c {
                'S' => Location::Start,
                '^' => Location::Splitter,
                '.' => Location::Empty,
                _ => panic!("Invalid location type"),
            }
        );

        let mut reached_splitters: HashSet<Coordinate> = HashSet::new();

        let mut beam_coordinates: VecDeque<Coordinate> = grid.iter()
            .filter_map(|(c, l)| match l {
                Location::Start => Some(c),
                _ => None,
            })
            .collect();

        while let Some(next) = beam_coordinates.pop_front() {
            if let Some(next_coord) = grid.neighbor_in_direction(next, Direction::Down) {
                match grid[next_coord] {
                    Location::Start => { panic!("Encountered a second start!"); },
                    Location::Splitter => {
                        if !reached_splitters.contains(&next_coord) {
                            reached_splitters.insert(next_coord);
                            [
                                grid.neighbor_in_direction(next_coord, Direction::Left),
                                grid.neighbor_in_direction(next_coord, Direction::Right),
                            ]
                                .into_iter()
                                .flatten()
                                .for_each(|c| beam_coordinates.push_back(c));
                        }
                    },
                    Location::Empty => { beam_coordinates.push_back(next_coord); }
                }
            }
        }

        reached_splitters
            .len()
            .to_string()
    }

    fn part2(&self, input: String) -> String {
        let grid = Grid::from_grid_string(
            &input,
            |c| match c {
                'S' => Location::Start,
                '^' => Location::Splitter,
                '.' => Location::Empty,
                _ => panic!("Invalid location type"),
            }
        );

        let start_coord = grid.iter().find(|(_, l)| **l == Location::Start).unwrap().0;

        let mut num_timelines: HashMap<Coordinate, usize> = HashMap::new();
        count_timelines(&mut num_timelines, &grid, 1, start_coord);

        num_timelines[&start_coord].to_string()
    }
}

fn count_timelines(
    num_timelines: &mut HashMap<Coordinate, usize>,
    grid: &Grid<Location>,
    timelines_so_far: usize,
    coordinate: Coordinate,
) {
    if num_timelines.contains_key(&coordinate) {
        return
    }

    match grid[coordinate] {
        Location::Splitter => {
            let neighbors: Vec<Coordinate> = [
                grid.neighbor_in_direction(coordinate, Direction::Left),
                grid.neighbor_in_direction(coordinate, Direction::Right),
            ]
                .into_iter()
                .flatten()
                .collect();

            neighbors.iter()
                .for_each(|next| {
                    count_timelines(num_timelines, grid, timelines_so_far, *next);
                });

            let timelines_from_here = neighbors.iter().map(|c| num_timelines[c]).sum();
            num_timelines.insert(coordinate, timelines_from_here);
        },
        _ => {
            let timelines_from_here = if let Some(next) = grid.neighbor_in_direction(coordinate, Direction::Down) {
                count_timelines(num_timelines, grid, timelines_so_far, next);
                num_timelines[&next]
            } else {
                timelines_so_far
            };

            num_timelines.insert(coordinate, timelines_from_here);
        },
    };
}

#[derive(PartialEq, Eq, Copy, Clone)]
enum Location {
    Start,
    Splitter,
    Empty,
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn part1() {
        let input = include_str!("../../inputs/test/07");

        let solver = Day07Solver {};
        assert_eq!("21", solver.part1(input.to_string()));
    }


    #[test]
    fn part2() {
        let input = include_str!("../../inputs/test/07");

        let solver = Day07Solver {};
        assert_eq!("40", solver.part2(input.to_string()));
    }
}

Go

Oh my scheduling God! Part 1 was easy. Then for part 2 I started tracing each particle one by one using goroutines, but spawning billions of goroutines seemed to make my poor laptop sad. So I implemented a whole thread pool with process management and stuff, but nothing worked. So at the end I started doing the unthinkable: using my brain.

It seems we can just reuse the same idea as part 1 one but with a clever counting scheme. Thing works and is as fast as part 1. I’m both happy and deeply sad not to have been able to leverage Go’s supposed killer features - which are actually rarely useful when programming things other than servers tbh.

Here we goooooo:

package main

import (
	"aoc/utils"
)

func parseStartLine(line string) []bool {
	runes := []rune(line)
	beams := make([]bool, len(runes))

	for idx, char := range runes {
		if char == 'S' {
			beams[idx] = true
		}
	}

	return beams
}

func stepOne(input chan string) (int, error) {
	beams := parseStartLine(<-input)
	splitCount := 0

	for line := range input {
		runes := []rune(line)
		for idx, char := range runes {
			if char == '^' && beams[idx] {
				splitCount++
				if idx > 0 {
					beams[idx-1] = true
				}
				if idx < len(runes)-1 {
					beams[idx+1] = true
				}
				beams[idx] = false
			}
		}
	}

	return splitCount, nil
}

func valueBeams(beams []bool) []int {
	valBeams := make([]int, len(beams))

	for idx, b := range beams {
		val := 0
		if b {
			val = 1
		}
		valBeams[idx] = val
	}

	return valBeams
}

func stepTwo(input chan string) (int, error) {
	beams := valueBeams(parseStartLine(<-input))

	for line := range input {
		runes := []rune(line)
		for idx, char := range runes {
			bc := beams[idx]
			if char == '^' && bc > 0 {
				beams[idx] = 0
				if idx > 0 {
					beams[idx-1] += bc
				}
				if idx < len(runes)-1 {
					beams[idx+1] += bc
				}
			}
		}
	}

	sum := 0
	for _, bc := range beams {
		sum += bc
	}

	return sum, nil
}

func main() {
	inputFile := utils.FilePath("day07.txt")
	utils.RunStep(utils.ONE, inputFile, stepOne)
	utils.RunStep(utils.TWO, inputFile, stepTwo)
}

C

Accidentally solved part 2 first but had the foresight to save the code. Otherwise my solution looks similar to what other people are doing, just with more code 😅

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <assert.h>

#define GW 148
#define GH 74

static char g[GH][GW];
static uint64_t dp[2][GW];
static int w,h;

/* recursively traces beam for part 1, marking visited splitters with 'X' */
static uint64_t
solve_p1(int x, int y)
{
	if (x<0 || x>=w || y<0 || y>=h || g[y][x] == 'X')
		return 0;
	else if (g[y][x] != '^')
		return solve_p1(x, y+1);
	else {
		g[y][x] = 'X';
		return 1 + solve_p1(x-1, y) + solve_p1(x+1, y);
	}
}

/* DP for part 2 */
static uint64_t
solve_p2(void)
{
	int x,y, odd;

	for (y=h-1; y>=0; y--)
	for (x=1; x<w-1; x++) {
		/* only two rows relevant at a time, so don't store any more */
		odd = y&1;

		if (g[y][x] == 'S') {
			printf("\n");
			return dp[!odd][x] + 1;
		}

		dp[odd][x] = g[y][x] == '^' || g[y][x] == 'X'
		    ? dp[!odd][x-1] + dp[!odd][x+1] + 1
		    : dp[!odd][x];
	}

	return 0;
}

int
main()
{
	int x,y, sx,sy;

	for (h=0; ; ) {
		/* one bottom row of padding */
		assert(h < GH-1);
		/* input already side padded, plus we have \n\0 */
		if (!fgets(g[h], GW, stdin))
			break;
		/* skip empty rows */
		for (x=0; g[h][x]; x++)
			if (g[h][x] == 'S' || g[h][x] == '^')
				{ h++; break; }
	}

	w = strlen(g[0])-1; /* strip \n */

	for (y=0; y<h; y++)
	for (x=0; x<w; x++)
		if (g[y][x] == 'S')
			{ sx=x; sy=y; break; }

	printf("07: %"PRIu64" %"PRIu64"\n", solve_p1(sx,sy), solve_p2());
}

My choice of representation for part 1 ended up being fortunate, as all I needed to change for part 2 was three setf in propagate that became incf. This was also another good use case for multiple return values, as it’s the beam vector you want to reduce on, but for part 1 the split count needs to come along for the ride.

(defun parse-line (line)
  (let ((result (make-array (length line) :initial-element 0)))
    (loop for i from 0 to (1- (length line))
          if (member (char line i) '(#\^ #\S))
            do (setf (aref result i) 1))
    result))

(defun read-inputs (filename)
  (let ((input-lines (uiop:read-file-lines filename)))
    (mapcar #'parse-line input-lines)))

(defun propagate (in-beams splitters)
  (let ((out-beams (make-array (length in-beams) :initial-element 0))
        (splits 0))
    (loop for i from 0 to (1- (length in-beams))
          if (not (zerop (aref in-beams i)))
            do (if (not (zerop (aref splitters i)))
                   (progn (incf (aref out-beams (1- i)) (aref in-beams i))
                          (incf (aref out-beams (1+ i)) (aref in-beams i))
                          (incf splits))
                   (incf (aref out-beams i) (aref in-beams i))))
    (values out-beams splits)))

(defun propagate-all (initial-beam-vector splitter-map)
  (let* ((total-splits 0)
         (final-beam-vector
           (reduce #'(lambda (in-beams splitters)
                       (multiple-value-bind (out-beams this-splits)
                           (propagate in-beams splitters)
                         (incf total-splits this-splits)
                         out-beams))
                   splitter-map :initial-value initial-beam-vector)))
    (values final-beam-vector total-splits)))

(defun main-1 (filename)
  (let ((grid (read-inputs filename)))
    (multiple-value-bind (final-beam-vector total-splits)
        (propagate-all (car grid) (cdr grid))
      total-splits)))

(defun main-2 (filename)
  (let ((grid (read-inputs filename)))
    (multiple-value-bind (final-beam-vector total-splits)
        (propagate-all (car grid) (cdr grid))
      (reduce #'+ (coerce final-beam-vector 'list)))))

Haskell

import Control.Arrow
import Control.Monad
import Control.Monad.Writer.Strict
import Data.Array
import Data.Functor
import Data.List
import Data.Maybe
import Data.Monoid

parse content = (elemIndices 'S' x, filter (not . null) $ elemIndices '^' <$> xs)
  where
    (x : xs) = lines content

split :: [Int] -> Int -> Writer (Sum Int) [Int]
split splitters beam
    | beam `elem` splitters = tell 1 $> [pred beam, succ beam]
    | otherwise = pure [beam]

part1 = getSum . execWriter . uncurry process
  where
    process start =
        foldl'
            (\beams splitters -> nub . concat <$> (beams >>= mapM (split splitters)))
            (pure start)

part2 :: ([Int], [[Int]]) -> Int
part2 (start, splitterList) = go (head start, 0)
  where
    go (i, j)
        | j >= depth = 1
        | hasSplitter i j = r ! (pred i, succ j) + r ! (succ i, succ j)
        | otherwise = r ! (i, succ j)

    r = listArray bounds [go (i, j) | (i, j) <- range bounds]
    bounds = ((0, 0), (width, depth))

    hasSplitter i j = j < length splitterList && i `elem` splitterList !! j

    depth = length splitterList
    width = succ . maximum $ concat splitterList

main = getContents >>= print . (part1 &&& part2) . parse