My original pure-functional design of Intcode definitely payed off today! Doing a breadth-first search (which I used to solve both parts) was straightforward. There is some duplication in the solution for each part, and I spend short time figuring out if I could make a more generic BFS abstraction, but in the end I didn’t feel it’s worth it. I might explore this in the future. Anyway, here’s my solution.

I just had a look at everyone’s solution and I’m surprised to notice that (for once) my solution is among the shortest (30 lines for each part).

Nothing fancy in my code though, it was a pretty straightforward tree traversal

I changed my Intcode module back to be fully functional (not using processes or message passing).

Here is my solution.

Slightly overcomplicated it by using Dijkstra for part 1, which is not required due to all edges being the same weight:
https://git.sr.ht/~ihabunek/aoc2019/tree/master/lib/day15.ex

I animated the result again:
https://asciinema.org/a/80grPw4uGhl2GTBQjz33hSKip

Since I was short on time, I just went roomba style for part 1 (random movement, preffering unexplored areas to old areas) and drew the map, then I calculated part 1 and 2 by hand (not proud at all).

I will try to do this in a proper way later.
“solution”, part 2 is broken since I tried to explore correctly, by I’m out of time