AOC 2025

https://adventofcode.com/2025/

Overview

I've been coding for over 10 years, but all in high-level languages. I am interested in learning a lower-level language. I'm not quite sure which one to pick. I figured Rust was a good choice as it is quite popular in the Python ecosystem, so I can probably find some use for it. I've also dabbled in Rust previously.

So far, I have solved 5/12 of the Advent of Code 2025 puzzles in Rust.

Stipulations:

• No vibe-coding, no cheating on solutions algorithmically.
• Rust docs and StackOverflow are allowed. Ignore Gemini Google search results.
• Claude can assist with compiler errors only or very nuanced things specific to Rust. Every question to AI must be meant with the intent to learn something about Rust and not to take a shortcut.
• Use minimal 3rd party libs. So far, only Clap and Clippy.
• No copy + pasting across modules. Looking at another module is perfectly fine, but all repeat code must be re-typed out by hand. Otherwise, no muscle memory is built.

The code here is almost certainly sloppy. I'm interested in keeping this as a time capsule where I can one day look back and see my progress.

Learnings

• Day 1: This day took the longest and involved the most learning! First, it took a while to get things running the way I want. I wanted a simple Clap-based setup without any frills. Most code examples show clap::Parser with a struct and not an enum. Second, I had to brush up on a lot of Rust things: impls, Results, match, iterators, and so on. You can see very much a "path of resistance" in this code, with plenty of questionable things like map_err and not using Self when I should have. I also relied a lot on Claude Code for this first day just to get everything compiling.
• Day 2: More impl, and also how to cast strings to numbers. (This is the last time I used Claude Code until day 5.)
• Day 3: Casting, constants, usize.
• Day 4: saturating_sub().
• Day 5: I got extremely ambitious for this one. I got tired of just writing basic code and wanted to see how far I could push things. I learned: by_ref() for an iterator, std::cmp::Ordering + some other traits, and a few more things. I did have to ask Claude for a hint when I tried to use Vec::binary_search_by and it pointed me to Vec::partition_point. I also had a logical bug that drove me crazy and which I got stuck on that I just couldn't find, and Claude helped out (I was setting j to 0 instead of 1-- oops!).
• Day 6: For the previous 5 problems, I had written fairly procedural code. Today I wanted to experiment with writing much more functional code. Sadly, there is still a for-loop in part 2, but part 1 is entirely for-loop free, and even part 2 is mostly functional. Uses of AI: I ran into a couple type issues I had to pass by Claude relating to types (some things about str versus String, and trying to slice an iterator); and after-the-fact, I had Claude re-write transpose_matrix to be more functional (I'd used a procedural implementation for this). Although having AI rewrite code goes against the spirit of what I am trying to do, I really just wanted to see the functional implementation!
• Day 7: Did this on the Amtrak. I wanted to continue my functional programming streak. I did OK, but still ultimately relied on a mutable variable in an outer scope for part 1. I learned about Iterator::fold(), which was exactly what I needed, and I learned about if conditions in match statements. No Claude.