Instead of hints, try asking for, or finding, test cases. Several of us posted additional test cases that found most people's problems in part 2. Once you have a test case (with a small input size, so you can easily step through it by hand) you can usually identify the particular problems in your solution. This will also help you understand why it works.

If that still doesn't work, then find working solutions and try to understand them. Some people's solutions get really clever, don't worry about those. If it's not even remotely clear to you what's happening in it, it's not a solution you need to study yet. Find the simplest, brute force solutions first. Then find ones that look similar but are optimized in some way.

I find that to be the best balance between challenge and learning something new. You will mentally burn yourself out if you keep bashing against the wall for hours or more, not quite a healthy thing to do :)

Meanwhile, people who actually try to compete on this stuff have already developed rich library of specialized algorithms to leap ahead of average programmer. Well, I guess nowadays a lot of it is LLM assisted too.

Certainly using nushell means anything beyond the true basics seems to be beyond most LLMs.

When I first started programming it used to get me down if I couldn't solve these things. But as I've got old and a little more experienced, I can now admit things like AoC are just not my thing. It's like crossword puzzles or low level algos. I find them extremely hard to reason about.

Just copying someone else's solution, or getting an LLM to fix it for you will be very low struggle, so you won't learn much.

To add some struggle, maybe look up a solution in a different language and translate to your language? You could choose a solution in a language similar to your language, so if you are solving in C, perhaps look up a C# solution or to make it harder look up a solution in a different paradigm. Find a Haskell solution or a Prolog one and see if that gives you enough hints.

I'm using nushell this year for AoC. It's functional by nature though you can make it imperative, so it's out of my comfort zone.

However the problem is math based whatever language you're using, involving modulo and int division. I had an hunch it was about that but no sense of what to do or how to approach.

Having looked at multiple ways of doing it I still have no clue what's going on, only that it works.

You would probably learn more vibe coding a solution than just reading code.

Because your engagement would be active, you would practice useful technique, and have an opportunity to iterate.

Good luck.

I think there's something to this except if you're using a language the llm doesn't know, like I am with nushell.

Day 2 for example I easily completed on my own with 2 very similar regex strings. However knowing regex to be slow I went looking for a more optimised solution. I found some python code that I mostly understood and attempted to rewrite the iterative solution into the functional nushell style and failed miserably. I turned to an llm for help only for it to waste my time with nonsense because nushell appears to be too new and changing too quickly for it.

This is exactly what I was getting at.

You will learn more from trying to vibe nushell code than reading solutions.

From thinking about how nushell is like other languages that can be vibe coded to considering why you want to learn nushell.

Or to put it another way, the dog did not actually eat your homework.