Depends if you're really stuck on a problem. I'd rather learn by getting a look at how it's done, even if the code is incorrect, than be completely suck with no idea why or what to do.
This is his entire point: getting to this brick wall is _where_ the real learning happens. When you start scrambling, reading old stackoverflow posts, and breaking out the calculus text book you're pushing the boundaries of what is possible for you.
Body builders don't grow by watching other people lift weights.
That isn't an accurate comparison. Asking an AI is functionally equivalent to searching the web. Your comparison is like saying you can't get strong by filling up the buckets you have handy with water and lifting them, you have to do what real bodybuilders do and lift weights. Whether I'm researching the answer using a search engine or an AI prompt, the result is the same.
How do they become pioneers? By digging into the body of knowledge. They know what books and authors are good, what blogs to read, they posted questions on SO. They don't search for direct answers, but instead they search for the knowledge that allowed the person who made the question to make the question, and person who found solutions to find the solution.
Looking how it's done doesn't equal you doing it. It's apples to oranges. OP was referring to needing to do work in the gym to get results. You didn't get results if you looked at others and did the same in the advent of code etc. You are merely entertaining yourself, but you are not competitive participant. With that approach you'll never be on top.
There are people who use frameworks and people who write frameworks. The first group struggles when there is no official documentation or tutorial on a topic they need. The second group when faced with the same goes through the source code and gets the answers.
Bottom line is the question of what you want to achieve - just go through the problems and be done with them, or deep dive into new topics ans material, to widen your knowledge?
I don't see much to disgree with in what you're saying. Were clearly talking about different things.
What happens when you dig in and exhaust all your best efforts and still don't find a solution? Are you a lost cause because you couldn't figure it out for yourself? Or are you allowed to be shown the way it's done so you can learn from that?
Many times in my life have I not been able to figure something out for myself, and needed shown how the thing was done. Just because I couldn't do it on my own doesn't mean I wasn't capable of either doing it once shown, or that by being shown I didn't learn anything.
I think the difference between the textbooks and AI is that the AI can answer exactly your question without providing the necessary context/reasoning behind how it arrived there.
When you are drawing connections between SO posts, textbooks, whatever you do a fair amount of reasoning yourself.
Now that I think about it it isn't AI specific (more knowledge specific) - I'd say the same if you wait a few hours and then look up the AOC solutions on github. Sure you'll have "solved" the problem but you never spent the time to actually figure out the solution.
> And at the start, novice body builders will of course watch other body builders to learn how it's done.
"Learn how it's done" - not get bigger. This is the important difference. If those novice body builders watched others and learned how its done without putting in the reps themselves they'd get nowhere.
Say I have no idea how to complete day 12 of AoC. Racked my brains, done some searching, researched as best I can, can't find anything that makes sense. I'm stuck.
By your logic I would gain absolutely nothing, learn nothing, but either looking up the answer or getting AI to give me the answer. I don't accept that.
Reason being? I've been in the exact same situation. Guess what, I learned different approaches to the problem I could not solve. I improved and can now solve similar problems as a result. Is it cheating? Maybe. Do I care? No. Why? Because I'm not doing AoC for the struggle, I'm doing it for fun, and I know my limits too.
I think in that case you can only guarantee that the code you're looking at isn't broken in some ways neither I or the author on Reddit know enough to spot.
If Claude generates it and it produces correct output for the challenge but is subtly broken on some edge case not in the challenge input, how is it difference from pulling some code from Reddit that produces correct output for the challenge but is subtly broken on some edge case not in the challenge input?