...and I still don't get it. I paid for a month of Pro to try it out, and it is consistently and confidently producing subtly broken junk. I had tried doing this before in the past, but gave up because it didn't work well. I thought that maybe this time it would be far along enough to be useful.
The task was relatively simple, and it involved doing some 3d math. The solutions it generated were almost write every time, but critically broken in subtle ways, and any attempt to fix the problems would either introduce new bugs, or regress with old bugs.
I spent nearly the whole day yesterday going back and forth with it, and felt like I was in a mental fog. It wasn't until I had a full night's sleep and reviewed the chat log this morning until I realized how much I was going in circles. I tried prompting a bit more today, but stopped when it kept doing the same crap.
The worst part of this is that, through out all of this, Claude was confidently responding. When I said there was a bug, it would "fix" the bug, and provide a confident explanation of what was wrong... Except it was clearly bullshit because it didn't work.
I still want to keep an open mind. Is anyone having success with these tools? Is there a special way to prompt it? Would I get better results during certain hours of the day?
For reference, I used Opus 4.6 Extended.
Your parent has no idea what they're talking about. This is what happens when AI companies are given more money than they know what to do with; their marketing gets more and more unhinged, and eventually leaks out to the laymen.
I guess what might help you convince them is some perspective. It's true that LLMs can generate code, but what a laymen doesn't see is that the code they generate is in a language built by humans, using frameworks and tools built by humans, running on platforms built by humans, and distributed through infrastructure built by humans. Additionally, LLMs themselves are built by humans.
Could all of that be AI generated too? No. Anthropic already tried to build a browser and a compiler with Claude recently, and both projects failed spectacularly. The transformer model powering current LLMs has hit a wall; advancements have slowed down significantly in recent years, and the only thing that will get things going is another breakthrough. You're better off gambling your tuition at a casino than betting we'll have a breakthrough within 7 years (or worse, predicting the impact it will have on job markets)
And of course, the idea that an administrative role is safe is silly. There are already experiments to have LLMs run physical businesses on their own. If I had to make a prediction, I'd say business roles are much easier for an LLM to completely replace than engineering.
And finally, the bubble. LLMs today are affordable because of a frenzied market of degenerate gamblers; when the bubble pops (which will happen within 7 years), these AI companies will be forced to restructure or die. Anthropic charges $200/mo for their top models, and that has rate limits that people hit regularly. If Anthropic can't afford unmetered access at $200/mo while they're swimming in cash, then there's no way in hell those prices don't explode post-bubble. I'm not saying AI companies will die, but you and I aren't going to be able to use them.
Like @Newsteinleo@infosec.pub said, don't let them decide this for you. If they won't help pay your tuition, then get your own student loans, and go to a public university to save money.