Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Now, with this realization, assess the narrative that every AI company is pushing down our throat and tell me how in the world we got here. The reckoning can’t come soon enough.


What narrative? I'm too deep in it all to understand what narrative being pushed onto me?


No, wasn't directed at someone in particular. More of an impersonal "you". It was just a comment against the AI inevitabilism that has profoundly polluted the tech discourse.


We're all too deep! You could even say that we're fully immersed in the likely scenario. Fellow humans are gathered here and presently tackling a very pointed question, staring at a situation, and even zeroing in on a critical question. We're investigating a potential misfire.


I doubt there will be a reckoning.

Yes, the tools still have major issues. Yet, they have become more and more usable and a very valuable tool for me.

Do you remember when we all used Google and StackOverflow? Nowadays most of the answers can be found immediately using AI.

As for agentic AI, it's quite useful. Want to find something in the code base, understand how something works? A decent explanation might only be one short query away. Just let the AI do the initial searching and analysis, it's essentially free.

I'm also impressed with the code generation - I've had Gemini 3 Pro in Antigravity generate great looking React UI, sometimes even better than what I would have come up with. It also generated a Python backend and the API between the two.

Sometimes it tries to do weird stuff, and we definitely saw in this post that the command execution needs to be on manual instead of automatic. I also in particular have an issue with Antigravity corrupting files when trying to use the "replace in file" tool. Usually it manages to recover from that on its own.


AI pulls its answers from stack overflow.

What will happen when SO is gone? When the problems go beyond the corpus the AI was trained on?


Which is weird because SO is trash and has been a long time… every top few answers might as well be skipped, and you’ll find the correct answer to the 3rd comment half way down the page

AI does the exact same thing in my experience. First try is not right. Nor the second. If you're lucky the third might.

At that point I'd rather look at SO


I imagine we will document the solution somewhere, preferably indexable for AI's search, so that it will be available before the next model is trained on the latest data.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: