[gonzalohm]

It's about the user of that information.

As a programmer, I'm comfortable judging the coding output of an LLM. But now, anyone can go and start building without any knowledge, and at first it may look fine, but you are creating software using a pretty weak foundation, bad maintainability, etc.

I think LLMs allow everyone to skip an important step of building anything, which is understanding how things work.

[clapthewind]

Why is that (framed as) a bad thing? We don't go around life understanding everything that we do because it would be exhausting.

Sometimes all you want to do is use something and it works.

[gonzalohm]

It depends on the application. A low stakes vibe coded personal project? Sure go ahead, that's a perfect use. A proof of concept? Also pretty good use.

For the main product of your company I think it's fair to say that people should know and understand what they are doing

[ragequittah]

You can understand deeply what you're doing and still make a huge amount of mistakes. This is what we're ironically (in terms of this conversation) seeing AI point out over and over now that companies have mythos in their hands. If I had to bet on it at some point we'll see the shift where if AI hasn't fully reviewed and edited code it's considered unsafe to use. Similar to how we all understand how once self-driving cars are good enough human driving should probably be minimized because we're so bad at it.