I think the truth is somewhere in the middle, but I could be wrong. I work in a non-programming technical field and I am also impressed and concerned by the progress the models have made over the last few years, especially in terms of being able to "do my job".

Like the other guy was saying, there are still issues with its implementations, and the second you begin to think it's perfect or immutable, you notice issues that would have been show stoppers if they made it through. And at the end of the day, at least for what I do, if nobody is here to press the buttons, the AI will never be able to do its thing to begin with.

I also personally predict that the cost of tokens and utilizing AI services will eventually become overly expensive and encumbered, as does everything under capitalism. Name one thing that hasn't gotten shittier and more expensive overall. The AI companies haven't even begun to enter their enshittification stage yet.

I maintain that AI in its current form will NOT be able to produce truly novel thought or original work, as it can only put together the Legos that it already has ingested. Perhaps grunt work programming will require less manpower, but cutting-edge developments are still beyond the LLM's grasp at the moment.

Anyways, I welcome other opinions

I implied you were neither of those things - I just quite openly implied you're a shill. Perhaps you should ask an AI for tips on improving your reading comprehension?

I know there's a temptation to call "skill issue" when you see anecdotal evidence that runs counter to your own experience, but I can say with some degree of professional certainty and a smidge of personal snark that no, I in fact do know what I'm doing. These tools have their uses, but their general applicability is vastly overstated (if I were the kind of cynical person that enjoyed being right, I would say intentionally overstated in an effort to attain profitability on bootstraps made of hype). And I wanna clarify that I do indeed use some AI tools (ex: security checkers) with regularity. They do absolutely have worth - they just don't have trillions of dollars worth... of worth. Ahem.

The hype-doomer thing that some people have about this, oh woe it's going to destroy an entire conceptual industry, that just isn't what's happening, but it's very much what the advertising would like you to believe it's doing.

I have both used agents and have been the "victim" of heavy agent users.

If you can provide an utterly well bounded problem with perfectly verifiable criteria for it to retry against until the tests pass and the tests are full and valid for the use case, it can work. Similarly, if your code's reality is forgiving and flexible, and a 'close enough' result is good enough to get the human software user in the right ball park, you might be able to extract decent behavior.

However, if there is a means by which the answer can 'look correct' yet be wrong in the real world and the real world scenarios require accuracy and precision, there's huge gaps.

Especially if the agents control the coding and the test case generation, seen plenty of times where it talked itself out of a test case that was failing when the test case was in fact correctly showing a flaw.