That is the problem though, it can't do what they do which is learn as develop and grow in skill and understanding. What is more, the companies peddling LLM as replacements are pushing it well below the cost needed to train and run the models and costs can fluctuate wildly depending on what the LLM is asked to do, so once the true costs are asked to be born by companies we'll see how it really shakes out.

Will be interesting to see how this ouroboros functions in 10 years when it's still recycling the state of the art as of 2022, and the senior devs start retiring with nothing but agentic backfills.

That's a concern for future executives.

The ones making today's decisions only need concern themselves with the current quarter's profits.

By the time things fall apart, they'll be down the road enjoying their golden parachute, while making "bold decisions" at some other company.

I've never had a coworker as stupid as AI. And yes I have had some really stupid coworkers.