Want to actually see it deliver tangible results that are of sufficient quality to merge to main? Oops! Look at that, we're out of time. The actual examples of BS-generated code I've seen are routinely horribly flawed and the implementer often can't even explain what they've done.
Of course, bullshit pedalers always fall back on "welp, ultimately the responsibility is on the engineer to understand and audit the output," but this weirdly never seems to happen...
It's rare to hear about organizations getting much out of it and the biggest bullshitters in the market are pushing it the hardest. For me that says it all.
The one minimal countervailing force against my dismissal of the BS is the handful of colleagues and friends who sincerely assert they find value in it. But it often reads like they're backfilling their opinion about its benefit to them from the hype spewed by bullshitters.
Their claims about boosts they personally see are modest, but they will talk up much grander claims from others. "So-and-so announced they produced 50 production-grade applications in a month and were generating huge revenue!" Did so-and-so substantiate this? "Well, I didn't look deeply into it."