The review bottleneck weighs more than the calls to slow down

On the heads of major AI developers calling one after another for development to slow down, a conversation puts forward the reading that those remarks are closer to blame-dodging. The grounds are governance: a founder holding a real ownership stake could simply scale back their own company first instead of raising it publicly, and there is no sign that the pace actually changed after the remarks. What is singled out as the real cause is a structural shift: as the ones making new products changed from people to models, release cycles overtook how fast people can review, and "out of control" is redefined to mean not that it acts on its own but that there is no time to check whether things were done right. This is followed by the paradox that hiring more people to review the delegated work cancels out the reason for adopting it, the claim that something generated in 10 minutes takes 5–10 hours to review, and the scale argument that doing the work of 500 people also means making 500 mistakes at once. However, most of the key events and figures are secondhand or estimates, so it is hard to use as citable material.

This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.