> At a major semiconductor manufacturer, agents reduced chip optimization work from six weeks to one day.

I call BS right there. If you can actually do that, you’d spin up a “chip optimization” consultancy and pocket the massive efficiency gain, not sell model access at a couple bucks per million-tokens.

There should be a massive “caveats and terms apply” on that quote.

So far the AI productivity gains have been all bark and no bite. I’ll believe when I see either faster product development, higher quality or lower prices (which indeed happened with other technological breakthroughs, whether the printing press or the loom) - if anything, software quality is going down suggesting we aren’t there yet.