Nice article. I've been saying this from intuition. Quality is more valuable than speed. It saves more in the long run. The only time you want speed over quality is in temporary solutions, in the order of weeks or months. Article is basically saying this.

> yet delivery stability dropped about 7%. More code arrived, and more of it broke. A metric that stops at the keystroke cannot see that cost.

AI value metrics are short-sighted. You really want quality code that never breaks as much as possible. You don't want to build a skyscraper with mud bricks, and most of the industry builds software that benefits from working well for many years.