I feel two things simultaneously reading this.

On one hand, the original study seems faulty, on its face (a human who isn't involved in a project can't possibly accurately assess how much actual effort went into a particular commit - much less, an AI model.) And the media's interpretation of the results (10% of engineers do nothing all day) is obviously laughable.

On the other hand, despite the premise and interpretation being laughable... I feel like the conclusion kind of isn't. We know from our own experiences that many engineers at large companies are not producing much value with their time. 10% is an underestimate. And it's a problem I reflect on quite a lot.

But the reasons are a lot more complex than, they log in to work everyday and then go take a nap. No, rather most of the time an engineer is not producing much value because she is not being asked to produce much value, she's just being asked to look busy. In other words it's an organizational problem.

Per Parkinson's Law, you can take any engineering task previously done by an individual and expand it into work for an entire team. (Take that API and turn it into a service, take that service and make it into a bunch of microservices, take those microservices and deploy them on Kubernetes, take that Kubernetes cluster and automate its maintenance and deployments and upgrades, and now take those maintenance processes and wrap them in an internal service...) And at every step of the way, you can easily justify to your boss, and your boss's boss, that the work is important. And some of it is! But it's a difficult job figuring out where to draw the line exactly, and most leaders have no incentive to do so (bigger org = bigger paycheck).