Mike Swanson's latest post explores the concept of backseat software, describing how modern applications increasingly act as intrusive channels rather than tools. He draws a parallel between annoying in-car tutorials and the constant interruptions and tutorials found in many apps, arguing that this behavior is a result of software evolving from something users operate to something that operates on users. This shift has normalized an interruption model that would be unthinkable in other contexts, such as driving.

Swanson traces the evolution of software from physical media, where updates were infrequent and user feedback was direct but costly, to the always-connected era of the internet. The introduction of the back channel allowed software to send crash reports and receive updates, initially improving quality and safety. However, this capability expanded to include user analytics, leading to a focus on metrics like DAU and MAU, which can often mislead product teams into optimizing for behavior rather than usability.

The post delves into the dangers of relying on analytics, where metrics can become targets and distort product decisions. Swanson cites an example from his time at Microsoft, where a team wanted to remove a feature based on low usage, only to realize that the feature had been buried in the UI over time. This illustrates how metrics can be correct but misleading, leading to poor product decisions. Swanson argues that experimentation in production, such as A/B testing, further blurs the line between product and laboratory, making it difficult to maintain a clear product vision.