Annie Vella

Everyone agrees that the software development lifecycle needs reimagining because of AI, yet what we generally see are pictures of the traditional process with little robot faces sprinkled on it. The truth is it's hard to reimagine something from within it - so the lifecycle kept its shape, and the engineer's job absorbed the difference.

Writing the code got dramatically faster and everything downstream is struggling to keep up. My Masters research found engineers feel they're spending less time on nearly every engineering activity, except for reviewing code. What replaced the writing is a job most of us didn't sign up for: directing, evaluating and correcting code we didn't write, while staying accountable for every line of it. I call this supervisory engineering work.

Other industries have been supervising fast automated systems for decades, and their conclusion isn't particularly encouraging - the human monitor, as one 1983 paper put it, has been given an impossible task. But that same paper ends on a brief rather than a warning: fixing this takes more engineering, not more vigilance.

This is a talk about the job we're actually doing now, why it feels the way it does, and why the next version needs designing on purpose - and far more boldly than we've managed so far.