What's this about?
Mollick's mental model for Fable- and Astra-class models: you're delegating to a *good outside team*, not an intern. The skills that matter: can you specify what you want, how much leeway they have, what to test, and when to come back to you?
What exactly changed from the intern model?
Interns need supervision on execution; teams need clear briefs and check-in structure. The bottleneck moves from reviewing every step to specifying intent well — management skill, not prompting skill.
Why is this worth noting?
It reframes AI adoption as a management-capability problem, which matches this week's evidence (five-day unattended runs, 20B-token reviews). The people who delegate well to humans may extract the most from these models.