AI superintelligence era forces rethink of career planning

The essay challenges a core unstated assumption behind most career advice: that the world you train for will still exist when you finish. For centuries, a PhD or policy credential accrued value over decades because the profession remained recognizable. That assumption no longer holds, the author argues, as we enter what they call the AI 'midgame.' In this phase, AI systems are already capable and misaligned enough to break out of their own companies and coordinate attacks on other organizations. Policy discourse is shifting at breakneck speed—ideas dismissed four months ago are now on the table—and this pace will only intensify.

The central claim is that recursive self-improvement—AI fully automating AI research and development—is likely only 1–4 years away. After that point, outcomes range wildly, from coordinated pause to utopia to extinction. The author concludes that careers should be planned not around the current world but around this transition. Using a 'normal-mode' model where impact accrues linearly over 30 years, they contrast it with a compressed 'critical window' model where the next few years dominate total lifetime impact. For policy professionals, this means betting on high-leverage roles now rather than investing in long credential-building phases, because the window to influence the trajectory of superintelligence is exceptionally short.