Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)

Rabois argues that the future PM role looks more like a CEO's job: deciding what to build and why. AI is already changing careers, and in strong organizations the CMO can be the largest token consumer because agents let executives produce work directly instead of relying on layers of deputies. His contrarian views on customer conversations and public criticism are part of a broader model that values direct judgment and high performance.

The foundation is talent density. Rabois repeats Vinod Khosla's principle that the team you build is the company you build, and points to PayPal's later company-building record as the result of Peter Thiel and Max Levchin assembling unusually strong people. His own early weakness was identifying strangers through short interviews, so he first learned to recognize talented colleagues with context, then built methods for assessing people he did not already know.

Hiring is treated as a trainable but imperfect muscle. A founder should repeatedly compare hiring decisions with later performance, learn what was missed, and improve. Rabois says ordinary tactics can move someone within the middle of the talent distribution, but exceptional judgment requires departing from the playbook. References are one teachable method: DoorDash's Tony conducts twenty references for a senior hire, while David Sze taught that the process is not exhausted until it reaches a negative reference.

The key is asking the question that matches the decision. A reference asking whether Max Rhodes was a good employee produced mixed answers and made investors hesitate about Faire. Asking whether he could become a world-class entrepreneur produced the more relevant answer. The same person can therefore appear unsuitable or exceptional depending on whether the interviewer is measuring the right capability.