The Geek Way

In the [YOU] on AI Field Guide

The geek way connects to the cycle that began with [YOU] on AI on two levels. On the practical level, it describes the organizational culture most likely to deploy AI effectively: an organization that settles questions about AI by experiment rather than executive intuition, gives its people autonomy to find new uses, iterates quickly, and stays open to surprising results is positioned to harness the technology. An organization that approaches AI through its existing hierarchy, demanding that every application be planned and approved from the top, will move too slowly and learn too little to keep pace.

On the deeper level, McAfee’s emphasis on humanity’s superpower carries an argument about what remains distinctively human in the age of intelligent machines. The capacities he identifies—intense cooperation and rapid social learning—are collective and cultural rather than individual and computational. An AI system can learn from data, but it does not cooperate in the human sense, does not participate in the dense web of mutual influence and shared purpose through which human groups generate and refine knowledge. McAfee’s analysis suggests that even as machines absorb more individual cognitive tasks, the distinctively human achievement of organized collective intelligence remains, for now, ours. The geek way is a method for cultivating that achievement, and in an age when individual cognition is increasingly shared with machines, the human capacity to learn and cooperate together may be the complement to the machine that McAfee has urged all along to develop.

The geek way also speaks directly to the cycle’s portrait of the Trivandrum training session: a week spent working alongside twenty engineers, not simply handing them a new tool but developing practices that redirected the technology toward their development as well as their productivity. The four norms were present in that room—disputes settled by what the system actually produced rather than by seniority; genuine ownership of the redefined work; rapid iteration; and openness to the uncomfortable finding that many of the engineers’ prior tasks could be absorbed while their architectural judgment became more valuable than ever.

Origin

McAfee developed the concept in his 2023 book The Geek Way, drawing on years of studying the organizations most successfully adapted to digital technology. The observation that these organizations shared a distinctive culture was not new—Silicon Valley observers had noticed the informal, flat, fast-moving character of the most successful technology companies for decades. What McAfee added was an analytical account of why the norms work, grounded in his reading of evolutionary anthropology and the science of human cooperation.

The four norms emerged from examining what the most successful technology organizations actually do differently. Science means a commitment to evidence that can override authority: anyone who can muster better evidence can win the argument regardless of rank. Ownership means genuine devolution of autonomy and responsibility, so that the people doing the work own their outcomes rather than waiting for direction from above. Speed means preferring the fast experiment to the elaborate forecast, because in a rapidly changing environment the experiment yields real information while the forecast yields only the illusion of control. Openness means the willingness to be wrong—to let the hierarchy be corrected by those below it, to change course when evidence demands, and to let go of ideas that are not working. McAfee argues that when all four operate together, the result is a culture that is simultaneously fast-moving, egalitarian, evidence-driven, and genuinely adaptive.

Key Ideas

Humanity’s superpower. McAfee’s explanation for why the geek way works is grounded in a claim about what makes our species unusual. Human beings possess a unique combination: the ability to cooperate intensely in large groups and the ability to learn rapidly from one another. This pairing is, in his phrase, humanity’s superpower, the thing that sets our species apart and accounts for our extraordinary success. Science and openness accelerate learning by ensuring that good ideas win and bad ideas are abandoned quickly. Ownership and speed intensify cooperation and engagement by giving people real stakes and rapid feedback. The geek way is calibrated to amplify this superpower.

Science settles disputes. The first and most radical norm is the commitment to settling disagreements through evidence rather than authority. In a geek culture, the question is not who has the most authority but what the data show, and anyone who can produce better evidence can win the argument regardless of rank. This inverts the traditional corporate hierarchy and makes it genuinely uncomfortable for those accustomed to winning arguments by virtue of position. McAfee argues the discomfort is the point: a culture that can hear hard truths from below is a culture that learns.

Openness as organizational immune system. The fourth norm—openness, the willingness to be wrong—functions as the immune system of the geek organization. Without it, the other norms decay: science without openness becomes a search for confirming evidence, ownership without openness becomes entrenched fiefdoms, speed without openness becomes the rapid execution of wrong decisions. Openness is what allows the hierarchy to be corrected by those below it and what permits the organization to update rather than defend its existing commitments.

Implications for AI deployment. Organizations with geek cultures are best positioned to deploy AI effectively because the technology requires exactly the norms McAfee describes. A geek organization treats questions about AI use as empirical questions to be settled by experiment, distributes authority to try new AI applications to the people closest to the work, iterates rapidly on what works, and remains open to the finding that the AI changes the nature of the work itself. The alternative—a hierarchical organization that approaches AI through top-down planning, requiring approval before any new use is attempted—will move too slowly and learn too little to capture the technology’s value.