The Waymo effect: how AI is quietly making research less collaborative

TEXT START: On a recent trip to San Francisco I did the thing that every visitor to San Francisco now does: I summoned a car with no one in it.

The Dissection

The article identifies a real secondary effect of AI: frictionless individual productivity can dissolve the social encounters that generated dissent, apprenticeship, serendipity, and durable understanding. Its Waymo analogy is effective. The collaborator is not merely an inefficient service provider; inconvenient disagreement is part of the intellectual production process.

But the article ultimately converts structural displacement into an institutional-design problem. It argues that funders can preserve collaboration by subsidizing travel, co-location, workshops, and deliberate human friction. That is a request to maintain a costlier production mode after AI has made much of its labor economically optional.

The Core Fallacy

The text mistakes a symptom of automation for the central mechanism of automation.

AI is not merely removing conversation from research. It is severing the connection between human participation and economically necessary output. Under the Discontinuity Thesis, once AI becomes cheaper and superior across cognitive work, human collaboration is no longer a stable economic domain that institutions can preserve at scale. It becomes a luxury, a status ritual, or a selectively maintained moat around AI capital.

The proposed remedy—engineer dissent back into the workflow—also assumes that human oversight will remain competent and authoritative. Bainbridge’s irony points in the opposite direction: automation degrades the very skill humans are supposedly retaining. A researcher who supervises machine-generated analysis may remain nominally in command while becoming substantively dependent on systems they can no longer reproduce or challenge.

The article sees the empty driver’s seat. It does not fully see the wage relation that used to make the driver necessary.

Hidden Assumptions

• Institutions can collectively resist the productivity race when competitors using AI cannot.
• Funding can make human collaboration economically rational rather than merely desirable.
• Human-generated novelty and dissent will continue to command enough market value to justify their cost.
• Researchers will retain enough expertise to function as genuine pilots after delegating core cognitive work.
• Metrics can be redesigned faster than competitive pressure forces them back toward speed and throughput.
• Preserving elite research culture meaningfully addresses the wider collapse of productive participation.
• The owners and controllers of AI capital will tolerate expensive human processes when automated substitutes produce acceptable results.

The most important omission is ownership. The article discusses researchers as users of AI, not as people divided between Sovereigns who control the systems and Servitors who remain indispensable only in narrow domains. That omission turns a power transition into a lifestyle debate.

Social Function

This is a partial truth functioning as transition management, elite self-exoneration, and prestige signaling.

It is not empty copium. Its account of cognitive atrophy, convergence, lost serendipity, and incentive pressure is materially correct. But its prescription allows research leaders to acknowledge the damage without confronting the terminal implication: human collaboration may survive only where someone with control over AI capital chooses to finance it.

The article preserves the self-image of the research class as indispensable stewards who merely need better incentives. It does not ask whether the system still requires most of their labor. Fund the friction is therefore less a solution than hospice care for a declining professional order.

The Verdict

The article is an accurate warning about decollaboration and a weak theory of what follows. It correctly describes how convenience erases invisible social infrastructure, but it assumes that recognizing the loss creates the power to reverse it.

Under P1, AI dominates cognitive production. Under P2, institutions cannot preserve large human-only research domains against competitive pressure. Under P3, productive participation collapses. Human collaboration may persist in protected enclaves, high-status rituals, fieldwork, verification, maintenance, and legitimacy management. It will not remain the default engine of research.

The ride will indeed be smooth. The deeper fact is worse than the article admits: the missing driver is not merely a collaborator. It is the economic necessity of the human being.